| Age | Commit message (Collapse) | Author |
|---|
|
darcs-hash:20080928142541-00fcc-90dd4da3c4fd9e689c7d1558f384f21766f07e86
|
|
darcs-hash:20080928142326-00fcc-4ba83303f54bdc9c0c27807e0a8e2f64e4a1b86a
|
|
This suppresses the compiler's warning about a function not known to
be defined.
darcs-hash:20080928135222-00fcc-5f4d902e13be97833fd28c5e611147c577ec1c41
|
|
darcs-hash:20080928134937-00fcc-7caddb09c14e2cfde0f950d150a9d2a18f2b9d78
|
|
Paredit 21 no longer has a message releasing it into the public domain.
Instead it is copyright (C) 2008, Taylor R. Campbell, and available
under the terms of the 3-clause BSD licence. See the paredit.el file
for details.
*** Style and Bugs
- Use of the terms `parenthesis', `bracket', and `delimiter' has been
clarified throughout the file:
. /Parentheses/ are nestable objects indicated by matching delimiter
pairs. For example, the text `(foo (bar baz) quux)' has two
parentheses, `(foo ... quux)' and `(bar baz)'; each is delimited
by matching round delimiters.
. /Delimiters/ are the actual atomic objects that delimit
parentheses. They may be shaped differently; for example, `[' is
an opening square delimiter, and `{' is an opening curly
delimiter.
. The terms `bracket', `brace', and `brocket' (the latter a
misspelling of `broket', for `broken bracket') are eschewed.
The names `paredit-open-parenthesis' &c. have been changed to
`paredit-open-round' &c., and likewise with `bracket' -> `square',
`brace' -> `curly', and `brocket' -> `angled'. The old names with
the morphemes `parenthesis' and `bracket' have been preserved so that
users need not update initialization files immediately, but will
instead be confused when a subsequent minor update breaks all their
initialization files for no good reason.
- Some commands now check their context more strictly, so that they
won't try to do anything in comments or strings. Whether this is the
right thing, I don't know.
- Several small elements of style were adjusted. This should make no
functional difference on the code. (If it does, then I have made a
mistake.)
- paredit.el no longer defines `kill-region-new'; the function is
instead called `paredit-hack-kill-region', to reflect its hackish
nature and to avoid name clashes. I believe that the only name
remaining defined by paredit.el without `paredit-' prefixed is
`backward-down-list', whose definition is so obvious that it would be
silly to define it any other way, so that any name conflict does not
worry me. (I suppose `kill-region-new' is obvious, too -- it is to
`kill-region' as `kill-new' is to `kill'. But the omission from a
pattern isn't as clear as with `backward-down-list'.)
- `paredit-comment-dwim' should work in GNU Emacs 21 now, whose
newcomment.el defines no `comment-or-uncomment-region'. More and
more ugly compatibility hacks are accruing in paredit.el as a
consequence, to my dismay. Oh well.
- The release notes are now formatted more legibly, paginated nicely,
and organized with Outline Mode. The gross discrepancy in writing
style over the years has not changed.
- The introductory comments in the file are more clearly written now.
- Fixed a bug in S-expression slurpage with mixed delimiters.
*** Altered Behaviour
- The bindings for `)' and `M-)' have been exchanged: `)' is bound to
`paredit-close-round' and `M-)' to `paredit-close-round-and-newline',
so that you can now without any glaring exceptions type new code
using the same keystrokes with and without Paredit Mode. You can
also now paste into Emacs through a terminal without leaving spurious
blank lines in the buffer. You are, of course, free to revert to the
old behaviour in your .emacs file.
- `paredit-semicolon' is no more. Now you may insert semicolons as
you please without the infuriating escape of the following text on
the line.
- `paredit-move-past-close-and-newline' will now leave comments on the
same line only if they are margin comments and no S-expressions are
intervening between the point and the comment; that is, only if it
looks like the margin comment really pertains to the expression being
closed will paredit leave it on the same line.
- `paredit-backward-delete', `paredit-forward-delete', and
`paredit-kill' formerly accepted prefix arguments to run the basic
`backward-delete-char', `delete-char', and `kill-line' commands
instead, without regard for the value of the prefix argument. Now
`C-u' will run the basic commands, but
. `paredit-kill' will pass a numeric argument on to `kill-line', and
. `paredit-backward-delete' and `paredit-forward-delete' will both
delete N characters if given a numeric prefix argument N.
