Re: [PATCH/RFC] parse-options: add facility to make options configurable

From: Jeff King <peff@peff.net>
Date: Mar 28, 2017, 05:17 UTC
Message-ID: <20170328051723.h2xi2do6iclm64pi@sigill.intra.peff.net>
In-Reply-To: <CACBZZX6=_Jh-emAr=g1-VQwgA4MnDpu=zSOqPK5QHAa7uef_LQ@mail.gmail.com>

On Sat, Mar 25, 2017 at 11:32:02PM +0100, Ævar Arnfjörð Bjarmason wrote:

Show 11 quoted lines

> > So hopefully it's clear that the two are functionally equivalent, and
> > differ only in syntax (in this case we manually decided which options
> > are safe to pull from the config, but we'd have to parse the options.log
> > string, too, and we could make the same decision there).
> 
> I like the simplicity of this approach a lot. I.e. (to paraphrase it
> just to make sure we're on the same page): Skip all the complexity of
> reaching into the getopt guts, and just munge argc/argv given a config
> we can stick ahead of the getopt (struct options) spec, inject some
> options at the beginning if they're in the config, and off we go
> without any further changes to the getopt guts.

Yep, I think that's an accurate description.

Show 21 quoted lines

> There's two practical issues with this that are easy to solve with my
> current approach, but I can't find an easy solution to using this
> method.
> 
> The first is that we're replacing the semantics of:
> 
> "If you're specifying it on the command-line, we take it from there,
> otherwise we use your config, if set, regardless of how the option
> works"
> 
> with:
> 
> "We read your config, inject options implicitly at the start of the
> command line, and then append whatever command-line you give us"
> 
> These two are not the same. Consider e.g. the commit.verbose config.
> With my current patch if have commit.verbose=1 in your config and do
> "commit --verbose" you just end up with a result equivalent to not
> having it in your config, but since the --verbose option can be
> supplied multiple times to increase verbosity with the injection
> method you'd end up with the equivalent of commit.verbose=2.

Right, for anything where multiple options are meaningful, they'd have to give "--no-verbose" to reset the option. In a sense that's less friendly, because it's more manual. But it's also less magical, because the semantics are clear: the config option behaves exactly as if you gave the option on the command line. So for an OPT_STRING_LIST(), you could append to the list, or reset it to empty, etc, as you see fit.

But I do agree that it's more manual, and probably would cause some confusion.

> I can't think of a good way around that with your proposed approach
> that doesn't essentially get us back to something very similar to my
> patch, i.e. we'd need to parse the command-line using the options spec
> before applying our implicit config.

Yes, the semantics you gave require parsing the options first. I think it would be sufficient to just give each "struct option" a "seen" flag (rather than having it understand the config mechanism), having parse_options() set the flag, and then feeding the result to a separate config/cmdline mapping mechanism. That keeps the complexity out of the options code.

It does tie us back in to requiring parse-options, which not all the options use.

In a lot of cases that "seen" flag is effectively a sentinel value in whatever variable the option value is stored in. But some of the options don't have reasonable sentinel values (as you noticed with the "revert -m" handling recently).

> The second issue is related, i.e. I was going to add some flag an
> option could supply to say "if I'm provided none of these other
> maybe-from-config options get to read their config". This is needed
> for hybrid plumbing/porcelain like "git status --porcelain".

Yeah, I agree you can't make that decision until you've seen the command-line options. I think we currently do some hairy stuff where we speculatively read config into a variable, and then apply the config-based defaults only when we know we're in non-porcelain mode (see status_deferred_config in builtin/commit.c).

That came about because we didn't want to parse the config a second time. These days the deferred config should probably just be read from the cached configset, after we've read the other options.

But I think this can be done after the full option-parsing is finished by applying the mapping then. I.e., something like:

    parse_options(argc, argv, options, ...);
    if (status_format != STATUS_FORMAT_PORCELAIN)
	apply_config_mapping(status_mapping, options);

-Peff