notmuch/notmuch-git.py

1157 lines
39 KiB
Python
Raw Normal View History

#!/usr/bin/env python3
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
#
# Copyright (c) 2011-2014 David Bremner <david@tethera.net>
# W. Trevor King <wking@tremily.us>
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see https://www.gnu.org/licenses/ .
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
"""
Manage notmuch tags with Git
"""
from __future__ import print_function
from __future__ import unicode_literals
import codecs as _codecs
import collections as _collections
import functools as _functools
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
import inspect as _inspect
import locale as _locale
import logging as _logging
import os as _os
import re as _re
import shutil as _shutil
import subprocess as _subprocess
import sys as _sys
import tempfile as _tempfile
import textwrap as _textwrap
from urllib.parse import quote as _quote
from urllib.parse import unquote as _unquote
import json as _json
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
_LOG = _logging.getLogger('notmuch-git')
_LOG.setLevel(_logging.WARNING)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
_LOG.addHandler(_logging.StreamHandler())
NOTMUCH_GIT_DIR = None
TAG_PREFIX = None
FORMAT_VERSION = 0
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
_HEX_ESCAPE_REGEX = _re.compile('%[0-9A-F]{2}')
_TAG_DIRECTORY = 'tags/'
_TAG_FILE_REGEX = ( _re.compile(_TAG_DIRECTORY + '(?P<id>[^/]*)/(?P<tag>[^/]*)'),
_re.compile(_TAG_DIRECTORY + '([0-9a-f]{2}/){2}(?P<id>[^/]*)/(?P<tag>[^/]*)'))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
# magic hash for Git (git hash-object -t blob /dev/null)
_EMPTYBLOB = 'e69de29bb2d1d6434b8b29ae775ad8c2e48c5391'
def _hex_quote(string, safe='+@=:,'):
"""
quote('abc def') -> 'abc%20def'.
Wrap urllib.parse.quote with additional safe characters (in
addition to letters, digits, and '_.-') and lowercase hex digits
(e.g. '%3a' instead of '%3A').
"""
uppercase_escapes = _quote(string, safe)
return _HEX_ESCAPE_REGEX.sub(
lambda match: match.group(0).lower(),
uppercase_escapes)
def _xapian_quote(string):
"""
Quote a string for Xapian's QueryParser.
Xapian uses double-quotes for quoting strings. You can escape
internal quotes by repeating them [1,2,3].
[1]: https://trac.xapian.org/ticket/128#comment:2
[2]: https://trac.xapian.org/ticket/128#comment:17
[3]: https://trac.xapian.org/changeset/13823/svn
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
"""
return '"{0}"'.format(string.replace('"', '""'))
def _xapian_unquote(string):
"""
Unquote a Xapian-quoted string.
"""
if string.startswith('"') and string.endswith('"'):
return string[1:-1].replace('""', '"')
return string
def timed(fn):
"""Timer decorator"""
from time import perf_counter
def inner(*args, **kwargs):
start_time = perf_counter()
rval = fn(*args, **kwargs)
end_time = perf_counter()
_LOG.info('{0}: {1:.8f}s elapsed'.format(fn.__name__, end_time - start_time))
return rval
return inner
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
class SubprocessError(RuntimeError):
"A subprocess exited with a nonzero status"
def __init__(self, args, status, stdout=None, stderr=None):
self.status = status
self.stdout = stdout
self.stderr = stderr
msg = '{args} exited with {status}'.format(args=args, status=status)
if stderr:
msg = '{msg}: {stderr}'.format(msg=msg, stderr=stderr)
super(SubprocessError, self).__init__(msg)
class _SubprocessContextManager(object):
"""
PEP 343 context manager for subprocesses.
'expect' holds a tuple of acceptable exit codes, otherwise we'll
raise a SubprocessError in __exit__.
"""
def __init__(self, process, args, expect=(0,)):
self._process = process
self._args = args
self._expect = expect
def __enter__(self):
return self._process
def __exit__(self, type, value, traceback):
for name in ['stdin', 'stdout', 'stderr']:
stream = getattr(self._process, name)
if stream:
stream.close()
setattr(self._process, name, None)
status = self._process.wait()
_LOG.debug(
'collect {args} with status {status} (expected {expect})'.format(
args=self._args, status=status, expect=self._expect))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if status not in self._expect:
raise SubprocessError(args=self._args, status=status)
def wait(self):
return self._process.wait()
def _spawn(args, input=None, additional_env=None, wait=False, stdin=None,
stdout=None, stderr=None, encoding=_locale.getpreferredencoding(),
expect=(0,), **kwargs):
"""Spawn a subprocess, and optionally wait for it to finish.
This wrapper around subprocess.Popen has two modes, depending on
the truthiness of 'wait'. If 'wait' is true, we use p.communicate
internally to write 'input' to the subprocess's stdin and read
from it's stdout/stderr. If 'wait' is False, we return a
_SubprocessContextManager instance for fancier handling
(e.g. piping between processes).
For 'wait' calls when you want to write to the subprocess's stdin,
you only need to set 'input' to your content. When 'input' is not
None but 'stdin' is, we'll automatically set 'stdin' to PIPE
before calling Popen. This avoids having the subprocess
accidentally inherit the launching process's stdin.