(`paredit-kill' should probably do the same, but repeating the
command N times is probably not what you want -- what you probably
want is to kill at most N *lines*, but `paredit-kill' N times might
kill many more lines than that. I don't know what the right thing is
here, but I welcome feedback from users who want to do something like
this.)
- With a `C-u' prefix argument, `paredit-wrap-sexp' now wraps all
S-expressions following the point until the end of the buffer or a
closing delimiter.
- `paredit-splice-sexp' with a `C-u' prefix argument (also known as
`paredit-splice-sexp-killing-backward') will now kill from the point,
rather than from the beginning of the next S-expression (or, with
`C-u C-u', from the end of the previous S-expression). This means
that it is more likely to do what you mean with
(let ((a b) (c d))
|;; I want to preserve this comment after `C-u M-s'.
(foo bar baz))
- `paredit-splice-sexp' now splices strings, by removing backslash
escapes, or signals an error if doing so would destroy the structure
of the buffer.
- I have finally introduced the first bit of code to try to deal
sensibly with broken buffers. It will probably go only downhill from
here, and continue in an interminable series of kludges to handle
every possible way in which the buffer can go *wrong* (it's bad
enough how many ways it can be *right*). If you try type a closing
delimiter from within a partial S-expression that has an opening
delimiter but no closing delimiter, then it will honk at you and
insert the closing delimiter -- or if what you typed doesn't match
the opening delimiter, it will just honk and refuse to do anything.
Also, `DEL' and `C-d' will delete spurious (but not [yet] mismatched)
opening and closing delimiters, respectively. (Thanks to John
Wiegley for inspiring me to do these dreary deeds.)
*** New Commands
- New command `paredit-yank-pop' cooperates with `paredit-wrap-sexp' by
behaving either like `yank' or like `yank-pop' if the previous
command was `paredit-wrap-sexp' or `paredit-yank-pop', and with the
added bonus of reindenting the newly wrapped form. It is in need of
a key to be bound to it; since it is different from both `yank' and
`yank-pop', I decided not to override `C-y' or `M-y', and I
considered `C-c C-y', but I imagine that many major modes want to
take that.
- New command `paredit-focus-on-defun' moves display to the top of the
definition at the point.
- New command `paredit-reindent-defun', which `M-q' is bound to in
Paredit Mode, indents the definition the point is on, or, if the
point is in a string or comment, fills the paragraph instead.
(Thanks to John Wiegley for the idea.)
- New variations on slurpage, barfage, and joinage. I'm still looking
for keys to bind to these commands. Find them with the strings
`add-to-{previous,next}-list' and `join-with-{previous,next}-list' in
their names. (Again, thanks to John Wiegley for the idea.)
- New command `paredit-convolute-sexp' performs the combined function
of `paredit-splice-sexp-killing-backward', `paredit-wrap-sexp', and
`yank'. Example:
(let ((foo bar))
(let ((baz quux))
|(zot mumble)
(frotz))
(gargle mumph))
->
(let ((baz quux))
(let ((foo bar))
(zot mumble)
(frotz)
(gargle mumph)))
This would otherwise have been done with the key sequence `C-u M-s
C-M-u M-( C-y C-M-u C-M-q'. `C-u M-s' could be `M-up', and `C-y
C-M-u C-M-q' could be `C-c C-y' if that key is chosen for
`paredit-yank-pop', making the sequence `M-up C-M-u M-( C-c C-y'. If
there is a good key for `paredit-convolute-sexp', it could be a nice
improvement over even that terser sequence. (Once again, this was
inspired by John Wiegley's idea (and name).)
[Observe, though, that the form (FROTZ) stuck with (ZOT MUMBLE) the
whole time, and was not carried along as the `end' of the (LET ((BAZ
QUUX)) ...) form. Hence this is *not* the idea mentioned below by
the name `paredit-interchange-sexps', but a simpler approximation of
the idea.]
