git prompt: faster count of untracked files

[nomaed]

In a repository with 18k untracked files (by accident, node_modules not in .gitignore), counting them using count (git ls-files) was causing a considerable lag with prompt generation.

After changing it to git ls-files | wc -l, the count process is at least twice as fast. Not ideal, but it avoids caching the entire git ls-files result as a string to count the number of lines by piping the result through wc -l directly.

[nomaed]

Test/demo:

$ cat count-fish
#!/usr/bin/env fish
count (git ls-files --others)

$ cat count-wc 
#!/usr/bin/env fish
git ls-files --others | wc -l

$ time -p ./count-fish
20930
real 0.18
user 0.15
sys 0.02

$ time -p ./count-fish
20930
real 0.17
user 0.14
sys 0.02

$ time -p ./count-fish
20930
real 0.19
user 0.15
sys 0.02

$ time -p ./count-wc
20930
real 0.09
user 0.04
sys 0.03

$ time -p ./count-wc
20930
real 0.06
user 0.03
sys 0.02

$ time -p ./count-wc
20930
real 0.09
user 0.04
sys 0.02

[faho]

Minor nit: You can pipe to string trim as well (wc -l | string trim) instead of nesting command substitutions.

[nomaed]

@faho Good point, done.

[krader1961]

Why is | string trim required? The wc -l portion of that command will return a single line that the fish sub-command will turn into a single value since it strips (i.e., ignores) a trailing newline.

[nomaed]

@faho commented about it, so I made the change
#3294 (comment)

Edit: I thought at first that you were talking about why piping instead of using a string trim with a sub-command. As @floam wrote, the reason is exactly that there are leading spaces that are saved and the string needs to be trimmed.

[krader1961]

Right. I forgot that there are brain-dead wc implementations (e.g., the BSD one) that left-pad the value with spaces.

[floam]

@krader1961 The spaces will not be ignored:

> set -l foo (command git ls-files --others --exclude-standard | wc -l)
> echo $foo
     102

[krader1961]

Thanks. Merged as dc02587.