"""
_LOG.debug('spawn {args} (additional env. var.: {env})'.format(
args=args, env=additional_env))
if not stdin and input is not None:
stdin = _subprocess.PIPE
if additional_env:
if not kwargs.get('env'):
kwargs['env'] = dict(_os.environ)
kwargs['env'].update(additional_env)
p = _subprocess.Popen(
args, stdin=stdin, stdout=stdout, stderr=stderr, **kwargs)
if wait:
if hasattr(input, 'encode'):
input = input.encode(encoding)
(stdout, stderr) = p.communicate(input=input)
status = p.wait()
_LOG.debug(
'collect {args} with status {status} (expected {expect})'.format(
args=args, status=status, expect=expect))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if stdout is not None:
stdout = stdout.decode(encoding)
if stderr is not None:
stderr = stderr.decode(encoding)
if status not in expect:
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
raise SubprocessError(
args=args, status=status, stdout=stdout, stderr=stderr)
return (status, stdout, stderr)
if p.stdin and not stdin:
p.stdin.close()
p.stdin = None
if p.stdin:
p.stdin = _codecs.getwriter(encoding=encoding)(stream=p.stdin)
stream_reader = _codecs.getreader(encoding=encoding)
if p.stdout:
p.stdout = stream_reader(stream=p.stdout)
if p.stderr:
p.stderr = stream_reader(stream=p.stderr)
return _SubprocessContextManager(args=args, process=p, expect=expect)
def _git(args, **kwargs):
args = ['git', '--git-dir', NOTMUCH_GIT_DIR] + list(args)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
return _spawn(args=args, **kwargs)
def _get_current_branch():
"""Get the name of the current branch.
Return 'None' if we're not on a branch.
"""
try:
(status, branch, stderr) = _git(
args=['symbolic-ref', '--short', 'HEAD'],
stdout=_subprocess.PIPE, stderr=_subprocess.PIPE, wait=True)
except SubprocessError as e:
if 'not a symbolic ref' in e:
return None
raise
return branch.strip()
def _get_remote():
"Get the default remote for the current branch."
local_branch = _get_current_branch()
(status, remote, stderr) = _git(
args=['config', 'branch.{0}.remote'.format(local_branch)],
stdout=_subprocess.PIPE, wait=True)
return remote.strip()
def _tag_query(prefix=None):
if prefix is None:
prefix = TAG_PREFIX
return '(tag (starts-with "{:s}"))'.format(prefix.replace('"','\\\"'))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def count_messages(prefix=None):
"count messages with a given prefix."
(status, stdout, stderr) = _spawn(
args=['notmuch', 'count', '--query=sexp', _tag_query(prefix)],
stdout=_subprocess.PIPE, wait=True)
if status != 0:
_LOG.error("failed to run notmuch config")
sys.exit(1)
return int(stdout.rstrip())
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def get_tags(prefix=None):
"Get a list of tags with a given prefix."
(status, stdout, stderr) = _spawn(
args=['notmuch', 'search', '--query=sexp', '--output=tags', _tag_query(prefix)],
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
stdout=_subprocess.PIPE, wait=True)
return [tag for tag in stdout.splitlines()]
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def archive(treeish='HEAD', args=()):
"""
Dump a tar archive of the current notmuch-git tag set.
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
Using 'git archive'.
Each tag $tag for message with Message-Id $id is written to
an empty file
tags/hash1(id)/hash2(id)/encode($id)/encode($tag)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
The encoding preserves alphanumerics, and the characters
"+-_@=.:," (not the quotes). All other octets are replaced with
'%' followed by a two digit hex number.
"""
_git(args=['archive', treeish] + list(args), wait=True)
def clone(repository):
"""
Create a local notmuch-git repository from a remote source.
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
This wraps 'git clone', adding some options to avoid creating a
working tree while preserving remote-tracking branches and
upstreams.
