This change addresses two known issues with large sets of changes to
the database. The first is that as reported by Steven Allen [1],
notmuch commits are not "flushed" when they complete, which means that
if there is an open transaction when the database closes (or e.g. the
program crashes) then all changes since the last commit will be
discarded (nothing is irrecoverably lost for "notmuch new", as the
indexing process just restarts next time it is run). This does not
really "fix" the issue reported in [1]; that seems rather difficult
given how transactions work in Xapian. On the other hand, with the
default settings, this should mean one only loses less than a minutes
worth of work. The second issue is the occasionally reported "storm"
of disk writes when notmuch finishes. I don't yet have a test for
this, but I think committing as we go should reduce the amount of work
when finalizing the database.
[1]: id:20151025210215.GA3754@stebalien.com
Since the library searches in several locations for a config file, the
caller does not know which of these is chosen in the usual case of
passing NULL as a config file. This changes provides an API for the
caller to retrieve the name of the config file chosen. It will be
tested in a following commit.
This removes duplication between the struct element and the
configuration string_map entry. Create a simple wrapper for setting
the database path that makes sure the trailing / is stripped.
This will allow re-opening in a different mode (read/write
vs. read-only) with current Xapian API. It will also prove useful when
updating the compact functions to support more flexible database
location.
The main goal is to allow configuration information to be temporarily
overridden by a separate config file. That will require further
changes not in this commit.
The performance impact is unclear, and will depend on the balance
between number of queries and number of distinct metadata items read
on the first call to n_d_get_config.
database.cc is uncomfortably large, and some of the static data
structures do not need to be shared as much as they are.
This is a somewhat small piece to factor out, but it will turn out to
be helpful to further refactoring.
static_cast is a bit tricky to understand and error prone, so add a
second pointer to (potentially the same) Xapian database object that
we know has the right subclass.
In order to mimic the "best effort" API of Xapian to provide
information from a closed database when possible, do not
destroy the Xapian database object too early.
Because the pointer to a Xapian database is no longer nulled on close,
introduce a flag to track whether the notmuch database is open or not.
This will be mandatory as of Xapian 1.5. The API is also more
consistent with the FieldProcessor API, which helps code re-use a bit.
Note that this switches to using the built-in Xapian support for
prefixes on ranges (i.e. deleted code at beginning of
ParseTimeRangeProcessor::operator(), added prefix to constructor).
Another side effect of the migration is that we are generating smaller
queries, using one OP_VALUE_RANGE instead of an AND of two OP_VALUE_*
queries.
This will be used to avoid needing a database access to resolve a db
prefix from the corresponding UI prefix (e.g. when indexing). Arguably
the setup of the separate header map does not belong here, since it is
about indexing rather than querying, but we currently don't have any
other indexing setup to do.
The new `body:` field (in Xapian terms) or prefix (in slightly
sloppier notmuch) terms allows matching terms that occur only in the
body.
Unprefixed query terms should continue to match anywhere (header or
body) in the message.
This follows a suggestion of Olly Betts to use the facility (since
Xapian 1.0.4) to add the same field with multiple prefixes. The double
indexing of previous versions is thus replaced with a query time
expension of unprefixed query terms to the various prefixed
equivalent.
Reindexing will be needed for 'body:' searches to work correctly;
otherwise they will also match messages where the term occur in
headers (demonstrated by the new tests in T530-upgrade.sh)
C99 stdbool turned 18 this year. There really is no reason to use our
own, except in the library interface for backward
compatibility. Convert the lib internally to stdbool.
'database.cc' is becoming a monster, and it's hard to follow what the
various static functions are used for. It turns out that about 1/3 of
this file notmuch_database_add_message and helper functions not used
by any other function. This commit isolates this code into it's own
file.
Some side effects of this refactoring:
- find_doc_ids becomes the non-static (but still private)
_notmuch_database_find_doc_ids
- a few instances of 'string' have 'std::' prepended, avoiding the
need for 'using namespace std;' in the new file.
The dynamic generation of the linker version script for libnotmuch
exports has grown rather complicated.
Reverse the visibility control by hiding symbols by default using
-fvisibility=hidden, and explicitly exporting symbols in notmuch.h
using #pragma GCC visibility. (We could also use __attribute__
((visibility ("default"))) for each exported function, but the pragma
is more convenient.)
The above is not quite enough alone, as it would "leak" a number of
weak symbols from Xapian and C++ standard library. Combine it with a
small static version script that filters out everything except the
notmuch_* symbols that we explicitly exposed, and the C++ RTTI
typeinfo symbols for exception handling.
