The previous code was assuming it was fine to push_next without
growing, which used to be the case with the previous implementation
because we always have poped the current thread that we try to push.
However now that we use a ring-buffer, m_next_begin == m_next_end can
either mean full, or empty. We solve this by assuming it means empty
and never allowing the buffer to become full, which means we need
to grow after pushing to next if we get full.
Fixes#4859
Handle begin/end paste directly in paste csi, manage paste buffer
out of get_char, filter Key::Invalid earlier.
get_next_key returning Key::Invalid means there was some input but
it could not be represented as a Key. An empty optional means there
was no input at all.
Text pasted into Kakoune's normal mode is interpreted as command
sequence, which is probably never what the user wants. Text
pasted during insert mode will be inserted fine but may trigger
auto-indentation hooks which is likely not what users want.
Bracketed paste is pair of escape codes sent by terminals that allow
applications to distinguish between pasted text and typed text.
Let's use this feature to always insert pasted text verbatim, skipping
keymap lookup and the InsertChar hook. In future, we could add a
dedicated Paste hook.
We need to make a decision on whether to paste before or after the
selection. I chose "before" because that's what I'm used to.
TerminalUI::set_on_key has
EventManager::instance().force_signal(0);
I'm not sure if we want the same for TerminalUI::set_on_paste?
I assume it doesn't matter because they are always called in tandem.
Closes#2465
If, for example, the buffer path now is a directory, MappedFile will
throw on construction. Using a try block to explicitely allow errors
fixes the issue.
Instead of potentially decoding for each thread, always decode as
its only slightly slower than finding next codepoint (which will
be necessary anyway) and pass the codepoint to each thread.
The command line "hook -group xyz " should get scope completions but
it actually gets hook completions because "xyz" is wrongly interpreted
as positional argument.
Fix this by using the parameters parser to compute positional
arguments.
Fixes#4840
Some plugins (*cough* kak-lsp) and help texts tend to have immensely long content
in a single line. This generates info boxes that span the entire terminal width.
This is made especially worse on widescreen monitors or at small text size.
This grants user control over how wide these boxes are.
I deliberately avoid pushing this change to `kak-lsp` because it's not the only
plugin that this could help--see the `hook` help text for an example of this
problem in vanilla Kakoune. I would also suggest that since this is a rendering
concern, it be handled by the terminal rendering logic.
We only grow when the ring buffer is full, which allows for a nice
simplification of the code.
Tell grow_ifn if we pushed in current or next so that we can
distinguish between filled by next or filled by current when
m_current == m_next_begin
Instead of two stacks growing from the two ends of a buffer, use
a ring buffer growing from the same mid spot.
This avoids the costly memory copy every step when we set next
threads as the current ones.
Erasing fully trimmed display atoms one by one means we have to
shift all the remaining ones every time. This is wasteful and we
can just erase all the fully trimmed atom in one go.
Fixes#4797
Previously it would result in a stray single-character selection at the
beginning of the input text.
For example:
[abcabc] -> split on 'a' -> [a][bc]a[bc]
or
[foobarfoobar] -> split on 'foo' -> [f]oo[bar]foo[bar]
Note that this behavior was not occuring if the input text was at the
beginning of the buffer
Commit 933e4a599 (Load buffer in command line order, 2022-12-06)
introduced a regression: the command
$ kak /boot/grub/grub.cfg
Fatal error: no such buffer '/boot/grub/grub.cfg'
quits with no indication of the underlying error.
Prior to 933e4a599, it would open the *scratch* buffer instead,
and show an error in the status line, pointing to the debug buffer,
which would contain:
error while opening file '/boot/grub/grub.cfg':
/boot/grub/grub.cfg: Permission denied
Let's fix this scenario by matching the old behavior.
Recent changes for selection-undo added an assertion that triggers
when a mouse-drag overlaps with an insert mode, because both events
record selection history. However this is actually fine. The one
that finishes last concludes the selection edition, while the other
one will be a nop.