"""
with _tempfile.TemporaryDirectory(prefix='notmuch-git-clone.') as workdir:
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
_spawn(
args=[
'git', 'clone', '--no-checkout', '--separate-git-dir', NOTMUCH_GIT_DIR,
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
repository, workdir],
wait=True)
nmbug: Accept failures to unset core.worktree in clone Since 6311cfaf (init: do not set unnecessary core.worktree, 2016-09-25, 2.11.0 [1]), Git has no longer set core.worktree when --separate-git-dir is used. This broke clone with: $ nmbug clone http://nmbug.notmuchmail.org/git/nmbug-tags.git Cloning into '/tmp/nmbug-clone.33gg442e'... Checking connectivity: 16674, done. ['git', '--git-dir', '/home/wking/.nmbug', 'config', '--unset', 'core.worktree'] exited with 5 $ echo $? 1 The initial discussion that lead to the Git change is in [2], and there is some more discussion around this specific change in [3]. There is some useful background on working trees in this 2009 message [4]. There is also a git-worktree(1) since df0b6cfb (worktree: new place for "git prune --worktrees", 2015-06-29, 2.5.0 [5]) which grew the ability to add new worktrees in 799767cc (Merge branch 'es/worktree-add', 2015-07-13, 2.5.0 [6]). Folks relying on core.worktree in the --separate-git-dir case fall into the "former case" in [4], and as Junio pointed out in that message, Git operations like 'add' don't really work there. In nmbug we don't want core.worktree, because our effective working tree is the notmuch database. By accepting failed core.worktree unsets, clone will work with Gits older and younger than 2.11.0. [1]: https://github.com/git/git/commit/6311cfaf93716bcc43dd1151cb1763e3f80d8099 [2]: https://public-inbox.org/git/CALqjkKZO_y0DNcRJjooyZ7Eso7yBMGhvZ6fE92oO4Su7JeCeng@mail.gmail.com/ [3]: https://public-inbox.org/git/87h94d8cwi.fsf@kyleam.com/ [4]: https://public-inbox.org/git/7viqbsw2vn.fsf@alter.siamese.dyndns.org/ [5]: https://github.com/git/git/commit/df0b6cfbda88144714541664fb501146d6465a82 [6]: https://github.com/git/git/commit/799767cc98b2f8e6f82d0de4bef9b5e8fcc16e97 Reported-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
2017-10-11 00:49:50 +02:00
_git(args=['config', '--unset', 'core.worktree'], wait=True, expect=(0, 5))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
_git(args=['config', 'core.bare', 'true'], wait=True)
(status, stdout, stderr) = _git(args=['show-ref', '--verify',
'--quiet',
'refs/remotes/origin/config'],
expect=(0,1),
wait=True)
if status == 0:
_git(args=['branch', 'config', 'origin/config'], wait=True)
existing_tags = get_tags()
if existing_tags:
_LOG.warning(
'Not checking out to avoid clobbering existing tags: {}'.format(
', '.join(existing_tags)))
else:
checkout()
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def _is_committed(status):
return len(status['added']) + len(status['deleted']) == 0
class CachedIndex:
def __init__(self, repo, treeish):
self.cache_path = _os.path.join(repo, 'notmuch', 'index_cache.json')
self.index_path = _os.path.join(repo, 'index')
self.current_treeish = treeish
# cached values
self.treeish = None
self.hash = None
self.index_checksum = None
self._load_cache_file()
def _load_cache_file(self):
try:
with open(self.cache_path) as f:
data = _json.load(f)
self.treeish = data['treeish']
self.hash = data['hash']
self.index_checksum = data['index_checksum']
except FileNotFoundError:
pass
except _json.JSONDecodeError:
_LOG.error("Error decoding cache")
_sys.exit(1)
def __enter__(self):
self.read_tree()
return self
def __exit__(self, type, value, traceback):
checksum = _read_index_checksum(self.index_path)
(_, hash, _) = _git(
args=['rev-parse', self.current_treeish],
stdout=_subprocess.PIPE,
wait=True)
with open(self.cache_path, "w") as f:
_json.dump({'treeish': self.current_treeish,
'hash': hash.rstrip(), 'index_checksum': checksum }, f)
@timed
def read_tree(self):
current_checksum = _read_index_checksum(self.index_path)
(_, hash, _) = _git(
args=['rev-parse', self.current_treeish],
stdout=_subprocess.PIPE,
wait=True)
current_hash = hash.rstrip()
if self.current_treeish == self.treeish and \
self.index_checksum and self.index_checksum == current_checksum and \
self.hash and self.hash == current_hash:
return
_git(args=['read-tree', self.current_treeish], wait=True)
def check_safe_fraction(status):
safe = 0.1
conf = _notmuch_config_get ('git.safe_fraction')
if conf and conf != '':
safe=float(conf)
total = count_messages (TAG_PREFIX)
if total == 0:
_LOG.error('No existing tags with given prefix, stopping.'.format(safe))
_LOG.error('Use --force to override.')
exit(1)
change = len(status['added'])+len(status['deleted'])
fraction = change/total
_LOG.debug('total messages {:d}, change: {:d}, fraction: {:f}'.format(total,change,fraction))
if fraction > safe:
_LOG.error('safe fraction {:f} exceeded, stopping.'.format(safe))
_LOG.error('Use --force to override or reconfigure git.safe_fraction.')
exit(1)
def commit(treeish='HEAD', message=None, force=False):
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
"""
Commit prefix-matching tags from the notmuch database to Git.
"""
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
status = get_status()
if _is_committed(status=status):
_LOG.warning('Nothing to commit')
return
if not force:
check_safe_fraction (status)
with CachedIndex(NOTMUCH_GIT_DIR, treeish) as index:
try:
_update_index(status=status)
(_, tree, _) = _git(
args=['write-tree'],
stdout=_subprocess.PIPE,
wait=True)
(_, parent, _) = _git(
args=['rev-parse', treeish],
stdout=_subprocess.PIPE,
wait=True)
(_, commit, _) = _git(
args=['commit-tree', tree.strip(), '-p', parent.strip()],
input=message,
stdout=_subprocess.PIPE,
wait=True)
_git(
args=['update-ref', treeish, commit.strip()],
stdout=_subprocess.PIPE,
wait=True)
except Exception as e:
_git(args=['read-tree', '--empty'], wait=True)
_git(args=['read-tree', treeish], wait=True)
raise
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
@timed
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def _update_index(status):
with _git(
args=['update-index', '--index-info'],
stdin=_subprocess.PIPE) as p:
for id, tags in status['deleted'].items():
for line in _index_tags_for_message(id=id, status='D', tags=tags):
p.stdin.write(line)
for id, tags in status['added'].items():
for line in _index_tags_for_message(id=id, status='A', tags=tags):
p.stdin.write(line)
def fetch(remote=None):
"""
Fetch changes from the remote repository.
