The main effort is changing from the old argv style config list
iterators to the new more opaque ones provided by the library (and
backed by the database+file config cache).
This will allow transitioning individual subcommands to the new
configuration framework. Eventually when they are all converted we can
remove the notmuch_config_t * argument.
For now, live with the parameter shadowing in some some subcommands;
it will go away when they are converted.
By using an enum we can have better error detection than copy pasting
key strings around.
The question of what layer this belongs in is a bit
tricky. Historically most of the keys are defined by the CLI. On the
other hand features like excludes are supported in the
library/bindings, and it makes sense to configure them from the
library as well.
The somewhat long prefix for notmuch_config_t is to avoid collisions
with the existing usage in notmuch-client.h.
The renaming and extra values will make sense when we start to convert
subcommands to the new configuration framework. It will also avoid
collisions with a new enum for configuration keys to be introduced in
a future commit.
Fill in the remainder of the documented functionality for
n_d_open_with_config with respect to config file location. Similar
searching default locations of the database file still needs to be
added.
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.
When prompting for one or more tags to add or remove to/from one or
more threads, ensure that the set of tags offered for completion
contains no duplicates.
Some completion packages (e.g. selectrum) will include every member of
the offered list, resulting in the same tag being indicated as a
possibility several times.
In ee897cab8b the upgrade tests from pre v3 databases were
removed. The reasons for that are still valid, but we should still
test the code paths that do the upgrade, and it is relatively
straightforward to do that for v3 to v3 upgrades.
The output of "notmuch show --format=sexp --format-version=4"
may contain `:content-type' entries with `nil' as the value,
when it fails to detect the correct value. Account for that
in a few places where we would otherwise risk a type error.
Note that `string=' does not choke on `nil' because it uses
the `symbol-name' when encountering a symbol.
Users may type some text into the buffer on an address line, before
actually invoking address completion. We now use that text as the
initial input when we begin address completion.
Previously we did knowingly replace the actual initial input with some
completion candidate that happens to match. Which candidate is used is
essentially random, at least when the actual initial input is short.
As a result users very often had to begin completion by deleting the
less than helpful "initial input".
IMO Notmuch should not override the default completion mechanism by
default, at least not globally. But since users are already used to
this behavior it is probably too late to change it. Do the next best
thing and at least allow users to opt out.
When called from code, then this function returns non-nil when the
message at point is a matched message. However it does nothing at all
to present that information to the user when it called interactively.
It is therefore safe to conclude that nobody is using this as a
command.
Like `cl-lib' and `pcase', which are already available in all
libraries, `subr-x' also provided many useful functions that
we would like to use.
Making `subr-x' available in every library from the get-go means
that we can use the functions it defines without having to double
check every single time, whether the feature is already available
in the current library.
We need to load `cl-lib' at run-time because we use more from it than
just macros. Never-the-less many, but not all libraries required it
only at compile-time, which we got away with because at least some
libraries already required it at run-time as well.
We use `cl-lib' and (currently to a lesser extend) `pcase' throughout
the code-base, which means that we should require these features in
most libraries.
In the past we tried to only require these features in just the
libraries that actually need them, without fully succeeding. We did
not succeed in doing so because that means we would have to check
every time that we use a function from these features whether they
are already being required in the current library.
An alternative would be to add the `require' forms at the top of every
library but that is a bit annoying too.
In order to make sure that these features are loaded when needed but
also to keep the noise down we only require them in "notmuch-lib.el",
which most other libraries require, and in most of the few libraries
that do not do so, namely "notmuch-draft.el", "notmuch-message.el" and
"notmuch-parser.el". ("coolj.el", "make-deps.el", various generated
libraries, and "notmuch-compat.el" are left touched.)
To some extend this is a personal preference, but the preference is
strongly dependent on whether one is used to a language that makes it
necessary to use variables like this.
This makes it perfectly clear that we are first getting and then using
a "foo":
(use-foo (get-foo))
Sure this has to be read "inside out", but that's something one better
gets used to quickly when dealing with lisp. I don't understand why
one would want to write this instead:
(let ((the-foo (get-foo)))
(use-foo the-foo))
Both `get-foo' and `use-foo' are named in a way that make it very
clear that we are dealing with a "foo". Storing the value in an
additional variable `the-foo' does not make this any more clear.
On the contrary I makes the reader wonder why the author choose to
use a variable. Is the value used more than once? Is the value
being retrieved in one context and then used in another (e.g. when
the current buffer changes)?
No longer use the function `notmuch-search-get-query', which does
nothing but return the value of that variable. That function was
added in [1: f47eeac0] for use in `notmuch-read-query' along-side
related `notmuch-show-get-query' and `notmuch-tree-get-query' but
using it here makes little sense.