The test could be simpler (i.e. not require sleeps) but I figured it
doesn't hurt add this since we don't have any comparable tests.
After buffer modification - in particular after deletion - adjacent
selection history entries may correspond to the same effective
selection when applied to the current buffer. This means that we
sometimes need to press <c-h> multiple times to make one visible
change. This is not what the user expects, so let's keep walking the
selection history until we hit an actual change.
Alternatively, we could minimize the selection history after buffer
changes but I think that would make the it worse after content
undo+redo.
Each selection undo operation is surrounded by pair of
begin_edition()/end_edition() calls.
The original reason for adding these was that in one of my preliminary
versions, a WinDisplay hook could break an undo chain, even if the
hook did not affect selections at all. This has since been fixed.
By surrounding the undo with begin_edition()/end_edition(), try to
ensure that any selection modification that happens in a WinDisplay
hook would not break the undo chain. Essentially this means that,
after using <c-h> to undo a buffer change, this was meant to
make sure that <c-k> could redo that buffer change.
However, it turns out this actually doesn't work. The attached test
case triggers an assertion. As described in the first paragraph,
the only real-world motivation for this is gone, so let's simplify
the behavior.
The assertion fix means that we can test the next commit better.
It turns out that neither <a-f> or <a-t> make sense when run at the
beginning of the buffer. When I first created the check, I thought
that <a-f> made sense if the character under the cursor was the
character being searched for. I was wrong, <a-f> should always go
at least one character backwards.
This commit fixes a bug in Buffer::advance where it would first access
m_lines[-1], and then check whether or not that access would have
segfaulted. This commit moves the check to before the segfault would
occur.
is run as
> ./kak -f '2oK.k<c-n><ret><c-n>' /dev/null
The crash occurs because when <c-n> is pressed for the second time,
it attempts to use m_completions from the first press of <c-n>.
This only happens when kakoune is run with -f, because when this is done
interactively, there is a client, which means that m_completions gets
reset. This removes the check that causes that difference.
I am not *completely* sure that this is the best way to solve the
problem, since I am not completely sure why that check was put there
in the first place.
Using BufferIterator adds overhead, but we know that DisplayAtoms
cannot span multiple buffer lines and hence we can directly iterate
using char pointers.
Each draft context gets its own private copy of the selections.
Any selection changes will be thrown away when the draft context
is disposed. Since selection-undo is only supported as top-level
command, it can never be used inside a draft context, so let's stop
recording it.
No functional change.
Calculating the length of an atom means we need to decode every
codepoint and compute its column width. This can prove quite expensive
in trim_from as we can have full buffer lines, so on buffer with long
lines we might have to go through megabytes of undisplayed data.
Pass the first buffer on the the command line explicitely to client
creation. This ensure the buffer list matches the command line, which
makes buffer-next/buffer-previous a bit more useful.
Fixes#2705
With overlapping selections, pasting after breaks assumption of
SelectionList::for_each as our changes are no longer happening in
increasing locations.
We hence cannot rely on the ForwardChangeTracker in that case and
have to rely on the more general (and more costly) ranges update logic.
This interacts poorly with paste linewise pastes and we try to preserve
the current behaviour by tracking the last paste position.
Overall, this change really begs for overlapping selections to be
removed, but we will fix them like that for now.
Fixes#4779
Comparing iterators between buffers should never happen, and the
only place we did was with default constructed BufferIterator which
we replace by casting the iterator to bool.
This should improve performance on iterator heavy code.
When the file system runs out of space, "write -force" will fail but
doesn't print "No space left on device".
Let's fix this by including such an underlying error. Untested.
Backstory: I alias "w" to a command that runs "write -force %arg{@}".
so I can overwrite files that already exist outside the editor (I
should probably get used to the new behavior).
Commit 69053d962 (Use menu behavior when completing change-directory,
2022-07-19) made ":cd dir/" actually run ":cd dir/first-subdir",
which can be surprising.
This is usually irrelevant because you rarely type the trailing slash.