See 'merge' to bring those changes into notmuch.
"""
args = ['fetch']
if remote:
args.append(remote)
_git(args=args, wait=True)
def init(remote=None):
"""
Create an empty notmuch-git repository.
This wraps 'git init' with a few extra steps to support subsequent
status and commit commands.
"""
from pathlib import Path
parent = Path(NOTMUCH_GIT_DIR).parent
try:
_os.makedirs(parent)
except FileExistsError:
pass
_spawn(args=['git', '--git-dir', NOTMUCH_GIT_DIR, 'init',
'--initial-branch=master', '--quiet', '--bare'], wait=True)
_git(args=['config', 'core.logallrefupdates', 'true'], wait=True)
# create an empty blob (e69de29bb2d1d6434b8b29ae775ad8c2e48c5391)
_git(args=['hash-object', '-w', '--stdin'], input='', wait=True)
# create a blob for the FORMAT file
(status, stdout, _) = _git(args=['hash-object', '-w', '--stdin'], stdout=_subprocess.PIPE,
input='1\n', wait=True)
verhash=stdout.rstrip()
_LOG.debug('hash of FORMAT blob = {:s}'.format(verhash))
# Add FORMAT to the index
_git(args=['update-index', '--add', '--cacheinfo', '100644,{:s},FORMAT'.format(verhash)], wait=True)
_git(
args=[
'commit', '-m', 'Start a new notmuch-git repository'
],
additional_env={'GIT_WORK_TREE': NOTMUCH_GIT_DIR},
wait=True)
def checkout(force=None):
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
"""
Update the notmuch database from Git.
This is mainly useful to discard your changes in notmuch relative
to Git.
"""
status = get_status()
if not force:
check_safe_fraction(status)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
with _spawn(
args=['notmuch', 'tag', '--batch'], stdin=_subprocess.PIPE) as p:
for id, tags in status['added'].items():
p.stdin.write(_batch_line(action='-', id=id, tags=tags))
for id, tags in status['deleted'].items():
p.stdin.write(_batch_line(action='+', id=id, tags=tags))
def _batch_line(action, id, tags):
"""
'notmuch tag --batch' line for adding/removing tags.
Set 'action' to '-' to remove a tag or '+' to add the tags to a
given message id.
"""
tag_string = ' '.join(
'{action}{prefix}{tag}'.format(
action=action, prefix=_ENCODED_TAG_PREFIX, tag=_hex_quote(tag))
for tag in tags)
line = '{tags} -- id:{id}\n'.format(
tags=tag_string, id=_xapian_quote(string=id))
return line
def _insist_committed():
"Die if the the notmuch tags don't match the current HEAD."
status = get_status()
if not _is_committed(status=status):
_LOG.error('\n'.join([
'Uncommitted changes to {prefix}* tags in notmuch',
'',
"For a summary of changes, run 'notmuch-git status'",
"To save your changes, run 'notmuch-git commit' before merging/pull",
"To discard your changes, run 'notmuch-git checkout'",
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
]).format(prefix=TAG_PREFIX))
_sys.exit(1)
def pull(repository=None, refspecs=None):
"""
Pull (merge) remote repository changes to notmuch.
'pull' is equivalent to 'fetch' followed by 'merge'. We use the
Git-configured repository for your current branch
(branch.<name>.repository, likely 'origin', and
branch.<name>.merge, likely 'master').
"""
_insist_committed()
if refspecs and not repository:
repository = _get_remote()
args = ['pull']
if repository:
args.append(repository)
if refspecs:
args.extend(refspecs)
with _tempfile.TemporaryDirectory(prefix='notmuch-git-pull.') as workdir:
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
for command in [
['reset', '--hard'],
args]:
_git(
args=command,
additional_env={'GIT_WORK_TREE': workdir},
wait=True)
checkout()
def merge(reference='@{upstream}'):
"""
Merge changes from 'reference' into HEAD and load the result into notmuch.
The default reference is '@{upstream}'.
"""
_insist_committed()
with _tempfile.TemporaryDirectory(prefix='notmuch-git-merge.') as workdir:
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
for command in [
['reset', '--hard'],
['merge', reference]]:
_git(
args=command,
additional_env={'GIT_WORK_TREE': workdir},
wait=True)
checkout()
def log(args=()):
"""
A simple wrapper for 'git log'.
After running 'notmuch-git fetch', you can inspect the changes with
'notmuch-git log HEAD..@{upstream}'.
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
"""
# we don't want output trapping here, because we want the pager.
args = ['log', '--name-status', '--no-renames'] + list(args)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
with _git(args=args, expect=(0, 1, -13)) as p:
p.wait()
def push(repository=None, refspecs=None):
"Push the local notmuch-git Git state to a remote repository."
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if refspecs and not repository:
repository = _get_remote()
args = ['push']
if repository:
args.append(repository)
if refspecs:
args.extend(refspecs)
_git(args=args, wait=True)
def status():
"""
Show pending updates in notmuch or git repo.
