Re: [PATCH] config: add an includeIf.env{Exists,Bool,Is,Match}

From: Jeff King <peff@peff.net>
Date: Sep 27, 2021, 20:15 UTC
Message-ID: <YVImeFHxY7hmb3wY@coredump.intra.peff.net>
In-Reply-To: <xmqqo88eq8um.fsf@gitster.g>

On Mon, Sep 27, 2021 at 09:30:41AM -0700, Junio C Hamano wrote:

Show 7 quoted lines

> >> This asserts what? FOO=" bar"?
> >
> > Whoops, that should have been "envIs", asserting that $FOO contains
> > "bar".
> 
> Oh, "can we check with a literal with leading whitespace?" was what
> my question was about ;-)

My assumption was that nobody would really care about doing so. It is true that it's less flexible, though (and is a decision we can't easily take back later).

Show 12 quoted lines

> > As I said, I think it matters more with the infix operators, as:
> >
> >   [includeIf "env:FOO == bar"]
> >
> > is more readable than:
> >
> >   [includeIf "env:FOO==bar"]
> 
> Sure, but at that point, we'd probably want some quoting mechanism
> for the literal to be compared, e.g.
> 
> 	[includeIf "env:PATH ~= \"(:|^)/usr/bin(:|$)\""]

Ick. The extra quoting of the internal double-quotes is pretty horrid to look at. Also, how does one match a double-quote in the value? \\\"?

If it were optional, that would make the common cases easy (no dq, no whitespace), and the hard ones possible.

I think this is getting into a bit of a digression, though. I'm willing to defer to Ævar, who is doing the actual work, and I don't know if he has found any of this compelling. ;)

Show 12 quoted lines

> > But I do think:
> >
> >   [includeIf "envIs:FOO:bar"]
> >
> > is harder to read than even:
> >
> >   [includeIf "envIs:FOO: bar"]
> 
> Hmph, that's quite subjective, I am afraid.  When I see the latter
> in the configuration file, "do I have to have a single space before
> 'bar' in the value of $FOO" would be the first question that would
> come to my mind.

I think it's just the mashed-up colons that I find ugly in the first one. But I agree the latter isn't that nice either, and introduces the ambiguity you describe.

> With an understanding that our syntax is so limited that we cannot
> even write '=' and need to resort to Is: instead, I'd actually find
> that the former less confusing than the latter.

That I think is the most interesting question: is the "=" actually out-of-bounds? I tend to think not, based on our responses earlier in the thread.

-Peff