However it does happen after correcting an error with `<backspace>`
and friends. For for example,
:cd d<tab>/f<backspace>
results in
:cd dir/
We should probably fix user expectations here. Do this by adding "dir/"
as valid completion. This requires us to allow empty candidates in
"RankedMatch" but there's no harm in that. This means we need to
filter out empty completions from shell-script-candidates elsewhere.
Alternative fix: we could revert 69053d962. This would remove the
convenient menu behavior but that wouldn't be a huge deal.
Fixes#4775
Mapping upper case keys is legitimate, for exampled so that they behave
the same as a lower case mapping. The current rejection of those mappings
is a misguided attempt to prevent mapping *to* upper case keys as those
will never get triggered.
Closes#4769
After a failed
:write file-that-already-exists
a user might want to type ":<up> -f<ret>" to force-overwrite.
This doesn't work because :write's switches must precede the filename.
It's dual :edit does not have this restriction.
Some commands require switches to precede positional arguments for a
good reason; for example because positional arguments might start with
"-" (like ":echo 1 - 1").
There seems to be no reason for the :write restriction, so remove
it. Same for :enter-user-mode.
Thanks to alexherbo2 for reporting.
Running %sYeti<ret>casdf on file
[example.journal.txt](https://github.com/mawww/kakoune/issues/4685#issuecomment-1193243588)
can cause noticeable lag. This is because we insert text at 6000
selections, which means we need to update highlighters in those lines.
The runtime for updating range highlighters is quadratic in the
number of selections: for each selection, we call on_new_range(),
which calls add_matches(), which calls std::rotate(), which needs
needs linear time.
Fix the quadratic runtime by calling std::inplace_merge() once instead
of repeatedly calling std::rotate(). This is works because ranges
are already sorted.
I used this script to benchmark the improvements.
(In hindsight I could have just used "-ui json" instead of tmux).
#!/bin/sh
set -ex
N=${1:-100}
kak=${2:-./kak.opt}
for i in $(seq "$N")
do
echo -n "\
2022-02-06 * Earth
expense:electronics:audio 116.7 USD
liability:card -116.7 USD
2022-02-06 * Blue Yeti USB Microphone
expense:electronics:audio 116.7 USD
liability:card -116.7 USD
"
done > big-journal.ledger
echo > .empty-tmux.conf 'set -sg escape-time 5'
test_tmux() {
tmux -S .tmux-socket -f .empty-tmux.conf "$@"
}
test_tmux new-session -d "$kak" big-journal.ledger
test_tmux send-keys '%sYeti' Enter c 1234567890
sleep .2
test_tmux send-keys Escape
while ! test_tmux capture-pane -p | grep 123
do
sleep .1
done
test_tmux send-keys ':wq' Enter
while test_tmux ls
do
sleep .1
done
rm -f .tmux-socket .empty-tmux.conf
This script's runtime used to grow super-linearly but now it grows
linearly:
kak.old kak.new
N=10000 1.142 0.897
N=20000 2.879 1.400
Detailed results:
$ hyperfine -w 1 './bench.sh 10000 ./kak.opt.'{old,new}
Benchmark 1: ./bench.sh 10000 ./kak.opt.old
Time (mean ± σ): 1.142 s ± 0.072 s [User: 0.252 s, System: 0.059 s]
Range (min … max): 1.060 s … 1.242 s 10 runs
Benchmark 2: ./bench.sh 10000 ./kak.opt.new
Time (mean ± σ): 897.2 ms ± 19.3 ms [User: 241.6 ms, System: 57.4 ms]
Range (min … max): 853.9 ms … 923.6 ms 10 runs
Summary
'./bench.sh 10000 ./kak.opt.new' ran
1.27 ± 0.09 times faster than './bench.sh 10000 ./kak.opt.old'
$ hyperfine -w 1 './bench.sh 20000 ./kak.opt.'{old,new}
Benchmark 1: ./bench.sh 20000 ./kak.opt.old
Time (mean ± σ): 2.879 s ± 0.065 s [User: 0.553 s, System: 0.126 s]
Range (min … max): 2.768 s … 2.963 s 10 runs
Benchmark 2: ./bench.sh 20000 ./kak.opt.new
Time (mean ± σ): 1.400 s ± 0.018 s [User: 0.428 s, System: 0.083 s]
Range (min … max): 1.374 s … 1.429 s 10 runs
Summary
'./bench.sh 20000 ./kak.opt.new' ran
2.06 ± 0.05 times faster than '../repro.sh 20000 ./kak.opt.old'
LineRangeSet::add_range() calls Vector::erase() in a loop over the
same vector. This could cause performance problems when there are many
selections. Fix this by only calling Vector::erase() once. I didn't
measure anything because my benchmark is dominated by another issue
(see next commit).