Prints lines of the form
ng Message-Id tag
where n is a single character representing notmuch database status
* A
Tag is present in notmuch database, but not committed to notmuch-git
(equivalently, tag has been deleted in notmuch-git repo, e.g. by a
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
pull, but not restored to notmuch database).
* D
Tag is present in notmuch-git repo, but not restored to notmuch
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
database (equivalently, tag has been deleted in notmuch).
* U
Message is unknown (missing from local notmuch database).
The second character (if present) represents a difference between
local and upstream branches. Typically 'notmuch-git fetch' needs to be
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
run to update this.
* a
Tag is present in upstream, but not in the local Git branch.
* d
Tag is present in local Git branch, but not upstream.
"""
status = get_status()
# 'output' is a nested defaultdict for message status:
# * The outer dict is keyed by message id.
# * The inner dict is keyed by tag name.
# * The inner dict values are status strings (' a', 'Dd', ...).
output = _collections.defaultdict(
lambda : _collections.defaultdict(lambda : ' '))
for id, tags in status['added'].items():
for tag in tags:
output[id][tag] = 'A'
for id, tags in status['deleted'].items():
for tag in tags:
output[id][tag] = 'D'
for id, tags in status['missing'].items():
for tag in tags:
output[id][tag] = 'U'
if _is_unmerged():
for id, tag in _diff_refs(filter='A'):
output[id][tag] += 'a'
for id, tag in _diff_refs(filter='D'):
output[id][tag] += 'd'
for id, tag_status in sorted(output.items()):
for tag, status in sorted(tag_status.items()):
print('{status}\t{id}\t{tag}'.format(
status=status, id=id, tag=tag))
def _is_unmerged(ref='@{upstream}'):
try:
(status, fetch_head, stderr) = _git(
args=['rev-parse', ref],
stdout=_subprocess.PIPE, stderr=_subprocess.PIPE, wait=True)
except SubprocessError as e:
if 'No upstream configured' in e.stderr:
return
raise
(status, base, stderr) = _git(
args=['merge-base', 'HEAD', ref],
stdout=_subprocess.PIPE, wait=True)
return base != fetch_head
@timed
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def get_status():
status = {
'deleted': {},
'missing': {},
}
with PrivateIndex(repo=NOTMUCH_GIT_DIR, prefix=TAG_PREFIX) as index:
maybe_deleted = index.diff(filter='D')
for id, tags in maybe_deleted.items():
(_, stdout, stderr) = _spawn(
args=['notmuch', 'search', '--output=files', 'id:{0}'.format(id)],
stdout=_subprocess.PIPE,
wait=True)
if stdout:
status['deleted'][id] = tags
else:
status['missing'][id] = tags
status['added'] = index.diff(filter='A')
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
return status
class PrivateIndex:
def __init__(self, repo, prefix):
try:
_os.makedirs(_os.path.join(repo, 'notmuch'))
except FileExistsError:
pass
file_name = 'notmuch/index'
self.index_path = _os.path.join(repo, file_name)
self.cache_path = _os.path.join(repo, 'notmuch', '{:s}.json'.format(_hex_quote(file_name)))
self.current_prefix = prefix
self.prefix = None
self.uuid = None
self.lastmod = None
self.checksum = None
self._load_cache_file()
self._index_tags()
def __enter__(self):
return self
def __exit__(self, type, value, traceback):
checksum = _read_index_checksum(self.index_path)
(count, uuid, lastmod) = _read_database_lastmod()
with open(self.cache_path, "w") as f:
_json.dump({'prefix': self.current_prefix, 'uuid': uuid, 'lastmod': lastmod, 'checksum': checksum }, f)
def _load_cache_file(self):
try:
with open(self.cache_path) as f:
data = _json.load(f)
self.prefix = data['prefix']
self.uuid = data['uuid']
self.lastmod = data['lastmod']
self.checksum = data['checksum']
except FileNotFoundError:
return None
except _json.JSONDecodeError:
_LOG.error("Error decoding cache")
_sys.exit(1)
@timed
def _index_tags(self):
"Write notmuch tags to private git index."
prefix = '+{0}'.format(_ENCODED_TAG_PREFIX)
current_checksum = _read_index_checksum(self.index_path)
if (self.prefix == None or self.prefix != self.current_prefix
or self.checksum == None or self.checksum != current_checksum):
_git(
args=['read-tree', '--empty'],
additional_env={'GIT_INDEX_FILE': self.index_path}, wait=True)
query = _tag_query()
clear_tags = False
(count,uuid,lastmod) = _read_database_lastmod()
if self.prefix == self.current_prefix and self.uuid \
and self.uuid == uuid and self.checksum == current_checksum:
query = '(and (infix "lastmod:{:d}..")) {:s})'.format(self.lastmod+1, query)
clear_tags = True
with _spawn(
args=['notmuch', 'dump', '--format=batch-tag', '--query=sexp', '--', query],
stdout=_subprocess.PIPE) as notmuch:
with _git(
args=['update-index', '--index-info'],
stdin=_subprocess.PIPE,
additional_env={'GIT_INDEX_FILE': self.index_path}) as git:
for line in notmuch.stdout:
if line.strip().startswith('#'):
continue
(tags_string, id) = [_.strip() for _ in line.split(' -- id:')]
tags = [
_unquote(tag[len(prefix):])
for tag in tags_string.split()
if tag.startswith(prefix)]
id = _xapian_unquote(string=id)
if clear_tags:
for line in _clear_tags_for_message(index=self.index_path, id=id):
git.stdin.write(line)
for line in _index_tags_for_message(
id=id, status='A', tags=tags):
git.stdin.write(line)
@timed
def diff(self, filter):
"""
Get an {id: {tag, ...}} dict for a given filter.
