v0.50.0 not showing process data correctly about fzf HOT 10 CLOSED

@blaineventurine sets an alias for cat1

Thank you for doing what I should have done and giving a fully reproducible config.

from fzf.

Looks like your ps commmand is producing ANSI color codes when it shouldn't. You see colored output when you do ps -ef | cat? And what is the output of command -v ps?

from fzf.

Looks like your ps commmand is producing ANSI color codes when it shouldn't. You see colored output when you do ps -ef | cat? And what is the output of command -v ps?

No color output, and /bin/ps

from fzf.

@blaineventurine sets an alias for cat¹ and has a config file for bat with the --color=always flag. ²

alias cat='bat'

Additionally, there is no --ansi flag in the FZF_DEFAULT_OPTS³.

Minimal setup to reproduce the issue:

export FZF_DEFAULT_OPTS=""
export FZF_COMPLETION_TRIGGER=',,'
alias cat='bat --color=always --number'
source /Users/paria/Developer/fzf/shell/completion.zsh

It seems to have come from commit 152988c.

Prepending command to the cat command in the function below seems to fix the issue. However,
this does not explain why the issue only became apparent after commit 152988c. Anyone understands this?

Another cat command in that file already has been prefixed with command⁴, lets just follow suit ?

from fzf.

Prepending command to the cat command in the function below seems to fix the issue.

Confirmed.

But I have to mention that aliasing cat to bat with --color=always is a bad idea and should be discouraged. It will break existing scripts that expect the normal cat behavior. alias cat='bat --number' should be enough. You'll get pretty output when you cat something, and it will not break scripts as bat will not print color codes when the standard output is not a TTY device.

alias cat='bat --number'

# Colors
cat README.md

# No colors
cat README.md | command cat

from fzf.

Commit 152988c wrapped the emulate part in an if statement. With that change, the alias expansion no longer worked as expected.

Moving the interactive check after the emulate command seems to have solved the issue and prevents us from prepending every single command with command, as was described in #1944.

The movement of the interactive check to the top came with #3449, but it was just a simple one-line check, compared with the commit mentioned prior which wrapped everything in an if statement.

from fzf.

@LangLangBart Thanks for the investigation. Should we do it like this then?

diff --git a/shell/completion.zsh b/shell/completion.zsh
index 7067b06..8cd9737 100644
--- a/shell/completion.zsh
+++ b/shell/completion.zsh
@@ -9,9 +9,6 @@
 # - $FZF_COMPLETION_TRIGGER (default: '**')
 # - $FZF_COMPLETION_OPTS    (default: empty)
 
-if [[ -o interactive ]]; then
-
-
 # Both branches of the following `if` do the same thing -- define
 # __fzf_completion_options such that `eval $__fzf_completion_options` sets
 # all options to the same values they currently have. We'll do just that at
@@ -76,6 +73,8 @@ fi
 # `finally` in lesser languages. We use it to *always* restore user options.
 {
 
+if [[ -o interactive ]]; then
+
 # To use custom commands instead of find, override _fzf_compgen_{path,dir}
 #
 #   _fzf_compgen_path() {
@@ -346,10 +345,10 @@ fzf-completion() {
 zle     -N   fzf-completion
 bindkey '^I' fzf-completion
 
+fi
+
 } always {
   # Restore the original options.
   eval $__fzf_completion_options
   'unset' '__fzf_completion_options'
 }
-
-fi

from fzf.

@LangLangBart Thanks for the investigation. Should we do it like this then?

Yes, key-bindings.zsh may require similar modifications, and maybe in one file a hint ?

# The 'emulate' command should not be placed inside the interactive check,
# as doing so fails to disable alias expansion.
if [[ -o interactive ]]; then

from fzf.

@LangLangBart Sounds good. Can you open a pull request?

from fzf.

But I have to mention that aliasing cat to bat with --color=always is a bad idea and should be discouraged. It will break existing scripts that expect the normal cat behavior. alias cat='bat --number' should be enough. You'll get pretty output when you cat something, and it will not break scripts as bat will not print color codes when the standard output is not a TTY device.

Thanks for this - I believe I copied it out of the bat documentation at some point. Removing --color=always fix the issue I was having.

from fzf.