This change provides the following improvements:
- Changes the view of the command history list, so that the script no longer depends on perl for performance.
- Enables syntax color highlighting on fish v4.3.3 and newer.
- Provides a preview window with the selected commands, and expanded view of the highlighted command if available.
- Improves the delete functionality, by successfully handling very large numbers of selected commands.
- Inserts commands in their formatted form with `<Alt-Enter>`.
---
* fish: Change history list view
The view of the command history is changed, so that no manipulation is
performed on the output of history command: The history item count
number is replaced by the Unix time of the command, multi-line display
is disabled and line wrapping is enabled by default. On fish v4.3.3
and newer, the use of ANSI color codes for syntax highlighting is now
possible, while the script no longer depends on perl for performance.
Fixes#4661
* fish: Reformat selected history commands with ALT-ENTER
* Add $FZF_WRAP environment variable
The variable is set when line wrapping is enabled. It has the value
`word` if word wrapping mode is set, otherwise it has the value `char`.
* fish: Add command history preview
The preview shows the highlighted command and any selected commands,
after first being formatted and colorized by fish_indent. The timestamp
of the highlighted command in the preview is converted to strftime
format "%F %a %T" if the date command is available.
The preview is hidden on start, and is displayed if more than 100
columns are available and one of the following conditions are met:
- The highlighted item doesn't completely fit in list view (line
wrapping is enabled for the preview and is now disabled for the list).
- The highlighted item contains newlines (multi-line commands or
strings).
- The highlighted item contains chained commands in a single line, that
can be broken down by the formatter for cleaner view.
- One or more commands are marked as selected.
* fish: Handle deletion of large number of selected history entries
* fish: Change wrapping options for the preview-window
We have been checking the mawk version by extracting <x>, <y>, <z>,
and <d> part from "mawk <x>.<y>.<z> <d>" in the output of the "mawk -W
version" and testing <x>, <y>, <z>, and <d> using an arithmetic
evalaution. However, <d> is ensured to be an integer only in "x.y.z
>= 1.3.4". Otherwise, it may cause a syntax error in the arithmetic
evaluation. The mawk started to include the date as an integer in the
<d> position only from mawk-1.3.3-20090721. We should first check
that "x.y.z >= 1.3.4" and then check the value of "d". In case, "mawk
-W version" produces a completely different text, we should also
redirect stderr of the arithmetic commands to /dev/null.
For the line "Host host1 # this is a comment", the current
implementation generates words in an inline comment as hostnames.
This patch removes the comment before generating the hostname.
In ~/.ssh/config, "=" can also be used as a separator between the
field name and the value. The current master does not properly handle
this and generate a hostname "=" or one starting with "=". This patch
correctly handles it.
Comments are anyway removed in the subsequent call to `sub(/#.*/,
"")`, and it becomes a blank line. Blank lines do not have fields, so
they are ignored in the next for-loop.
In the current implementation, any hostnames in /etc/hosts containing
"0.0.0.0" as a part (such as "110.0.0.0" would be excluded. "0.0.0.0"
should be checked by the exact match.
An entry of the form `[example.com]:port,192.168.0.1 ...` in
~/.ssh/known_hosts are not properly processed. The current
implementation gives up the matching on the first occurrence of `]`,
the subsequent 192.168.0.1 would not be extracted. This patch
continues the analysis and removes "]" together with "[".
This patch also removes the ":port" part from the hostnames in
~/.ssh/known_hosts. One cannot use the form "hostname:port" in the
arguments to the ssh command anyway.
Solaris awk at /usr/bin/awk is meant for backward compatibility with
an ancient implementation of 1977 awk in the original UNIX. It lacks
many features of POSIX awk. To use a standard-conforming version in
Solaris, one needs to explicitly use /usr/xpg4/bin/awk.
macOS awk is a variant of nawk, but it contains a unique patch for the
UTF-8 support. However, this patch causes the problem. If the input
contains any non-UTF-8 data, macOS awk stops processing and does not
do anything, instead of ignoring the unrecognized data and continue
the processing. However, the contents of the ssh configuration and
/etc/hosts is not under the control of fzf, so we cannot fix the input
when those files contain non-UTF-8 data. To work around this
behavior, one can set the locale to LC_ALL=C to treat the input data
with the plain 8-bit encoding.
This is a rewrite of __fzf_parse_commandline function, that fixes the
following issues, when CTRL-T/ALT-C is used and current command line
token contains:
- Escaped newlines (\n): This never worked correctly, but after 282884a,
the string would split, and the script would enter an infinite loop
while trying to set $dir.
- Escaped bell (\a, \cg), backspace (\b), form feed (\v, \cl), carriage
return (\r), vertical tab (\v, \ck): walker-root would not set
correctly for existing directories containing any of those characters.
- Regular expression special characters (^, +, ? etc): $dir would not be
be stripped from $fzf_query if it contained any of those characters.
The lowest supported fish version is v3.1b. For optimal operation, the
function uses more recent commands when supported by the running
version. Specifically, for versions equal or newer than:
- v3.2.0: Sets variables using PCRE2 capture groups of `string match
--regex` when needing to preserve any trailing newlines and
simultaneously omit the extra newline that is appended by `string
collect -N`.
- v3.5.0: Uses the builtin path command for path normalization, dirname
extraction and existing directories check.
- v4.0.0: Uses the --tokens-expanded option of commandline, for
expansion and dealing with unbalanced quotes and incomplete escape
sequences. It also uses the regex style of string-escape, to prepare
variable contents for regex operations. This is not used in older
versions, because they don't escape newlines.
CTRL-T/ALT-C now works correctly when selecting files or directories
that contain newlines in their names. When external commands defined by
$FZF_CTRL_T_COMMAND/$FZF_ALT_C_COMMAND are used (for example the fd
command with -0 switch), the --read0 option must also be set through
$FZF_CTRL_T_OPTS/$FZF_ALT_C_OPTS.
Enable inserting multiple history commands. To disable, set `--no-multi`
through `$FZF_CTRL_R_OPTS`.
Also, remove the usage of `become` action, to avoid leaving behind
temporary files.
Co-authored-by: Junegunn Choi <junegunn.c@gmail.com>
- Remove check/set of FZF_TMUX_HEIGHT variable. It is already done by
__fzf_defaults.
- Remove unnecessary begin/end block.
- Simplify result variable check.
- Set the command line using a single call to commandline.
- Remove check/set of FZF_TMUX_HEIGHT variable. It is already done by
__fzf_defaults.
- Remove unnecessary begin/end block.
- Pass all fzf options (except query) through FZF_DEFAULT_OPTS variable.
- Remove check/set of FZF_TMUX_HEIGHT variable. It is already done by
__fzf_defaults.
- Remove unnecessary begin/end block.
- Simplify result variable check.
- Insert file names using a single call to commandline.
Append all arguments after the first one, so that functions don't have
to pass all appending options as a single string. Also, output
everything as a single string (an array of one item).
More natural processing of the query taken from command line, by
unquoting/unescaping the token. Unescaped open quotes are removed.
Because of how `string unescape` works, if both single and double quotes
are present, with the outer quotes open, only the outer quotes are
removed.
Examples:
`'foo bar'`, `"foo bar"`, `foo\ bar` becomes `foo bar`
`"foobar`, `'foobar`, `foo"bar`, `foo'bar` becomes `foobar`
`'"foo"'`, `'"foo"` becomes `"foo"`
`"'foo'"`, `"'foo'` becomes `'foo'`
`"'foo` becomes `'foo`
`'"foo` becomes `"foo`