For example, use 'A' to find added tags, and 'D' to find deleted tags.
"""
s = _collections.defaultdict(set)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
with _git(
args=[
'diff-index', '--cached', '--diff-filter', filter,
'--name-only', 'HEAD'],
additional_env={'GIT_INDEX_FILE': self.index_path},
stdout=_subprocess.PIPE) as p:
# Once we drop Python < 3.3, we can use 'yield from' here
for id, tag in _unpack_diff_lines(stream=p.stdout):
s[id].add(tag)
return s
def _read_index_checksum (index_path):
"""Read the index checksum, as defined by index-format.txt in the git source
WARNING: assumes SHA1 repo"""
import binascii
try:
with open(index_path, 'rb') as f:
size=_os.path.getsize(index_path)
f.seek(size-20);
return binascii.hexlify(f.read(20)).decode('ascii')
except FileNotFoundError:
return None
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def _clear_tags_for_message(index, id):
"""
Clear any existing index entries for message 'id'
Neither 'id' nor the tags in 'tags' should be encoded/escaped.
"""
dir = _id_path(id)
with _git(
args=['ls-files', dir],
additional_env={'GIT_INDEX_FILE': index},
stdout=_subprocess.PIPE) as git:
for file in git.stdout:
line = '0 0000000000000000000000000000000000000000\t{:s}\n'.format(file.strip())
yield line
def _read_database_lastmod():
with _spawn(
args=['notmuch', 'count', '--lastmod', '*'],
stdout=_subprocess.PIPE) as notmuch:
(count,uuid,lastmod_str) = notmuch.stdout.readline().split()
return (count,uuid,int(lastmod_str))
def _id_path(id):
hid=_hex_quote(string=id)
from hashlib import blake2b
if FORMAT_VERSION==0:
return 'tags/{hid}'.format(hid=hid)
elif FORMAT_VERSION==1:
idhash = blake2b(hid.encode('utf8'), digest_size=2).hexdigest()
return 'tags/{dir1}/{dir2}/{hid}'.format(
hid=hid,
dir1=idhash[0:2],dir2=idhash[2:])
else:
_LOG.error("Unknown format version",FORMAT_VERSION)
_sys.exit(1)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
def _index_tags_for_message(id, status, tags):
"""
Update the Git index to either create or delete an empty file.
Neither 'id' nor the tags in 'tags' should be encoded/escaped.
"""
mode = '100644'
hash = _EMPTYBLOB
if status == 'D':
mode = '0'
hash = '0000000000000000000000000000000000000000'
for tag in tags:
path = '{ipath}/{tag}'.format(ipath=_id_path(id),tag=_hex_quote(string=tag))
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
yield '{mode} {hash}\t{path}\n'.format(mode=mode, hash=hash, path=path)
def _diff_refs(filter, a='HEAD', b='@{upstream}'):
with _git(
args=['diff', '--diff-filter', filter, '--name-only', a, b],
stdout=_subprocess.PIPE) as p:
# Once we drop Python < 3.3, we can use 'yield from' here
for id, tag in _unpack_diff_lines(stream=p.stdout):
yield id, tag
def _unpack_diff_lines(stream):
"Iterate through (id, tag) tuples in a diff stream."
for line in stream:
match = _TAG_FILE_REGEX[FORMAT_VERSION].match(line.strip())
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if not match:
message = 'non-tag line in diff: {!r}'.format(line.strip())
if line.startswith(_TAG_DIRECTORY):
raise ValueError(message)
_LOG.info(message)
continue
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
id = _unquote(match.group('id'))
tag = _unquote(match.group('tag'))
yield (id, tag)
def _help(parser, command=None):
"""
Show help for an notmuch-git command.
Because some folks prefer:
$ notmuch-git help COMMAND
to
$ notmuch-git COMMAND --help
"""
if command:
parser.parse_args([command, '--help'])
else:
parser.parse_args(['--help'])
def _notmuch_config_get(key):
(status, stdout, stderr) = _spawn(
args=['notmuch', 'config', 'get', key],
stdout=_subprocess.PIPE, wait=True)
if status != 0:
_LOG.error("failed to run notmuch config")
_sys.exit(1)
return stdout.rstrip()
def read_format_version():
try:
(status, stdout, stderr) = _git(
args=['cat-file', 'blob', 'master:FORMAT'],
stdout=_subprocess.PIPE, stderr=_subprocess.PIPE, wait=True)
except SubprocessError as e:
_LOG.debug("failed to read FORMAT file from git, assuming format version 0")
return 0
return int(stdout)
# based on BaseDirectory.save_data_path from pyxdg (LGPL2+)
def xdg_data_path(profile):
resource = _os.path.join('notmuch',profile,'git')
assert not resource.startswith('/')
_home = _os.path.expanduser('~')
xdg_data_home = _os.environ.get('XDG_DATA_HOME') or \
_os.path.join(_home, '.local', 'share')
path = _os.path.join(xdg_data_home, resource)
return path
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser(
description=__doc__.strip(),
formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument(
'-C', '--git-dir', metavar='REPO',
help='Git repository to operate on.')
