Hi Kuba,

On Wed, 29 Jan 2020, Jakub Narebski wrote:

> Shourya Shukla <shouryashukla.oo@gmail.com> writes:
>
> > Hello,
> >
> > I was looking at the previous year projects[1] and a project intrigued me, namely:
> > "Convert scripts to builtins".
> >
> > Following from Christian's advice[2], I have decided to focus on my project proposal.
> > I noticed that various commands such as "git bisect", "git web--browse"(it particularly
> > interests me) are still in their "shell" form and will be needed to be converted into
> > their "C" form as per the project description.
> [...]
> > [1]: https://git.github.io/SoC-2019-Ideas/
> > [2]: https://lore.kernel.org/git/CAP8UFD2Fo=2suQDLwzLM-Wg3ZzXpqHw-x0brPtPV0d4dRsgs9A@mail.gmail.com/
>
> As far as I know, "git bisect" is currently being converted from shell
> to C by Miriam Rubio for Outreachy project [3], so I am not sure if it
> would be feasible as GSoC 2020 project.

Indeed. That one "is already taken".

> I'm not sure if it would be possible and if it would make sense to
> convert "git instaweb" and/or it's helper script "git web--browse" from
> shell to C.

I agree, both of those scripts seem not to benefit all that much from
being converted, while some people would still consider them to be
developed easier as shell scripts.

> I think trying to convert either "git stash" or "git submodule" to C
> would make more sense.

Oh, but `git stash` is already converted. The only two remainining shell
scripts for which I would consider a conversion to C to make sense are
`git submodule` and `git mergetool`.

Large parts of `git submodule` are already implemented in `git
submodule--helper`, so that's a head start (thanks Stephan Beller!).

The `git mergetool` command is a bit trickier, as it consists of three
scripts, really, with the `difftool--helper` depending on
`mergetool--lib`.

Realistically, I think that it would be possible for a GSoC student who is
already very familiar with the code base and with submodules to finish the
conversion of `git submodule` in one season.

The same is probably not true for `git mergetool`: it would require a
couple of seasons to convert, and a good chunk of the first month would be
taken by planning a conversion strategy.

As of the current `master`, the `git-*.sh` scripts are:

# In the process of being converted

git-bisect.sh

# Candidates for being converted

git-difftool--helper.sh
git-mergetool--lib.sh
git-mergetool.sh
git-submodule.sh

# Trivial conversions

git-merge-octopus.sh
git-merge-one-file.sh
git-merge-resolve.sh

# Already deprecated

git-filter-branch.sh
git-legacy-stash.sh
git-rebase--preserve-merges.sh

# Is this even needed anymore?

git-quiltimport.sh

# Probably better to leave them as shell scripts

git-instaweb.sh
git-request-pull.sh
git-web--browse.sh

# Not even Git commands

git-parse-remote.sh
git-sh-i18n.sh
git-sh-setup.sh

The situation of the Perl scripts to be converted is much nicer:

# Already in code review

git-add--interactive.perl

# Too complex/too dependent on the Perl packages

git-send-email.perl
git-svn.perl

# Support for legacy SCMs that are less and less used

git-archimport.perl
git-cvsexportcommit.perl
git-cvsimport.perl
git-cvsserver.perl

So: after `git add -i`, I think we're done with the conversion of the Perl
scripts. Took long enough ;-)

Ciao,
Dscho

>
> [3]: https://public-inbox.org/git/20200128144026.53128-1-mirucam@gmail.com/t/#u
>
> Best,
> --
> Jakub Narębski
>