Because fzf processes HTTP GET requests in the main event loop,
accessing the endpoint from within execute/transform actions would
result in a deadlock and hang fzf indefinitely. This commit sets
a 2 second timeout to avoid the deadlock.
So that it can determine if it should subtract 1 from $FZF_PREVIEW_LINES
to avoid scrolling issue of Sixel image that touches the bottom of the
screen.
Progress:
* Sixel image can now be displayed with other text, and is scrollable
* If an image can't be displayed entirely due to the scroll offset, fzf
will render a wireframe to indicate that an image should be displayed
* Renamed $FZF_PREVIEW_{WIDTH,HEIGHT} to $FZF_PREVIEW_PIXEL_{WIDTH,HEIGHT}
for clarity
* Added bin/fzf-preview.sh script to demonstrate how to display an image
using Kitty or Sixel protocol
An example:
ls *.jpg | fzf --preview='seq $((FZF_PREVIEW_LINES*9/10)); fzf-preview.sh {}; seq 100'
A known issue:
* If you reduce the size of the preview window, the image may extend
beyond the preview window
# Scrolling will behave similarly to CTRL-E and CTRL-Y of vim
fzf --bind scroll-up:offset-up,scroll-down:offset-down \
--bind ctrl-y:offset-up,ctrl-e:offset-down \
--scroll-off=5
Close#3456
Currently there is not option to bind ctrl-delete and shift-delete. As
suggested by issue #3240, shift-delete could be used to bind "delete
entry from history" as it is a common way to do so in other
applications, e.g. browsers.
This, however, does only implement to use the key combination itself and
does not assign a default action to any of them. This does enable to
call one's all predefined actions. With the exec action this can
expanded like the issue #3240 suggested.
If desirable, the key combinations could later get a default behavior.
Co-authored-by: Junegunn Choi <junegunn.c@gmail.com>
Can we find a better name? I have considered the followings.
* 'point', because "the pointer" points to the current item.
* 'shift', 'switch', 'move', etc. These are not technically correct
because the current item can change without cursor movement (--tac,
reload, search update)
* 'change' is already taken. 'change-current' feels a bit wordy and
sounds wrong, 'current-changed' is wordy and doesn't go well with the
other event names
* 'target', not straightforward
Close#3053
* Default border style on Windows is changed to `sharp` because some
Windows terminals are not capable of displaying `rounded` border
characters correctly.
* If your terminal emulator renders each box-drawing character with
2 columns, set `RUNEWIDTH_EASTASIAN` environment variable to `1`.
Even when {q} is empty. Because, why not?
While this can be seen as a breaking change, there is an easy workaround
to keep the old behavior.
# This will show // even when the query is empty
: | fzf --preview 'echo /{q}/'
# But if you don't want it,
: | fzf --preview '[ -n {q} ] || exit; echo /{q}/'
Close#2759
# Put the cursor on the 10th item
seq 100 | fzf --sync --bind 'start:pos(10)'
# Put the cursor on the 10th to last item
seq 100 | fzf --sync --bind 'start:pos(-10)'
Close#3069Close#395
This reverts commit 750b2a6313.
This can cause a deadlock if the endpoints are accessed in the core event
loop via execute action.
fzf --listen 6266 --bind 'space:execute:curl localhost:6266'
Technically, there's no reason to use the API because the information is
already available via `{}` and `{q}`, but I'd like to completely remove
the risk of misuse.
While reading the description of the --info flag, it's not
immediately obvious that the "finder info" is in fact the
UI element representing the match counters.
Favors the line with shorter matched chunk. A chunk is a set of
consecutive non-whitespace characters.
Unlike the default `length`, this new scheme works well with tabular input.
# length prefers item #1, because the whole line is shorter,
# chunk prefers item #2, because the matched chunk ("foo") is shorter
fzf --height=6 --header-lines=2 --tiebreak=chunk --reverse --query=fo << "EOF"
N | Field1 | Field2 | Field3
- | ------ | ------ | ------
1 | hello | foobar | baz
2 | world | foo | bazbaz
EOF
If the input does not contain any spaces, `chunk` is equivalent to
`length`. But we're not going to set it as the default because it is
computationally more expensive.
Close#2285Close#2537
- Not the exact solution to --tiebreak=length not taking --nth into account,
but this should work. And the added benefit is that it works well even
when --nth is not provided.
- Adding a bonus point to the last character of a word didn't turn out great.
The order of the result suddenly changes when you type in the last
character in the word producing a jarring effect.
