Re: [BUG] diff algorithm selection issue

From: ydirson@free.fr <ydirson@free.fr>
Date: May 26, 2020, 16:56 UTC
Message-ID: <253380167.1036247407.1590512216359.JavaMail.root@zimbra39-e7>
In-Reply-To: <xmqqd06qoec9.fsf@gitster.c.googlers.com>

Hi Junio,

Show 12 quoted lines

> > When the config has diff.algorithm=patience set, "git diff
> > --minimal" seems to
> > be ignored, and does not give the same output as "git diff
> > --diff-algorithm=minimal",
> > but really the same as "git diff --diff-algorithm=patience".
> 
> As I wrote, that is absolutely the intended behaviour.
> 
> When patience and other algorithm learns how to trade cycles off
> with output size, --minimal may make a difference, but unlike
> "--diff-algorithm=minimal" that forces Myers algorithm, the
> "--minimal" option should not change the underlying algorithm.

OK, so then the problem is just in the doc, where --diff-algorithm=minimal should rather be documented as something like:

 The basic greedy "myers" diff algorithm, spending extra time to make
 sure the smallest possible diff is produced (equivalent to
 `--diff-algorithm=myers --minimal`).

Or is it rather intended to be `--diff-algorithm=default --minimal`, whatever the default may be in the future ?

As for the other flags, --patience and --histogram should probably documented as backward-compatibility aliases for --diff-algorithm=, right ?

I'll send a formal patch if all of this sounds good.

And as a last point, there would be the problem shown by patience diff on the commit referenced in my original post - If patience is still considered for promotion to default some day, I guess we wouldn't want it to make such a bad choice.

Best regards,

-- 
Yann