Finally, as the symbol hiding test can no longer look at the generated
symbol table, switch the test to parse the functions from notmuch.h.
Replace multiple tables with some flags in a single table. This makes
the code in notmuch_database_open_verbose a bit shorter, and it should
also make it easier to add other options to fields, e.g. regexp
searching.
Many of the external links found in the notmuch source can be resolved
using https instead of http. This changeset addresses as many as i
could find, without touching the e-mail corpus or expected outputs
found in tests.
It's already kindof gross that this is hardcoded in two different
places. We will also need these later in field processors calling back
into the query parser.
This exposes the committed database revision to library users along
with a UUID that can be used to detect when revision numbers are no
longer comparable (e.g., because the database has been replaced).
This adds a new document value that stores the revision of the last
modification to message metadata, where the revision number increases
monotonically with each database commit.
An alternative would be to store the wall-clock time of the last
modification of each message. In principle this is simpler and has
the advantage that any process can determine the current timestamp
without support from libnotmuch. However, even assuming a computer's
clock never goes backward and ignoring clock skew in networked
environments, this has a fatal flaw. Xapian uses (optimistic)
snapshot isolation, which means reads can be concurrent with writes.
Given this, consider the following time line with a write and two read
transactions:
write |-X-A--------------|
read 1 |---B---|
read 2 |---|
The write transaction modifies message X and records the wall-clock
time of the modification at A. The writer hangs around for a while
and later commits its change. Read 1 is concurrent with the write, so
it doesn't see the change to X. It does some query and records the
wall-clock time of its results at B. Transaction read 2 later starts
after the write commits and queries for changes since wall-clock time
B (say the reads are performing an incremental backup). Even though
read 1 could not see the change to X, read 2 is told (correctly) that
X has not changed since B, the time of the last read. In fact, X
changed before wall-clock time A, but the change was not visible until
*after* wall-clock time B, so read 2 misses the change to X.
This is tricky to solve in full-blown snapshot isolation, but because
Xapian serializes writes, we can use a simple, monotonically
increasing database revision number. Furthermore, maintaining this
revision number requires no more IO than a wall-clock time solution
because Xapian already maintains statistics on the upper (and lower)
bound of each value stream.
This feature will exist in all newly created databases, but there is
no upgrade provided for it. If this flag exists, it indicates that
the database was created after the indexed MIME-types feature was
added.
Previously, our database schema was versioned by a single number.
Each database schema change had to occur "atomically" in Notmuch's
development history: before some commit, Notmuch used version N, after
that commit, it used version N+1. Hence, each new schema version
could introduce only one change, the task of developing a schema
change fell on a single person, and it all had to happen and be
perfect in a single commit series. This made introducing a new schema
version hard. We've seen only two schema changes in the history of
Notmuch.
This commit introduces database schema version 3; hopefully the last
schema version we'll need for a while. With this version, we switch
from a single version number to "features": a set of named,
independent aspects of the database schema.
Features should make backwards compatibility easier. For many things,
it should be easy to support databases both with and without a
feature, which will allow us to make upgrades optional and will enable
"unstable" features that can be developed and tested over time.
Features also make forwards compatibility easier. The features
recorded in a database include "compatibility flags," which can
indicate to an older version of Notmuch when it must support a given
feature to open the database for read or for write. This lets us
replace the old vague "I don't recognize this version, so something
might go wrong, but I promise to try my best" warnings upon opening a
database with an unknown version with precise errors. If a database
is safe to open for read/write despite unknown features, an older
version will know that and issue no message at all. If the database
is not safe to open for read/write because of unknown features, an
older version will know that, too, and can tell the user exactly which
required features it lacks support for.
Add a custom value range processor to enable date and time searches of
the form date:since..until, where "since" and "until" are expressions
understood by the previously added date/time parser, to restrict the
results to messages within a particular time range (based on the Date:
header).
If "since" or "until" describes date/time at an accuracy of days or
less, the values are rounded according to the accuracy, towards past
for "since" and towards future for "until". For example,
date:november..yesterday would match from the beginning of November
until the end of yesterday. Expressions such as date:today..today
means since the beginning of today until the end of today.
Open-ended ranges are supported (since Xapian 1.2.1), i.e. you can
specify date:..until or date:since.. to not limit the start or end
date, respectively.
CAVEATS:
Xapian does not support spaces in range expressions. You can replace
the spaces with '_', or (in most cases) '-', or (in some cases) leave
the spaces out altogether.
Entering date:expr without ".." (for example date:yesterday) will not
work as you might expect. You can achieve the expected result by
duplicating the expr both sides of ".." (for example
date:yesterday..yesterday).