parser.add_argument(
'-p', '--tag-prefix', metavar='PREFIX',
default = None,
help='Prefix of tags to operate on.')
parser.add_argument(
'-N', '--nmbug', action='store_true',
help='Set defaults for --tag-prefix and --git-dir for the notmuch bug tracker')
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
parser.add_argument(
'-l', '--log-level',
choices=['critical', 'error', 'warning', 'info', 'debug'],
help='Log verbosity. Defaults to {!r}.'.format(
_logging.getLevelName(_LOG.level).lower()))
help = _functools.partial(_help, parser=parser)
help.__doc__ = _help.__doc__
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
subparsers = parser.add_subparsers(
title='commands',
description=(
'For help on a particular command, run: '
"'%(prog)s ... <command> --help'."))
for command in [
'archive',
'checkout',
'clone',
'commit',
'fetch',
'help',
'init',
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
'log',
'merge',
'pull',
'push',
'status',
]:
func = locals()[command]
doc = _textwrap.dedent(func.__doc__).strip().replace('%', '%%')
subparser = subparsers.add_parser(
command,
help=doc.splitlines()[0],
description=doc,
formatter_class=argparse.RawDescriptionHelpFormatter)
subparser.set_defaults(func=func)
if command == 'archive':
subparser.add_argument(
'treeish', metavar='TREE-ISH', nargs='?', default='HEAD',
help=(
'The tree or commit to produce an archive for. Defaults '
"to 'HEAD'."))
subparser.add_argument(
'args', metavar='ARG', nargs='*',
help=(
"Argument passed through to 'git archive'. Set anything "
'before <tree-ish>, see git-archive(1) for details.'))
elif command == 'checkout':
subparser.add_argument(
'-f', '--force', action='store_true',
help='checkout a large fraction of tags.')
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
elif command == 'clone':
subparser.add_argument(
'repository',
help=(
'The (possibly remote) repository to clone from. See the '
'URLS section of git-clone(1) for more information on '
'specifying repositories.'))
elif command == 'commit':
subparser.add_argument(
'-f', '--force', action='store_true',
help='commit a large fraction of tags.')
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
subparser.add_argument(
'message', metavar='MESSAGE', default='', nargs='?',
help='Text for the commit message.')
elif command == 'fetch':
subparser.add_argument(
'remote', metavar='REMOTE', nargs='?',
help=(
'Override the default configured in branch.<name>.remote '
'to fetch from a particular remote repository (e.g. '
"'origin')."))
elif command == 'help':
subparser.add_argument(
'command', metavar='COMMAND', nargs='?',
help='The command to show help for.')
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
elif command == 'log':
subparser.add_argument(
'args', metavar='ARG', nargs='*',
help="Additional argument passed through to 'git log'.")
elif command == 'merge':
subparser.add_argument(
'reference', metavar='REFERENCE', default='@{upstream}',
nargs='?',
help=(
'Reference, usually other branch heads, to merge into '
"our branch. Defaults to '@{upstream}'."))
elif command == 'pull':
subparser.add_argument(
'repository', metavar='REPOSITORY', default=None, nargs='?',
help=(
'The "remote" repository that is the source of the pull. '
'This parameter can be either a URL (see the section GIT '
'URLS in git-pull(1)) or the name of a remote (see the '
'section REMOTES in git-pull(1)).'))
subparser.add_argument(
'refspecs', metavar='REFSPEC', default=None, nargs='*',
help=(
'Refspec (usually a branch name) to fetch and merge. See '
'the <refspec> entry in the OPTIONS section of '
'git-pull(1) for other possibilities.'))
elif command == 'push':
subparser.add_argument(
'repository', metavar='REPOSITORY', default=None, nargs='?',
help=(
'The "remote" repository that is the destination of the '
'push. This parameter can be either a URL (see the '
'section GIT URLS in git-push(1)) or the name of a remote '
'(see the section REMOTES in git-push(1)).'))
subparser.add_argument(
'refspecs', metavar='REFSPEC', default=None, nargs='*',
help=(
'Refspec (usually a branch name) to push. See '
'the <refspec> entry in the OPTIONS section of '
'git-push(1) for other possibilities.'))