- `define-paredit-pair' now defines commands `paredit-wrap-...' for
wrapping S-expressions with different delimiters, like
`paredit-wrap-sexp'. The function `paredit-wrap-sexp' now accepts
optional arguments for the delimiters to insert; the new commands are
defined in terms of the modified `paredit-wrap-sexp'. `M-[' is now
bound to `paredit-wrap-square'.
darcs-hash:20080928134400-00fcc-618d1b2da41f6a67aaef69d72d89ce15379372cc
|
|
*** Preliminary Support for XEmacs
This version introduces preliminary support for XEmacs. The changes
needed were:
- `check-parens' is called by the `paredit-mode' initialization only if
it is fbound.
- The forward and backward deletion keys are specified differently in
GNU Emacs and XEmacs.
- `paredit-forward' and `paredit-backward' use a "_" interactive
specification in XEmacs in order to preserve the mark.
- `paredit-region-active-p' is just `region-active-p' in XEmacs.
- Some hair was needed to handle S-expression parse error conditions
properly, and versions of XEmacs earlier than 21.5 may have obscure
problems as a result.
*** Style and Bugs
- rxvt-specific terminal escape sequences are no longer bound to the
commands that the keys those sequences denote are. Set your
environment variables correctly and keep your rxvt.el updated.
(Aren't terminals fun?)
- HTML output is now properly quoted. Some vestigial indirections in
the tables have been removed.
- Yet *ANOTHER* `paredit-kill' bug is fixed. I do not know what the
bug is or why it happened, but it seems to be gone now.
- Improved robustness of `paredit-join-sexps' and `paredit-splice-sexp'
with respect to their use in the middle of atoms, and made splicing
within a string illegal.
- Fixed several bugs in the paredit word killing commands. In the
process, however, I encountered what seems to be a bug in Emacs's
`parse-partial-sexp', which may mean bugs in other things...
- Eliminated dependency on `cl' feature.
- Fixed a bug in `paredit-backward-kill-word' that would cause deletion
of the first double-quote in `(foo "|")'.
- Fixed a bug with `paredit-backward-kill-word' at the end of buffer.
- Fixed a bug with `paredit-backward-kill-word' before any words in a
buffer.
*** Altered Behaviour and New Functionality
- `paredit-mode' now accepts a prefix argument to mean that it should
not check the parentheses in the buffer first. Supply the prefix
argument with care; though it is useful for editing small parts of a
buffer in Lisp when the rest is random, it might also screw the whole
buffer up in unpredictable ways, because most of paredit doesn't
even try to handle invalid structure.
- Parenthesis blinking is improved somewhat to better respect user
customization of `blink-matching-paren'.
- The paredit minor mode no longer exchanges C-j & RET; following the
GNU Emacs conventions, it now leaves RET alone and binds C-j to
`paredit-newline'. Those of you, such as me, who relied on the old
exchange will have to update your .emacs files.
- C-left and C-right are now bound to paredit-forward-barf-sexp and
paredit-forward-slurp-sexp, instead of M-left and M-right, whose word
motion bindings are no longer shadowed.
- The behaviour of (, ", M-(, and M-" with respect to active regions
and prefix arguments has been regularized:
. With neither an active region, i.e. an active mark and
transient-mark-mode, ( and " will insert a pair of delimiters at
the point, and M-( and M-" will wrap the following expression with
delimiters.
. With an active region and no prefix argument, if and only if the
region contains only balanced expressions, all four commands will
wrap the region with the respective delimiters.
. With a prefix argument N, all four commands will wrap the next N
expressions with the commands' respective delimiters.
" and M-", of course, escape any characters that need escaping first
if they are used to wrap regions.
- Implemented slurpage into strings.
- Made `M-- M-s' equivalent to `M-- M-1 M-s'.
- Changed `paredit-insert-pair' so that it will not skip whitespace
before inserting the pair.
- `paredit-splice-sexp' with a prefix argument and friends (i.e. `M-s',
`M-<up>', and `M-<down>') now always create a new entry on the kill
ring.
darcs-hash:20080928134307-00fcc-ad5658e27bb367859e56b76b9e376a2a384737d1
|
|
This version introduces support for delimiters other than round
brackets. Previously, if the major mode's syntax table considered
square brackets (and curly braces, &c.) to be delimiters, since no
paredit commands would insert them balanced, deleting them would be
tricky: paredit's DEL & C-d bindings would refuse to delete them
because they would be imbalanced unless you manually type the closing
delimiter.
