Re: move detection doesnt take filename into account

From: Elliot Wolk <elliot.wolk@gmail.com>
Date: Jul 1, 2014, 14:40 UTC
Message-ID: <53B2C870.4030406@gmail.com>
In-Reply-To: <287177519.16421.1404206204124.JavaMail.zimbra@dewire.com>

interesting that it considers suffixes {only suffixes following periods?}. this is insufficient, in my opinion.

with all other things being equal, it ought to find the closest match {using smith-waterman or some such algorithm}.

as a real-world use case, i have a repository with empty files that mirrors the file structure of a directory containing large binary files. when i move a dir, it seems to select the files renamed at random.

On 07/01/2014 05:16 AM, Robin Rosenberg wrote:

Show 20 quoted lines

>
> ----- Ursprungligt meddelande -----
>> Från: "Elliot Wolk" <elliot.wolk@gmail.com>
>> Till: git@vger.kernel.org
>> Skickat: måndag, 30 jun 2014 8:38:18
>> Ämne: move detection doesnt take filename into account
>>
>> if you move two identical {e.g.: empty} files to two new locations in a
>> single commit, the move detection picks them {seemingly?} arbitrarily.
>> it should use a statistical algorithm to compare the filenames and pick
>> a likely match.
> I think it does, but based on filename suffix. E.g. here is a rename of
> three empty files with a suffix.
>
>   3 files changed, 0 insertions(+), 0 deletions(-)
>   rename 1.a => 2.a (100%)
>   rename 1.b => 2.b (100%)
>   rename 1.c => 2.c (100%)
>
> -- robin