git@vger.kernel.org mailing list mirror (one of many)
 help / Atom feed
From: Jeff King <peff@peff.net>
To: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Cc: Git Mailing List <git@vger.kernel.org>, Junio C Hamano <gitster@pobox.com>, "brian m . carlson" <sandals@crustytoothpaste.net>, Matthieu Moy <Matthieu.Moy@grenoble-inp.fr>, Duy Nguyen <pclouds@gmail.com>, Brandon Williams <bmwill@google.com>
Subject: Re: [PATCH/RFC] parse-options: add facility to make options configurable
Date: Tue, 28 Mar 2017 01:17:23 -0400
Message-ID: <20170328051723.h2xi2do6iclm64pi@sigill.intra.peff.net> (raw)
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:

> > 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.

> 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

  reply index

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-03-19  9:57 Add configuration options for some commonly used command-line options (Was: [RFH] GSoC 2015 application) Duy Nguyen
2017-03-19 10:15 ` Add configuration options for some commonly used command-line options Matthieu Moy
2017-03-19 13:18   ` brian m. carlson
2017-03-19 13:43     ` Ævar Arnfjörð Bjarmason
2017-03-20 10:56       ` Duy Nguyen
2017-03-20 17:32         ` Brandon Williams
2017-03-20 18:18           ` Jeff King
2017-03-31 19:44             ` Brandon McCaig
2017-03-20 18:56           ` Junio C Hamano
2017-03-20 19:14             ` Jeff King
2017-03-20 21:57             ` Ævar Arnfjörð Bjarmason
2017-03-24 23:10       ` [PATCH/RFC] parse-options: add facility to make options configurable Ævar Arnfjörð Bjarmason
2017-03-25 16:47         ` Ævar Arnfjörð Bjarmason
2017-03-25 21:31           ` Jeff King
2017-03-25 22:32             ` Ævar Arnfjörð Bjarmason
2017-03-28  5:17               ` Jeff King [this message]
2017-03-25 21:28         ` brian m. carlson
2017-03-20 10:42     ` Add configuration options for some commonly used command-line options Duy Nguyen

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20170328051723.h2xi2do6iclm64pi@sigill.intra.peff.net \
    --to=peff@peff.net \
    --cc=Matthieu.Moy@grenoble-inp.fr \
    --cc=avarab@gmail.com \
    --cc=bmwill@google.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=pclouds@gmail.com \
    --cc=sandals@crustytoothpaste.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

git@vger.kernel.org mailing list mirror (one of many)

Archives are clonable:
	git clone --mirror https://public-inbox.org/git
	git clone --mirror http://ou63pmih66umazou.onion/git
	git clone --mirror http://czquwvybam4bgbro.onion/git
	git clone --mirror http://hjrcffqmbrq6wope.onion/git

Newsgroups are available over NNTP:
	nntp://news.public-inbox.org/inbox.comp.version-control.git
	nntp://ou63pmih66umazou.onion/inbox.comp.version-control.git
	nntp://czquwvybam4bgbro.onion/inbox.comp.version-control.git
	nntp://hjrcffqmbrq6wope.onion/inbox.comp.version-control.git
	nntp://news.gmane.org/gmane.comp.version-control.git

 note: .onion URLs require Tor: https://www.torproject.org/
       or Tor2web: https://www.tor2web.org/

AGPL code for this site: git clone https://public-inbox.org/ public-inbox