Now commands are defined for the opening and closing of parentheses
(round), brackets (square), braces (curly), and brockets (angled),
named `paredit-open-<type>', `paredit-close-<type>-and-newline', and
`paredit-close-<type>'; paredit-mode binds the opening and closing
square bracket keys to be `paredit-open-bracket' and `paredit-close-
bracket', respectively. The rest you can bind yourself; this minimal
pair of bindings will, I think, account for accidental insertion,
elisp vectors, and (bletch) the equation of square and round brackets
as parentheses in some non-standard Scheme extensions.
Also now supported in this version is insertion of delimiter pairs
around active regions in transient-mark-mode. If you mark a region
with transient-mark-mode enabled, you can use any of the delimiter
pair insertion keys (like opening round bracket, double-quote, &c.)
to insert a pair of delimiters around the region. There are now two
ways to parenthesize lists of expressions with visual feedback: using
M-( (paredit-wrap-sexp) followed by C-) (paredit-forward-slurp-sexp)
until satisfied, and now C-M-SPC (mark-sexp) until you have marked
the desired expressions and then any of the delimiter pair insertion
keys to insert the delimiters.
darcs-hash:20080928134228-00fcc-3ec209344dbaea07e93f2e02cddead13f7e39c6a
|
|
*** Style and Bugs
- Corrected terminal arrow key sequences *again*. M-left and M-right
were backwards.
- Put the save-excursion back in paredit-recentre-on-sexp. I don't
remember why it was taken out in version 13.
- Fixed HTML output to stop producing spurious <tr> tags.
- Corrected a number of paredit command examples.
- Aesthetic changes to the code:
. Regularized some aspects of code style.
. Repaginated so that all pages are at most 80 lines long, and most
are at least 40 lines.
. Formatted headings for an outline regexp to match so that
outline-minor-mode works nicely on paredit.el.
*** Altered Behaviour and New Functionality
- Implemented paredit-forward-kill-word & paredit-backward-kill-word,
or M-d & M-DEL, which are like kill-word & backward-kill-word, but
they will not kill parenthesis, string, or comment delimiters; they
will preserve the structure of S-expressions, while the built-in
Emacs word killing commands would destroy it.
- M-" is now bound to paredit-meta-doublequote, which has the old
behaviour of paredit-close-string-and-newline if within a string,
but which wraps the following S-expression (or N S-expressions) in
double-quotes if without a string; paredit-doublequote does the
same, but the default argument is 0, not 1.
- M-S (paredit-split-sexp) no longer deletes horizontal space in
strings before splitting them into two. The rationale, as
suggested by Zbigniew Szadkowski, is that whitespace is usually
significant in strings, while not in lists, and you can type M-\
M-S if you really do want the horizontal space deleted anyway.
- Reintroduced paredit-join-sexps as M-J. The implementation is now
more robust: it ensures that the two S-expressions to join match --
i.e. they are both lists, or they are both strings, or they are
both symbols --, and it correctly handles the atom case now as
well.
- Extended paredit command examples to allow multiple steps in
succession of a single example.
darcs-hash:20080928134213-00fcc-6ad0986678483423b9db18258e129b7824e33a01
|
|
*** Style and Bugs
- Rewrote all documentation strings to be in the imperative mood, per
GNU Emacs's guidelines. Some documentation strings may have been
corrected, too, but I can't be bothered to grovel through the diff
to pick out all changes to all documentation strings.
- Forced parenthesis flashing even with show-paren-mode enabled.
- Fixed bug in forward deletion within strings so that the empty
string can be deleted.
- Simplified determination of whether the point is in a comment.
*** Altered Behaviour and New Functionality
- Eliminated paredit-terminal-mode. All key bindings it had are now
incorporated into paredit-mode's keymap. I may have changed some
keybindings, too, but I don't remember what they were if I did. I
think I fixed some of the keybindings in the terminal.
- Added examples to documentation of all paredit commands, as well as
code to generate an HTML file containing the examples in nicely
formatted tables (sorry, web purists).
- Made paredit-mode refuse to be enabled in a buffer with imbalanced
parentheses.