Open-ended ranges won't work with pre-1.2.1 Xapian, but they don't
produce an error either.
Signed-off-by: Jani Nikula <jani@nikula.org>
notmuch_database_t now keeps a nesting count and we only start a
transaction or commit for the outermost atomic section.
Introduces a new error, NOTMUCH_STATUS_UNBALANCED_ATOMIC.
Replace _notmuch_convert_tags with this and simplify
_create_filenames_for_terms_with_prefix. This will also come in handy
shortly to get the message file name list.
Instead of having an API for setting a library-wide flag for
synchronization (notmuch_database_set_maildir_sync) we instead
implement maildir synchronization with two new library functions:
notmuch_message_maildir_flags_to_tags
and notmuch_message_tags_to_maildir_flags
These functions are nicely documented here, (though the implementation
does not quite match the documentation yet---as plainly evidenced by
the current results of the test suite).
This adds group [maildir] and key 'synchronize_flags' to the
configuration file. Its value enables (true) or diables (false) the
synchronization between notmuch tags and maildir flags. By default,
the synchronization is disabled.
This prevents any of the private functions from being leaked out
through the library interface (at least when compiling with a
recent-enough gcc to support the visibility pragma).
Previously we were using Xapian's add_document to allocate document ID
values for notmuch_message_t objects. This had the drawback of adding
a partially constructed mail document to the database. If notmuch was
subsequently interrupted before fully populating this document, then
later runs would be quite confused when seeing the partial documents.
There are reports from the wild of people hitting internal errors of
the form "Message ... has no thread ID" for example, (which is
currently an unrecoverable error).
We fix this by manually allocating document IDs without adding
documents. With this change, we never call Xapian's add_document
method, but only replace_document with either the current document ID
of a message or a new one that we have allocated.
Thanks to Michal Sojka <sojkam1@fel.cvut.cz> for pointing out the
correct fix, which I verified in the freely-available WG14/N1124 draft
(from the C99 working group) which is available here:
http://www.open-std.org/JTC1/SC22/wg14/www/docs/n1124.pdf
The sequential identifiers have the advantage of being guaranteed to
be unique (until we overflow a 64-bit unsigned integer), and also take
up half as much space in the "notmuch search" output (16 columns
rather than 32).
This change also has the side effect of fixing a bug where notmuch
could block on /dev/random at startup (waiting for some entropy to
appear). This bug was hit hard by the test suite, (which could easily
exhaust the available entropy on common systems---resulting in large
delays of the test suite).
The recent support for renames in the database is our first time
(since notmuch has had more than a single user) that we have a
database format change. To support smooth upgrades we now encode a
database format version number in the Xapian metadata.
Going forward notmuch will emit a warning if used to read from a
database with a newer version than it natively supports, and will
refuse to write to a database with a newer version.
The library also provides functions to query the database format
version:
notmuch_database_get_version
to ask if notmuch wants a newer version than that:
notmuch_database_needs_upgrade
and a function to actually perform that upgrade:
notmuch_database_upgrade
This patch adds a new function called notmuch_database_get_all_tags
which can be used to obtain a list of all tags from the database
(in other words, the list contains all tags from all messages). The
function produces an alphabetically sorted list.
To add support for the new function, we rip the guts off of
notmuch_message_get_tags and put them in a new generic function
called _notmuch_convert_tags. The generic function takes a
Xapian::TermIterator as argument and uses the iterator to find tags.
This makes the function usable with different Xapian objects.
Function notmuch_message_get_tags is then reimplemented to call the
generic function with message->doc.termlist_begin() as argument.
Similarly, we implement notmuch_message_database_get_all_tags, the
function calls the generic function with db->xapian_db->allterms_begin()
as argument.
Finally, notmuch_database_get_all_tags is exported through
lib/notmuch.h
Signed-off-by: Jan Janak <jan@ryngle.com>
The rudimentary aspect here is that the date ranges are specified with
UNIX timestamp values (number of seconds since 1970-01-01 UTC). One
thing that can help here is using the date program to determins
timestamps, such as:
$(date +%s -d 2009-10-01)..$(date +%s)
Long-term, we'll probably need to do our own query parsing to be able
to support directly-specified dates and also relative expressions like
"since:'2 months ago'".
If an earlier exception occurred, then it's not unexpected for the
flush to fail as well. So in that case, we'll silently catch the
exception. Otherwise, make some noise about things going wrong at the
time of flush.
We only rarely need to actually open the database for writing, but we
always create a Xapian::WritableDatabase. This has the effect of
preventing searches and like whilst updating the index.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Carl Worth <cworth@cworth.org>
