Re: [PATCH v2 01/18] cat-file: split expand_atom into 2 functions

From: Christian Couder <christian.couder@gmail.com>
Date: Jan 17, 2018, 23:04 UTC
Message-ID: <CAP8UFD1hnP3yab-qDKPkNcj5VadUD_HW5+V1e2nwPAB8r31zoQ@mail.gmail.com>
In-Reply-To: <20180117214906.GC13128@sigill.intra.peff.net>

On Wed, Jan 17, 2018 at 10:49 PM, Jeff King <peff@peff.net> wrote:

Show 15 quoted lines

> On Tue, Jan 16, 2018 at 10:22:23AM +0300, Оля Тележная wrote:
>
>> >> In other words, I think the endgame is that expand_atom() isn't there at
>> >> all, and we're calling the equivalent of format_ref_item() for each
>> >> object (except that in a unified formatting world, it probably doesn't
>> >> have the word "ref" in it, since that's just one of the items a caller
>> >> might pass in).
>>
>> Agree! I want to merge current edits, then create format.h file and
>> make some renames, then finish migrating process to new format.h and
>> support all new meaningful tags.
>
> I think we have a little bit of chicken and egg there, though. I'm
> having trouble reviewing the current work, because it's hard to evaluate
> whether it's doing the right thing without seeing the end state.

Yeah, to me it feels like you are at a middle point and there are many ways to go forward.

As I wrote in another email though, I think it might be a good time to consolidate new functionality by adding tests (and perhaps documentation at the same time) for each new atom that is added to ref-filter or cat-file. It will help you refactor the code and your patch series later without breaking the new functionality.

> So what
> I was suggesting in my earlier mails was that we actually _not_ try to
> merge this series, but use its components and ideas to build a new
> series that does things in a bit different order.

Yeah, I think you will have to do that, but the tests that you can add now for the new features will help you when you will build the new series.

And hopefully it will not be too much work to create this new series as you will perhaps be able to just use the interactive rebase to build it.

I also don't think it's a big problem if the current patch series gets quite long before you start creating a new series.