Re: [RFC PATCH 1/1] mv: integrate with sparse-index

From: Derrick Stolee <derrickstolee@github.com>
Date: Mar 23, 2022, 13:10 UTC
Message-ID: <e61303b8-10ad-5b5b-d48b-cac89ac53d29@github.com>
In-Reply-To: <CAJyCBORkauHAdDiHjQ2Agj3bNhLNPtKk-VW5=bNmBfNuQtv7hA@mail.gmail.com>

On 3/22/2022 4:38 AM, Shaoxuan Yuan wrote:

Show 15 quoted lines

> Hi all,
> 
> On Tue, Mar 22, 2022 at 3:45 AM Derrick Stolee <derrickstolee@github.com> wrote:
>> I'm particularly surprised in how much 'git mv' doesn't work very
>> well in the sparse-checkout environment already, which makes things
>> more difficult than "just" doing the normal sparse index things.
>>
>> It's good that we are discovering them and working to fix them.
>>
>> Thanks,
>> -Stolee
> 
> Really appreciate the mentoring and tips here, I'm trying to make some progress
> now. The problems facing here certainly push me to explore more and know
> better about the codebase. Appreciate all the help :-)

A thought occurred to me while thinking about these difficulties: perhaps it is better to start with 'git rm' since that does only half of what 'git mv' does. It should be a smaller lift as a first contribution. There is even a clear loop that is marked with "TODO: audit for interaction with sparse-index."

As we've discovered in this thread, the direction for integrating a builtin with the sparse index should follow this outline:

0. Test the builtin in t1092 with interactions inside and outside
   of the sparse-checkout cone.*

1. Add command_requires_full_index = 0 line to the builtin.

2. Check for failures and diagnose them.

3. Check for index expansion and remove them as necessary.
   (Go back to 2.)

4. Run performance tests.

(*) This step is the one we failed to focus enough on previously.

Of course, if you've already gotten really far on 'mv' and don't want to switch context, then keep at it.

Thanks, -Stolee