LineRangeSet::remove_range() also has a suspicious call to erase()
but that one is only used in test code, so it doesn't matter.
From the issue:
> It often happens to me that I carefully craft a selection with multiple
> cursors, ready to make changes elegantly, only to completely mess it
> up by pressing a wrong key (by merging the cursors for example). Being
> able to undo the last selection change (even if only until the previous
> buffer change) would make this much less painful.
Fix this by recording selection changes and allowing simple linear
undo/redo of selection changes.
The preliminary key bindings are <c-h> and <c-k>.
Here are some other vacant normal mode keys I considered
X Y
<backspace> <minus>
# ^ =
<plus> '
unfortunately none of them is super convenient to type. Maybe we
can kick out some other normal mode command?
---
This feature has some overlap with the jump list (<c-o>/<c-i>) and
with undo (u) but each of the three features have their moment.
Currently there's no special integration with either peer feature;
the three histories are completely independent. In future we might
want to synchronize them so we can implement Sublime Text's "Soft
undo" feature.
Note that it is possible to restore selections that predate a buffer
modification. Depending on the buffer modification, the selections
might look different of course. (When trying to apply an old buffer's
selection to the new buffer, Kakoune computes a diff of the buffers
and updates the selection accordingly. This works quite well for
many practical examples.)
This makes us record the full history of all selections for each
client. This seems wasteful, we could set a limit. I don't expect
excessive memory usage in practice (we also keep the full history of
buffer changes) but I could be wrong.
Closes#898
To be able to undo selection changes, we want to record selections
from all commands that modify selections. Each such command will get
its own private copy of the selections object.
This copy will live until the command is finished executing.
All child commands that are run while the command is executing,
will also use the same copy, because to the user it's all just one
selection change anyway.
Add an RAII object in all places where we might modify selections.
The next commit will use this to create the private selections copy
in the constructor (if there is none) and remove redundant history
items in the destructor.
We could avoid the RAII object in some places but that seems worse.
For lifetimes that don't correspond to a lexical scope, we use a
std::unique_ptr. For lambdas that require conversion to std::function,
we use std::shared_ptr because we need something that's copyable.
The next commit changes the selections to a history of
selections. Today we directly access the selections data member. Let's
instead use an accessor method, to reduce the number of changes in
the next commit.
clang/clangd complain about the new HashSet type:
hash_map.cc:98:20: warning: braces around scalar initializer [-Wbraced-scalar-init]
set.insert({10});
^~~~
The argument to HashSet<int>::insert is just an int, so we don't
need braces. Only an actual HashMap would need braces to construct
a HashItem object.
When passing a filename parameter to "write", the -force parameter
allows overwriting an existing file.
The "write!" variant (which allows writing files where the current
user does not have write permissions) already implies -force.
All other variants (like write-quit or write-all) do not take a
file parameter.
Hence -force is relevant only for "write". Let's hide it from the
autoinfo of the other commands.
It's difficult to avoid duplication when constructing the constexpr
SwitchMap because String is not constexpr-enabled. Today, all our
SwitchMap objects are known at compile time, so we could make SwitchMap
use StringView to work around this. In future we might want to allow
adding switches at runtime, which would need String again to avoid
lifetime issues.
