Re: The config include mechanism doesn't allow for overwriting

From: Jeff King <peff@peff.net>
Date: Oct 24, 2012, 00:51 UTC
Message-ID: <20121024005130.GA19638@sigill.intra.peff.net>
In-Reply-To: <CAEBDL5V0Ffyp306rVo-USCBy_AXXMHXN1yrWmkF1BhzFaq60nA@mail.gmail.com>

On Tue, Oct 23, 2012 at 08:46:47PM -0400, John Szakmeister wrote:

Show 26 quoted lines

> On Tue, Oct 23, 2012 at 10:13 AM, Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:
> [snip]
> > And git config --get foo.bar will give you:
> >
> >     $ git config -f /tmp/test --get foo.bar
> >     one
> >     error: More than one value for the key foo.bar: two
> >     error: More than one value for the key foo.bar: three
> >
> > I think that it would be better if the config mechanism just silently
> > overwrote keys that clobbered earlier keys like your patch does.
> >
> > But in addition can we simplify things for the consumers of
> > "git-{config,var} -l" by only printing:
> >
> >     foo.bar=three
> >
> > Or are there too many variables like "include.path" that can
> > legitimately appear more than once.
> 
> I frequently use pushurl in my remotes to push my master branch both
> to the original repo and my forked version.  I find it very helpful in
> my workflow, and would hate to lose that.  That said, I do like the
> idea of having a config file and the ability to override some of the
> variables.

No, that won't go anywhere. We really do have two classes of variables: things that are expected to be single values, and things that are expected to be lists.

From the perspective of the config code, we don't know or care which is which, and just feed all entries sequentially to a C callback. In practice, the callbacks do one of two things:

  1. Append the values into a list.

  2. Overwrite, and end up with the final value seen.

The trouble is that git-config has to print the values in a reasonable way, so it asks the caller to give a hint about which it wants (--get versus --get-all). But in the single-value case did not behave like the C callbacks, which is what my series fixes.

Using "git config -l" is more like asking the config machinery to just feed us everything, which is what the C callbacks see. Which is more flexible, but way less convenient for the caller. But it doesn't need to be fixed, since the caller has all the information to implement whatever semantics they like.

-Peff