- Updated documentary heading. It now explains how to customize keys
while still autoloading and how to make paredit's RET work nicely
with SLIME's REPL.
- Improved semicolon insertion: (a) to accept a numeric prefix
argument dictating a number of semicolons to insert, instead of a
prefix argument that forces the insertion of a semicolon without a
trailing newline, which can be effected with C-q anyway; and (b) to
allow insertion of semicolons before existing comments without
inserting a superfluous leading newline. To comment out code, you
must still use M-; or M-x comment-region.
darcs-hash:20080928134133-00fcc-8b43de19c101c35bdc2972ba2d6304468710367f
|
|
- Introduced M-S (paredit-split-sexp) to split lists or strings from
the middle.
- Fixed the definition of M-; to use (kbd "M-;") to correctly
identify the key sequence meta-semicolon, not "M-;" for M hyphen
semicolon.
darcs-hash:20080928134120-00fcc-f3d0c5f1318c5cc1c4de78c69666bbc126f305fd
|
|
- Rewrote comment insertion code.
- Implemented M-; (paredit-comment-dwim), which is like comment-dwim
but inserts comments more appropriately with respect to paredit.
darcs-hash:20080928134053-00fcc-89725dbfcfd855de4a0e07e0ffa1e133229c0482
|
|
- paredit-close-list-and-newline now refuses to move a margin comment
to another line; instead it will help to preserve the column of the
comment.
- The semicolon key is now bound to a command that will automatically
move any code following the point onto the next line, so that you
do not inadvertently comment out half expressions. You can still
use M-; (comment-dwim) to comment out specific regions that are not
meant to be code (e.g., old comments that were accidentally
uncommented) or whole S-expressions, usually in conjunction with
C-M-SPC (mark-sexp).
darcs-hash:20080928133903-00fcc-0f07413273679feeca677ed9d8e5464876c7f772
|
|
- Removed M-\ (paredit-join-sexps), whose key binding conflicts with
delete-horizontal-space and whose implementation was inadequate and
led to general uselessness of the command.
- Improved RET (paredit-newline) so that it does not do anything
fancy within strings and first tests whether the point is in a
comment before checking whether it is in a character.
- Changed paredit-skip-whitespace from skip-syntax-{forward,backward}
to skip-chars-{forward,backward}, because in the Lisp syntax table
newlines are not considered whitespace -- rather, they are comment
ends --, but we want to skip them nevertheless.
- Corrected paredit-kill in a way I don't understand.
- Minor code improvements, including:
. Changed call to previous-line to use forward-line instead.
. Removed unnecessary save-excursion in paredit-recentre-on-sexp.
. IF indentation changes.
darcs-hash:20080928133822-00fcc-703d3e78cd56715f9f46966923349ba2561d8a72
|
|
- Implemented M-r (paredit-raise-sexp), which raises a single
S-expression up the tree, deleting all of its siblings and its
enclosing list.
- Rearranged some arrow key bindings again.
- Made paredit-forward-delete and paredit-backward-delete check for
buffer bounds and also matching of the delimiters of empty lists.
- Added a buffer bounds check to paredit-kill.
- Made backward barfing signal an error, not just a message.
darcs-hash:20080928133744-00fcc-b6843f09a351f97bd75c8f768399e66f8e49a697
|
|
- Changed the key for paredit-splice-sexp from M-/, which is by
default the popular dabbrev-expand, to M-s, which I was surprised
to find no existing binding for.
- Added a prefix argument to paredit-splice-sexp; see the command's
documentation for details. M-up is now equivalent to C-u M-s;
M-down, to C-u C-u M-s.
- Fixed a lot of the terminal key sequences for the arrow key
bindings in paredit-terminal-mode.
- Modified the behaviour of paredit-forward and paredit-backward to
change only one level of nesting depth, not to repeat until there
is a sibling S-expression to move across.
- Changed a lot of code to use character syntax, instead of exact
character comparisons, for better generality.
- Rewrote much of paredit-kill, again.
darcs-hash:20080928133658-00fcc-6f28a122f0471a40b9f80ce8e60303ebf196f42f
|
|
- Introduced paredit-forward and paredit-backward, which are like
forward-sexp and backward-sexp but which will also go up a nesting
level if at the end of a list.
