Currently, the completion input for __fzf_complete_native is defined inline in the Fish completions:
|
set -l -- completions (eval complete -C \"$argv[1]\") |
This means if the user wants to change the completions in any way, they have to rewrite all the logic of this function. The bash and zsh completions can read from $_fzf_compgen_path and $_fzf_compgen_dir, which allow the user to change the completion input. It would be quite simple to add a similar variable for fish. Since fish has no internal separation between path and directory completion, it could be named something like $_fzf_complete_fish.
I personally want this to align the description text. With the default Fish completion, the descriptions are right-aligned:
While with the FZF completion, they're left-aligned.
I wanted to right-align them again, but FZF doesn't provide alignment logic for specific columns. The easiest solution would be to preprocess the input by piping the eval complete output into printf. That's when I discovered that the fish completion didn't have any command overridability.
I'd be happy to resolve this in a PR if I get approval for the idea (and if nobody beats me to it).
Currently, the completion input for
__fzf_complete_nativeis defined inline in the Fish completions:fzf/shell/completion.fish
Line 140 in cf57950
This means if the user wants to change the completions in any way, they have to rewrite all the logic of this function. The bash and zsh completions can read from
$_fzf_compgen_pathand$_fzf_compgen_dir, which allow the user to change the completion input. It would be quite simple to add a similar variable for fish. Since fish has no internal separation between path and directory completion, it could be named something like$_fzf_complete_fish.I personally want this to align the description text. With the default Fish completion, the descriptions are right-aligned:
While with the FZF completion, they're left-aligned.
I wanted to right-align them again, but FZF doesn't provide alignment logic for specific columns. The easiest solution would be to preprocess the input by piping the
eval completeoutput intoprintf. That's when I discovered that the fish completion didn't have any command overridability.I'd be happy to resolve this in a PR if I get approval for the idea (and if nobody beats me to it).