Re: [PATCH 4/6] builtin/submodule--helper: store update_clone information in a struct

From: Stefan Beller <sbeller@google.com>
Date: Jul 16, 2018, 23:55 UTC
Message-ID: <CAGZ79kaOy0pZJCob4d=66FQLaOu_q_BAJeFc94xkFa0SSqYigw@mail.gmail.com>
In-Reply-To: <xmqqo9f7t1jy.fsf@gitster-ct.c.googlers.com>

On Mon, Jul 16, 2018 at 12:37 PM Junio C Hamano <gitster@pobox.com> wrote:

Show 24 quoted lines

>
> Stefan Beller <sbeller@google.com> writes:
>
> > The information that is printed for update_submodules in
> > 'submodule--helper update-clone' and consumed by 'git submodule update'
> > is stored as a string per submodule. This made sense at the time of
> > 48308681b07 (git submodule update: have a dedicated helper for cloning,
> > 2016-02-29), but as we want to migrate the rest of the submodule update
> > into C, we're better off having access to the raw information in a helper
> > struct.
>
> The reasoning above makes sense, but I have a few niggles on the
> naming.
>
>  * Anything you'd place to keep track of the state is "information",
>    so a whole "_information" suffix to the structure name does not
>    add much value.  We've seen our fair share of (meaningless)
>    "_data" suffix used in many places but I think the overly long
>    "_information" that is equally meaningless trumps them.  I think
>    we also have "_info", but if we are not going to have a beter
>    name, let's not be original and stick to "_data" like other
>    existing codepath.  An alternative with more meaningful name is
>    of course better, though, but it is not all that much worth to
>    spend too many braincycles on it.

agreed.

>  * Is the fact that these parameters necessary to help populating
>    the submodule repository are sent on a line to external process
>    the most important aspect of the submodule_lines[] array?  As

In terms of what the submodule--helper does, it is. It is not the most important aspect in the big picture, or once we finish the migration to not have shell interpret its output.

>    this step is a preparation to migrate out of that line/text
>    oriented IPC, I think line-ness is the least important and
>    interesting thing to name the variable with.

ok.

Show 7 quoted lines

> If I were writing this patch, perhaps I'd do
>
>         struct update_clone_data *update_clone;
>         int update_clone_nr, update_clone_alloc;
>
> following my gut, but since you've been thinking about submodule
> longer than I have, perhaps you can come up with a better name.

That makes sense. We do not need to mention 'submodule' as that ought to be obvious from the file name already and 'update_clone' is just enough to describe what we are doing. Although there is already struct submodule_update_clone, which would be the better candidate for 'update_clone' and this new struct would be used as a helper in that struct, so update_clone_data

I'll think of adding a patch to rename that already existing struct (submodule_update_clone -> update_clone) and renaming this to update_clone_data.

Show 26 quoted lines

>
> > @@ -1463,8 +1469,9 @@ struct submodule_update_clone {
> >       const char *recursive_prefix;
> >       const char *prefix;
> >
> > -     /* Machine-readable status lines to be consumed by git-submodule.sh */
> > -     struct string_list projectlines;
> > +     /* to be consumed by git-submodule.sh */
> > +     struct submodule_update_clone_information *submodule_lines;
> > +     int submodule_lines_nr; int submodule_lines_alloc;
> >
> >       /* If we want to stop as fast as possible and return an error */
> >       unsigned quickstop : 1;
> > @@ -1478,7 +1485,7 @@ struct submodule_update_clone {
> >  #define SUBMODULE_UPDATE_CLONE_INIT {0, MODULE_LIST_INIT, 0, \
> >       SUBMODULE_UPDATE_STRATEGY_INIT, 0, 0, -1, STRING_LIST_INIT_DUP, 0, \
> >       NULL, NULL, NULL, \
> > -     STRING_LIST_INIT_DUP, 0, NULL, 0, 0}
> > +     NULL, 0, 0, 0, NULL, 0, 0, 0}
>
> The structure definition and this macro definition are nearby, so it
> is not crucial, but its probably not a bad idea to switch to C99
> initializers for a thing like this to make it more readable, once
> the code around this area stabilizes back again sufficiently (IOW,
> let's not distract ourselves in the middle of adding a new feature
> like this one).

Are we still in the phase of "test balloon" or do we now accept C99 initializers all over the code base?

But I agree to defer the conversion for now.

Thanks, Stefan