Re: [PATCH 1/2] python file more pytonic, adjust "if" and "for"

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 6, 2022, 17:16 UTC
Message-ID: <xmqqh70glvno.fsf@gitster.g>
In-Reply-To: <71da6f53a44cd3390d122ff2c0446824313e5101.1665056747.git.gitgitgadget@gmail.com>

"dsal3389 via GitGitGadget" <gitgitgadget@gmail.com> writes:

Show 12 quoted lines

> From: dsal3389 <dsal3389@gmail.com>
>
> L371
> redesign few lines to get rid of the "else" statement
>
> L404
> moved the if statement below another if statement that
> checks if it should exit the code, only if it doesnt need to,
> then we can iterate the for loop and decode the text
>
> Changes to be committed:
> 	modified:   git-p4.py

Compare this with the commits by others in "git log --no-merges" output of this project.

> Signed-off-by: Daniel Sonbolian <dsal3389@gmail.com>

Please have this on the in-body "From:" line we see above. I think GitGitGadget takes it from the author of the commit object it sends out, so you may have to go back to your branch and fix them with "rebase -i".

Show 18 quoted lines

> ---
>  git-p4.py | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/git-p4.py b/git-p4.py
> index d26a980e5ac..0ba5115fa2e 100755
> --- a/git-p4.py
> +++ b/git-p4.py
> @@ -368,10 +368,9 @@ def read_pipe(c, ignore_error=False, raw=False, *k, **kw):
>         """
>      retcode, out, err = read_pipe_full(c, *k, **kw)
>      if retcode != 0:
> -        if ignore_error:
> -            out = ""
> -        else:
> +        if not ignore_error:
>              die('Command failed: {}\nError: {}'.format(' '.join(c), err))
> +        out = ""

I think the code with or without the patch is about the same complexity, but people tend to have harder time understanding logic that involves double negation, so I can imagine that some readers may find the code with the patch harder to understand. In any case, the difference falls into the "it is minor enough that once it is written in one way, it is not worth the churn to rewrite it in the other way" category.

Show 11 quoted lines

> @@ -400,10 +399,10 @@ def read_pipe_lines(c, raw=False, *k, **kw):
>      p = subprocess.Popen(c, stdout=subprocess.PIPE, *k, **kw)
>      pipe = p.stdout
>      lines = pipe.readlines()
> -    if not raw:
> -        lines = [decode_text_stream(line) for line in lines]
>      if pipe.close() or p.wait():
>          die('Command failed: {}'.format(' '.join(c)))
> +    if not raw:
> +        lines = [decode_text_stream(line) for line in lines]
>      return lines

This is in the same "the difference is minor enough that once it is written in one way, it is not worth the churn to rewrite it in the other way" category. Your reasoning might be that massaging of the lines is only needed when we do not die() and it is more efficient to check and die first, but that is optimizing for the wrong case. The code should not die in its normal operation and there is no point optimizing for an error code path.

One thing that might deserve benchmarking and optimizing here is if we can do better than reading everything in lines array and holding onto the original until decoding the whole thing at once at the end of input. If converting each line and appending the result as it is read from the pipe turns out to be more efficient, it may be an optimization worth considering, as it is optimizaing for the normal case.