1: f47eeac0b0
emacs: set default in notmuch-read-query
Define these variables as automatically buffer-local, meaning that
they always become buffer-local when set unless explicitly told
otherwise using `setq-default' or when using the Custom interface.
Previously they were declared, which keeps the byte-compiler quiet but
is not actually the same as being defined. `notmuch-search-mode' then
made them buffer-local in the current buffer and then set the local
values. This works but is not kosher.
The definitions of the three non-option variables have to be moved up
a bit to enable the change in the next commit, which see.
We shouldn't force `ido-completion-read' on users who do not otherwise
use Ido. Unfortunately simply turning on `ido-mode' does not change
every `completing-read' into a `ido-completing-read', instead it only
changes file and buffer completion.
I do realize that existing Ido users will initially dislike this
change, but I would like to encourage them to see this as an
opportunity to learn about Fido.
Unlike `ido-mode', build-in `fido-mode' confirms to the standard
completion API, so turning it on causes every `completing-read' to
use the Fido completion mechanism and which is similar to the Ido
mechanism:
> An enhanced `icomplete-mode' that emulates `ido-mode'. This global
> minor mode makes minibuffer completion behave more like `ido-mode'
> than regular `icomplete-mode'."
`notmuch-show-insert-part-text/plain' calls
`notmuch-show-insert-text/plain-hook' with two arguments
MSG and DEPTH. Currently all hook functions ignore MSG but
third-party functions may not. One hook function uses DEPTH.
This function had a few issues.
- Neither its name nor the old comment before it is called made it
clear what it does.
- It took one argument but didn't do anything with it.
- It's doc-string made a few claims, which are untrue and generally
focused on details instead of that its purpose is.
The previous commit switched to lexical-binding but without dealing
with the new warnings about unused lexical arguments and variables.
This commit deals with most of them, in most cases by either removing
leftover bindings that are actually unnecessary, or by marking certain
arguments as "known to be unused" by prefixing their names with "_".
In the case of the functions named `notmuch-show-insert-...' the
amount of silencing that is required is a bit extreme and we might
want to investigate if there is a better way.
In the case of `notmuch-mua-mail', ignoring CONTINUE means that we do
not fully follow the intended behavior described in `compose-mail's
doc-string.
Doing so causes many new compile warnings. Some of these warnings
concern genuine changes in behavior that have to be addressed right
away.
Many other warnings are due to unused variables. Nothing has changed
here, except that the byte-compiler can now detect these pre-existing
and harmless issues. We delay addressing these issues so that we can
focus on the important ones here.
A third group of warnings concern arguments that are not actually used
inside the function but which cannot be removed because the functions
signature is dictated by some outside convention. Silencing these
warning is also delayed until subsequent commits.
`outline-minor-mode' treats comments that begin with three or more
semicolons as headings. That makes it very convenient to navigate
code and to show/hide parts of a file.
Elips libraries typically have four top-level sections, e.g.:
;;; notmuch.el --- run notmuch within emacs...
;;; Commentary:...
;;; Code:...
;;; notmuch.el ends here
In this package many libraries lack a "Commentary:" section, which is
not optimal but okay for most libraries, except major entry points.
Depending on how one chooses to look at it, the "... ends here" line
is not really a heading that begins a section, because it should never
have a "section" body (after all it marks eof).
If the file is rather short, then I left "Code:" as the only section
that contains code. Otherwise I split the file into multiple sibling
sections. The "Code:" section continues to contain `require' and
`declare-function' forms and other such "front matter".
If and only if I have split the code into multiple sections anyway,
then I also added an additional section named just "_" before the
`provide' form and shortly before the "...end here" line. This
section could also be called "Back matter", but I feel it would be
distracting to be that explicit about it. (The IMO unnecessary but
unfortunately still obligatory "... ends here" line is already
distracting enough as far as I am concerned.)
Before this commit some libraries already uses section headings, some
of them consistently. When a library already had some headings, then
this commit often sticks to that style, even at the cost inconsistent
styling across all libraries.
A very limited number of variable and function definitions have to be
moved around because they would otherwise end up in sections they do
not belong into.
Sections, including but not limited to their heading, can and should
be further improved in the future.
In practice this probably does not make a difference or we would
have heard about it many times, but better be safe than sorry.
Process sentinels are called not only when the process has finished
but also on other state changes.
When running "notmuch" we use its full path but when displaying the
command to the user we show just its name for readability reasons.
Avoid passing around both representations because it is very easy
to get the name from the path.
Notmuch itself uses the involved functions just for "notmuch" but
there might be extensions that use them for other executable so we
forgo other potential simplifications.