- Introduced C-c C-M-l (paredit-recentre-on-sexp), whose name is
self-explanatory.
- Added support for numeric prefix arguments to paredit-open-list.
- Fixed paredit-kill so that it would correctly kill whitespace
between parentheses, as in ( ).
- Changed suggestion of what to put in your .emacs from require to
autoload.
darcs-hash:20080928133616-00fcc-1983846a129d64ac4b85e04df49fe9f4b6e36416
|
|
- Introduced enable-paredit-mode and disable-paredit-mode to better
choose which one to enable.
- Forced blinkage of matching parentheses in paredit-close-list and
paredit-close-list-and-newline, even if show-paren-mode is enabled.
darcs-hash:20080928133359-00fcc-e05f09d8e69fcaa6077e1bd3610dccc46a6dc441
|
|
- Added paredit-terminal-mode, which is like paredit-mode but which
provides key bindings that work in terminals, while paredit-mode
contains many (such as controlled brackets) that do not work in
terminals. Thanks to Jorgen Schaefer for suggesting many of the
terminal key bindings.
- Exchanged RET and C-j: RET now inserts the fancy newline with auto-
indentation, while C-j inserts a literal line feed. While this
defies convention, and some people prefer not to do this, I have
found that it is more convenient for RET to have the behaviour of
the common case, where auto-indentation is desired, and for C-j to
insert the uncommon exact, literal line feed. You can always
customize the key bindings yourself, of course.
- Rearranged arrow key bindings.
- Implemented paredit-close-list-and-newline, which is like
paredit-close-list followed by RET (paredit-newline); and M-"
(paredit-close-string-and-newline), which is similar but for
strings.
The closing round bracket character now inserts the newline, while
the meta modifier inhibits this.
- Overhauled paredit-kill.
- Extended slurpage and barfage to permit their operation across
arbitrary depth changes.
- Fixed bizarre bug with backward barfage. I apologize for the
alliteration.
- Fixed a number of other small bugs.
- Prefixed `paredit-' to the remaining functions defined in the file
that did not already have the prefix.
- Defined backward-down-list, which for some reason is not provided
by lisp.el, although up-list, down-list, & backward-up-list all
are. (This is the sole exception to the above item. It deserves
no prefix because it ought to be defined in lisp.el with this name
anyway.)
darcs-hash:20080928133248-00fcc-cff6c6ab1db5bb274d4b53bf702b99804dc33ebd
|
|
- Changed paredit-open-list and paredit-close-list so that they will
refuse to insert parentheses inside character literals.
- Substituted new code for GNU Emacs's move-past-close-and-reindent.
I do not remember why now, and I no longer understand either code
well enough to discern differences, other than that Emacs's
m-p-c-a-r is probably faster because it incrementally parses each
step of the way. All I can see is that paredit.el's code now
inserts indentation before the closing parenthesis while Emacs's
m-p-c-a-r does not.
- Fixed bugs in deletion within character literals and strings.
darcs-hash:20080928133142-00fcc-9a9fb3492e906d5e09da1f1ccb68bd1944ab8269
|
|
- Flushed M-" (paredit-insert-doublequote), which was rather useless
and which " (paredit-doublequote) now subsumes the functionality of.
- Added instrumented forward deletion as well as backward deletion,
which now behave well inside strings.
- Flushed unnecessary individual round bracket insertion keys; use
C-q instead.
- Added C-left & C-right: backward-sexp & forward-sexp, respectively.
- Fixed the test of whether the point is in a character literal.
- Modified " (paredit-doublequote) to insert escaped double-quote if
in the middle of the string, rather than to jump past the string's
closing delimiter.
- Introduced bogus backslash escaping mechanism.
- Introduced new command for breaking the line and indenting, and
bound C-j, rather than RET, to it, according to convention.
- Improved C-k (paredit-kill), particularly in strings where it will
no longer kill the closing delimiter of the string.
- Changed the splicage, joinage, slurpage, and barfage commands so
that they will reindent only the modified list, not the whole
definition.
darcs-hash:20080928133010-00fcc-90a9f6931a49b99f633def9bc51157b3bc40a6bf
|
|
darcs-hash:20080928131127-00fcc-7af7487ee9835f4f8a88b30a93ef8fe43a60e59c
|
