A completion table can specify a `group-function` in its metadata.
The group function takes two arguments, a completion candidate and a
transform argument. The group function is used to group the
candidates after sorting and to enhance the completion UI with group
titles.
If the transform argument is nil, the function must return the title
of the group to which the completion candidate belongs. The function
may also return nil if the candidate does not belong to a group.
If the transform argument is non-nil, the function must return the
transformed candidate. For example, the transformation allows to
remove a redundant part of the candidate, which is then displayed in
the title.
The grouping functionality is guarded by the customizable variable
`completions-group` and turned off by default for the *Completions*
buffer.
The specific form of the `group-function` has been chosen in order to
allow allocation-free grouping. This is important for completion UIs,
which continously update the displayed set of candidates (Icomplete,
Vertico, Ivy, etc.). Only when the transform argument is non-nil the
candidate transformation is performed, which may involve a string
allocation as done in the function `xref--completing-read-group`.
The function `xref-show-definitions-completing-read` makes use of the
`group-function`, by moving the file name prefix to the title. If
grouping is enabled, the *Completions* are displayed as
"linenum:summary" instead of "file:linenum:summary". This way the
*Completions* buffer resembles the *Occur* buffer.
* doc/lispref/minibuf.texi: Add documentation.
* lisp/minibuffer.el (completion-metadata): Describe the
`group-function` in the docstring.
(completions-group): Add guard variable, off by default.
(completions-group-format): Add variable defining the format string
for the group titles.
(completions-group-title): Add face used by `completions-group-format`
for the group titles.
(completions-group-separator): Add face used by
`completions-group-format` for the group separator lines.
(minibuffer--group-by): New grouping function.
(minibuffer-completion-help): Use it.
(display-completion-list): Add optional GROUP-FUN argument.
(completion--insert-strings): Add optional GROUP-FUN argument. Insert
group titles if `completions-format` is `one-column`. Transform each
candidate with the GROUP-FUN. Attach the untransformed candidate to
the property `completion--string`.
* lisp/simple.el (choose-completion): Retrieve the untransformed
completion candidate from the property `completion--string`.
* lisp/progmodes/xref.el:
(xref--completing-read-group): New grouping function.
(xref-show-definitions-completing-read): Use it.
(sql-remove-continuation-prompt, sql-send-string): Move newline
insertion.
Previously, the preoutput filter `sql-remove-continuation-prompt'
inserted a leading newline in the interactive SQL buffer if it decided
that is has to remove continuation prompts and that it had found all
it was looking for.
1) This filter function was a doubtful place to do that (arguably, its
name does not suggest any action like this).
2) The behavior worked inconsistently, eg, when sending a single-line
"SELECT" statement, because it only ran when the filter function
needed to remove any prompts (for example, not when sending a
region without newlines).
This can lead to misaligned table headers, which explains why
emacswiki and stackoverflow both present several fixes to this
behavior.
* lisp/progmodes/bug-reference.el
(bug-reference-try-setup-from-rmail): Match the Rmail mbox filename
against GROUP-REGEXP in bug-reference-setup-from-mail-alist.
(bug-reference-mode-force-auto-setup): New function which forces
auto-setup even if bug-reference-bug-regexp and
bug-reference-url-format are already set.
* lisp/progmodes/bug-reference.el
(bug-reference-auto-setup-functions): New defvar so that other
packages can add their own auto-setup functions to it.
* lisp/progmodes/bug-reference.el (bug-reference--run-auto-setup): Use
the new variable instead of hard-coding the 4 functions we've had
already.
* lisp/progmodes/grep.el (grep-mode-font-lock-keywords):
Adapt regexp to match MS Windows-style shell-quoting.
* test/lisp/progmodes/grep-tests.el: New file.
(font-lock-fontify-syntactically-region): Use `comment-end-skip` as
fallback for `font-lock-comment-end-skip`, as is done for
`font-lock-comment-start-skip` (and as the name suggests).
* lisp/progmodes/opascal.el (opascal-mode): Revert last change,
made unnecessary.
* lisp/progmodes/cperl-mode.el (cperl-find-pods-heres): Delete
?? from the allowed bare regexp delimiters.
(cperl-short-docs): Delete ?...? from the documentation.
* test/lisp/progmodes/cperl-mode-tests.el (cperl-bug-47598):
Add tests for good, bad, and ambiguous use of ? as regex
delimiter (bug#47598).
* lisp/progmodes/ruby-mode.el (ruby-font-lock-keywords): puts and
printf can be called without arguments, so the font locking of
"bare" calls to either is incorrect. The fix is to font-lock them
as for other kernel methods which accepts zero or more arguments
(bug#48180).
* lisp/progmodes/ruby-mode.el (ruby-encoding-map):
Add entry for utf-8 (bug#48043).
(ruby--detect-encoding): Don't convert to string too early, so
that returning nil is meaningful.
(ruby-mode-set-encoding): Convert to string here.
* lisp/progmodes/cc-fonts.el (c-font-lock-cut-off-declarators): Instead of
using a crude 2,000 characters back limit for backward searching, which is
erroneous when that point is in a literal, use the already calculated
c-determine-limit result.
* doc/emacs/maintaining.texi (Xref Commands): Remove the
description of the TAB binding. Enhance the description of the
RET binding. (Bug#44611)
* etc/NEWS: Announce the obsolescence of TAB binding in XREF.
* lisp/progmodes/xref.el (xref-goto-xref): Improve doc string.
* lisp/progmodes/project.el (project-compilation-buffer-name-function):
New option.
(project-compile): Use it.
(project-prefixed-buffer-name): New function.
Copyright-paperwork-exempt: yes
* lisp/progmodes/project.el (project--find-regexp-in-files):
Don't unquote file names here.
* lisp/progmodes/xref.el (xref-matches-in-files): Do it here.
And only if the first element in the list is quoted (bug#47799).
* lisp/progmodes/cperl-mode.el (cperl-electric-else): Don't expand
scalar variables like '$continue' as keywords. (Bug#47902)
* test/lisp/progmodes/cperl-mode-tests.el
(cperl-test-hyperactive-electric-else): Verify that keywords are
expanded but variable names aren't.
This is only a band-aid; it would be better to fix xref.el to work
with quoted filenames as well.
* lisp/progmodes/project.el (project--find-regexp-in-files): Unquote
filenames before passing them to 'xref-matches-in-files'.
* test/lisp/progmodes/project-tests.el (project/quoted-directory):
Also test 'project-find-regexp'.
* lisp/progmodes/project.el (project--files-in-directory):
Quote LOCALDIR a bit earlier, to affect
xref--find-ignores-arguments as well (bug#47799).
* lisp/progmodes/xref.el (xref-matches-in-directory):
Quote the dir passed to xref--rgrep-command.