args = parser.parse_args()
nmbug_mode = False
notmuch_profile = _os.getenv('NOTMUCH_PROFILE','default')
if args.nmbug or _os.path.basename(__file__) == 'nmbug':
nmbug_mode = True
if args.git_dir:
NOTMUCH_GIT_DIR = args.git_dir
else:
if nmbug_mode:
default = _os.path.join('~', '.nmbug')
else:
default = _notmuch_config_get ('git.path')
if default == '':
default = xdg_data_path(notmuch_profile)
NOTMUCH_GIT_DIR = _os.path.expanduser(_os.getenv('NOTMUCH_GIT_DIR', default))
_NOTMUCH_GIT_DIR = _os.path.join(NOTMUCH_GIT_DIR, '.git')
if _os.path.isdir(_NOTMUCH_GIT_DIR):
NOTMUCH_GIT_DIR = _NOTMUCH_GIT_DIR
if args.tag_prefix:
TAG_PREFIX = args.tag_prefix
else:
if nmbug_mode:
prefix = 'notmuch::'
else:
prefix = _notmuch_config_get ('git.tag_prefix')
TAG_PREFIX = _os.getenv('NOTMUCH_GIT_PREFIX', prefix)
_ENCODED_TAG_PREFIX = _hex_quote(TAG_PREFIX, safe='+@=,') # quote ':'
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if args.log_level:
level = getattr(_logging, args.log_level.upper())
_LOG.setLevel(level)
# for test suite
for var in ['NOTMUCH_GIT_DIR', 'NOTMUCH_GIT_PREFIX', 'NOTMUCH_PROFILE', 'NOTMUCH_CONFIG' ]:
_LOG.debug('env {:s} = {:s}'.format(var, _os.getenv(var,'%None%')))
if _notmuch_config_get('built_with.sexp_queries') != 'true':
_LOG.error("notmuch git needs sexp query support")
_sys.exit(1)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
if not getattr(args, 'func', None):
parser.print_usage()
_sys.exit(1)
# The following two lines are used by the test suite.
_LOG.debug('prefix = {:s}'.format(TAG_PREFIX))
_LOG.debug('repository = {:s}'.format(NOTMUCH_GIT_DIR))
FORMAT_VERSION = read_format_version()
_LOG.debug('FORMAT_VERSION={:d}'.format(FORMAT_VERSION))
if args.func == help:
arg_names = ['command']
else:
(arg_names, varargs, varkw) = _inspect.getargs(args.func.__code__)
nmbug: Translate to Python This allows us to capture stdout and stderr separately, and do other explicit subprocess manipulation without resorting to external packages. It should be compatible with Python 2.7 and later (including the 3.x series). Most of the user-facing interface is the same, but there are a few changes, where reproducing the original interface was too difficult or I saw a change to make the underlying Git UI accessible: * 'nmbug help' has been split between the general 'nmbug --help' and the command-specific 'nmbug COMMAND --help'. * Commands are no longer split into "most common", "other useful", and "less common" sets. If we need something like this, I'd prefer workflow examples highlighting common commands in the module docstring (available with 'nmbug --help'). * 'nmbug commit' now only uses a single argument for the optional commit-message text. I wanted to expose more of the underlying 'git commit' UI, since I personally like to write my commit messages in an editor with the notes added by 'git commit -v' to jog my memory. Unfortunately, we're using 'git commit-tree' instead of 'git commit', and commit-tree is too low-level for editor-launching. I'd be interested in rewriting commit() to use 'git commit', but that seemed like it was outside the scope of this rewrite. So I'm not supporting all of Git's commit syntax in this patch, but I can at least match 'git commit -m MESSAGE' in requiring command-line commit messages to be a single argument. * The default repository for 'nmbug push' and 'nmbug fetch' is now the current branch's upstream (branch.<name>.remote) instead of 'origin'. When we have to, we extract this remote by hand, but where possible we just call the Git command without a repository argument, and leave it to Git to figure out the default. * 'nmbug push' accepts multiple refspecs if you want to explicitly specify what to push. Otherwise, the refspec(s) pushed depend on push.default. The Perl version hardcoded 'master' as the pushed refspec. * 'nmbug pull' defaults to the current branch's upstream (branch.<name>.remote and branch.<name>.merge) instead of hardcoding 'origin' and 'master'. It also supports multiple refspecs if for some crazy reason you need an octopus merge (but mostly to avoid breaking consistency with 'git pull'). * 'nmbug log' now execs 'git log', as there's no need to keep the Python process around once we've launched Git there. * 'nmbug status' now catches stderr, and doesn't print errors like: No upstream configured for branch 'master' The Perl implementation had just learned to avoid crashing on that case, but wasn't yet catching the dying subprocess's stderr. * 'nmbug archive' now accepts positional arguments for the tree-ish and additional 'git archive' options. For example, you can run: $ nmbug archive HEAD -- --format tar.gz I wish I could have preserved the argument order from 'git archive' (with the tree-ish at the end), but I'm not sure how to make argparse accept arbitrary possitional arguments (some of which take arguments). Flipping the order to put the tree-ish first seemed easiest. * 'nmbug merge' and 'pull' no longer checkout HEAD before running their command, because blindly clobbering the index seems overly risky. * In order to avoid creating a dirty index, 'nmbug commit' now uses the default index (instead of nmbug.index) for composing the commit. That way the index matches the committed tree. To avoid leaving a broken index after a failed commit, I've wrapped the whole thing in a try/except block that resets the index to match the pre-commit treeish on errors. That means that 'nmbug commit' will ignore anything you've cached in the index via direct Git calls, and you'll either end up with an index matching your notmuch tags and the new HEAD (after a successful commit) or an index matching the original HEAD (after a failed commit).
2014-10-03 20:20:57 +02:00
kwargs = {key: getattr(args, key) for key in arg_names if key in args}
try:
args.func(**kwargs)
except SubprocessError as e:
if _LOG.level == _logging.DEBUG:
raise # don't mask the traceback
_LOG.error(str(e))
_sys.exit(1)