Re: [RFC PATCH 0/1] Fuzzy blame

From: Barret Rhoden <brho@google.com>
Date: Apr 3, 2019, 15:25 UTC
Message-ID: <7540d14b-f225-39ec-b37e-54cb157d4a72@google.com>
In-Reply-To: <CAJDYR9R77_+gfOgLXX_Az8iODNRyDTHAT8BAubZeptEWJViYqA@mail.gmail.com>

Hi -

On 3/25/19 7:21 PM, Michael Platings wrote:

Show 6 quoted lines

>> The first thing that comes to mind for me is to plug your fuzzy logic
>> into my patch set.
> Please do! It should be easy to pluck fuzzy_find_matching_lines() and
> its dependencies out. Just to set your expectations, I have not yet
> optimised it and it is highly wasteful right now both in terms of time
> and memory.

I edited my patch set to allow changing the heuristic. I also made a commit that uses your fingerprinting code to match target lines to parent lines. I'll send it all out in another email and CC you. The last commit is still a work in progress.

Regarding stuff like the name of the ignore file, in my first version, I went with whatever git hyper-blame does. That was shot down, rightly so, I think. With a git-config setting, you can name the file whatever you want, or add multiple files. With my current patchset, you can disable the file too with --ignore-revs-file="".

As far as using notes or per-commit info, that might be nice, though it's not a huge burden to have a separate commit - you can wait til after things get merged (so we have the final object name (hash)) and it's not hugely burdensome. But I get that's just my opinion. =)

Barret