Re: [PATCHv2] builtin/blame: highlight interesting things

From: Stefan Beller <sbeller@google.com>
Date: Jul 26, 2017, 23:57 UTC
Message-ID: <CAGZ79kZ5x8p7CE6zGQ2HOJFPV_8_XKsENKt70ETACvvYZYKp2Q@mail.gmail.com>
In-Reply-To: <xmqqwp6uojzr.fsf@gitster.mtv.corp.google.com>

On Wed, Jul 26, 2017 at 4:29 PM, Junio C Hamano <gitster@pobox.com> wrote:

Show 26 quoted lines

> Stefan Beller <sbeller@google.com> writes:
>
>> When using git-blame lots of lines contain redundant information, for
>> example in hunks that consist of multiple lines, the metadata (commit name,
>> author) are repeated. A reader may not be interested in those, so darken
>> (commit, author) information that is the same as in the previous line.
>>
>> Choose a different approach for dates and imitate a 'temperature cool down'
>> for the dates. Compute the time range of all involved blamed commits
>> and then color
>>  * lines of old commits dark (aged 0-50% in that time range)
>>  * lines of medium age normal (50-80%)
>>  * lines of new age red (80-95%)
>>  * lines just introduced bright yellow (95-100%)
>>
>> Signed-off-by: Stefan Beller <sbeller@google.com>
>> ---
>>
>>   I played around with it a bit more, using a different color scheme
>>   for dates, http://i.imgur.com/redhaLi.png
>
> I do agree with what this one tries to do, in that a block of lines
> tend to share the same metainfo as they come from the same commit
> and it is distracting to see them repeatedly---doing something to
> make their "these are in one group" nature stand out will give us a
> much better presentation.

Well, we could also try a "zebra" as in sb/diff-color-move to show blocks with the same fancy border detection.

> But does this particular implementation work well for people who use
> black in on white background?  "Darken to make it less distracting"
> may not work on both white-on-black and black-on-white users.

correct. Once we have a shared understanding what the "interesting things" are and how to handle them, I would add color.blame.<slot> options to make it configurable.

> "Show the background only by replacing the letters with SP for
> metainfo that are same as previous line" would work for folks from
> either camp, I would imagine.  And that should be a single feature,
> that can be enabled independently from the age based coloring.

True, I had that as the very first step of this experiment, I lost the patch for it, but could redo it for presentation and discussion.

My impression was that this would remove _too_ much, e.g. if a commit spans more than one screen, you may not see the first line, but only blank space.

> The age coloring is much harder to make it work for folks from both
> camps at the same time with the same color selection.  Yellow on
> white would be terribly unreadable for black-on-white folks, for
> example.

Configuration is key here, I would think, both in the color space, as well as in the selection space. One could imagine that other people would rather have a defined time span, e.g. hard coding "2 weeks/one quarter/ more than a year" or relate that time span to the project history instead of the file history.

Show 7 quoted lines

> If you make "make it less distracting by blanking them out (not
> 'darken them')" feature without the age coloring, that can be usable
> immediately by folks from both camps, even if you cannot find a way
> to do the age coloring that would satisfy both groups.  One group
> can just leave the knob off and not use the age coloring, while the
> other group can use it and people from both camps will be happier
> than the status quo.

ok.

Thanks, Stefan