[PATCH 8 of 8] fileset: add a 'status(...)' predicate to control evaluation context

[Yuya Nishihara]

On Sun, 5 Mar 2017 14:18:06 +0100, Pierre-Yves David wrote:
> On 03/05/2017 12:49 PM, Yuya Nishihara wrote:
> > On Fri, 03 Mar 2017 14:40:45 +0100, Pierre-Yves David wrote:
> >> # HG changeset patch
> >> # User Pierre-Yves David <pierre-yves.david at ens-lyon.org>
> >> # Date 1488546500 -3600
> >> #      Fri Mar 03 14:08:20 2017 +0100
> >> # Node ID 3e95bf7ed38189f6f56de89571fffc667280bb56
> >> # Parent  9d6e733046b9aa7e2ded8c4207625fedcc2a8c04
> >> # EXP-Topic filesetrev-func
> >> # Available At https://www.mercurial-scm.org/repo/users/marmoute/mercurial/
> >> #              hg pull https://www.mercurial-scm.org/repo/users/marmoute/mercurial/ -r 3e95bf7ed381
> >> fileset: add a 'status(...)' predicate to control evaluation context
> >
> > The series looks generally good, but I don't queue as the first half was
> > originally written by me. Thanks for bringing my hack into production.
> >
> >> --- a/mercurial/fileset.py
> >> +++ b/mercurial/fileset.py
> >> @@ -463,6 +463,25 @@ def revs(mctx, x):
> >>                  result.append(f)
> >>      return result
> >>
> >> + at predicate('status(base, rev, pattern)')
> >> +def status(mctx, x):
> >> +    """``status(base, rev, revspec)``
> >> +
> >> +    Evaluate predicate using status change between ``base`` and
> >> +    ``rev``. Examples:
> >> +
> >> +    - ``status(3, 7, added())`` - matches files added from "3" to "7"
> >> +    """
> >> +    repo = mctx.ctx.repo()
> >> +    # i18n: "status" is a keyword
> >> +    b, r, x = getargs(x, 3, 3, _("status takes three arguments"))
> >> +    # i18n: "status" is a keyword
> >> +    baserevspec = getstring(b, _("first argument to revs must be a revision"))
> >> +    revspec = getstring(r, _("second argument to revs must be a revision"))
> >
> > s/revs/status/
> 
> Good catch, Do you want a V2 or a followup ?

I'm not the one who will queue this series, but perhaps this can be fixed in
flight.

> >> +    basectx = scmutil.revsingle(repo, baserevspec)
> >> +    ctx = scmutil.revsingle(repo, revspec)
> >
> > Perhaps it's better to reject an empty revspec string.
> 
> Revsingle natively refuse empty revset, (but we should add a test for it).

I meant status('', '', pattern), not an empty result.

> > And I slightly prefer picking the first rev for basectx just like
> > revset.rangeset() would do.
> 
> I do not really like the idea. It seems strange to me to have the same 
> revset interpreted as a differet revs when pased as 'base' or 'rev'.

Okay, I don't have a strong opinion about this. But I note that it's the common
behavior where revision pair is required. Try "hg status --rev : --rev :" for
example.