Re: [PATCH] completion: improve ls-files filter performance

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi drizzd,

On Wed, 4 Apr 2018, Clemens Buchacher wrote:

> From the output of ls-files, we remove all but the leftmost path
> component and then we eliminate duplicates. We do this in a while loop,
> which is a performance bottleneck when the number of iterations is large
> (e.g. for 60000 files in linux.git).
> 
> $ COMP_WORDS=(git status -- ar) COMP_CWORD=3; time _git
> 
> real    0m11.876s
> user    0m4.685s
> sys     0m6.808s
> 
> Replacing the loop with the cut command improves performance
> significantly:
> 
> $ COMP_WORDS=(git status -- ar) COMP_CWORD=3; time _git
> 
> real    0m1.372s
> user    0m0.263s
> sys     0m0.167s
> 
> The measurements were done with Msys2 bash, which is used by Git for
> Windows.

Those are nice numbers right there, so I am eager to get this into Git for
Windows as quickly as it stabilizes (i.e. when it hits `next` or so).

I was wondering about one thing, though:

> diff --git a/contrib/completion/git-completion.bash b/contrib/completion/git-completion.bash
> index 6da95b8..69a2d41 100644
> --- a/contrib/completion/git-completion.bash
> +++ b/contrib/completion/git-completion.bash
> @@ -384,12 +384,7 @@ __git_index_files ()
>  	local root="${2-.}" file
>  
>  	__git_ls_files_helper "$root" "$1" |
> -	while read -r file; do
> -		case "$file" in
> -		?*/*) echo "${file%%/*}" ;;

This is a bit different from the `cut -f1 -d/` logic, as it does *not
necessarily* strip a leading slash: for `/abc` the existing code would
return the string unmodified, for `/abc/def` it would return an empty
string!

Now, I think that this peculiar behavior is most likely bogus as `git
ls-files` outputs only relative paths (that I know of). In any case,
reducing paths to an empty string seems fishy.

I looked through the history of that code and tracked it all the way back
to
https://public-inbox.org/git/1357930123-26310-1-git-send-email-manlio.perillo@gmail.com/
(that is the reason why you are Cc:ed, Manlio). Manlio, do you remember
why you put the `?` in front of `?*/*` here? I know, it's been more than
five years...

Out of curiosity, would the numbers change a lot if you replaced the `cut
-f1 -d/` call by a `sed -e 's/^\//' -e 's/\/.*//'` one?

I am not proposing to change the patch, though, because we really do not
need to expect `ls-files` to print lines with leading slashes.

> -		*) echo "$file" ;;
> -		esac
> -	done | sort | uniq
> +	cut -f1 -d/ | sort | uniq
>  }
>  
>  # Lists branches from the local repository.

Ciao,
Dscho