[PATCH 1 of 3 V2 STABLE] changelog: introduce count argument to revs() function

[Matt Mackall]

On Tue, 2013-07-23 at 13:22 +0400, Alexander Plavin wrote:
> 2013/7/23 Alexander Plavin <me at aplavin.ru>:
> > 2013/7/23 Brodie Rao <brodie at sf.io>:
> >> On Mon, Jul 22, 2013 at 2:54 AM, Alexander Plavin <me at aplavin.ru> wrote:
> >>> # HG changeset patch
> >>> # User Alexander Plavin <me at aplavin.ru>
> >>> # Date 1374321959 -14400
> >>> #      Sat Jul 20 16:05:59 2013 +0400
> >>> # Node ID 82096c966790ec607efb16ebff2cfa53c643b700
> >>> # Parent  b97ce93726dbcbfc439458e244a266953c22940e
> >>> changelog: introduce count argument to revs() function
> >>>
> >>> This allows specifying count of revisions to yield, which is convenient when
> >>> some revs are filtered.
> >>>
> >>> diff -r b97ce93726db -r 82096c966790 mercurial/changelog.py
> >>> --- a/mercurial/changelog.py    Fri Jul 19 21:26:08 2013 +0400
> >>> +++ b/mercurial/changelog.py    Sat Jul 20 16:05:59 2013 +0400
> >>> @@ -147,11 +147,23 @@
> >>>
> >>>          return filterediter()
> >>>
> >>> -    def revs(self, start=0, stop=None):
> >>> +    def revs(self, start=0, stop=None, count=None):
> >>>          """filtered version of revlog.revs"""
> >>> +        if count is not None:
> >>> +            if count < 0:
> >>> +                stop = 0
> >>> +            elif count > 0:
> >>> +                stop = len(self)
> >>> +            else:
> >>> +                stop = start
> >>> +
> >>> +        curcount = 0
> >>>          for i in super(changelog, self).revs(start, stop):
> >>>              if i not in self.filteredrevs:
> >>>                  yield i
> >>> +                curcount += 1
> >>> +                if count is not None and curcount >= abs(count):
> >>> +                    break
> >>
> >> Instead of making this change inside changelog.revs(), couldn't you
> >> use itertools.islice() where you need to limit the number of
> >> revisions?
> >
> > I thought about such variant, but this one (from this patch) seemed
> > more convenient to use in the calling place and more universal.
> > ('revs(start, -count)' vs 'itertools.islice(revs(start, 0), count)').
> > If anyone else agrees that with islice it will be better, I will
> > change and resend.
> 
> Also, (forgot to add to the previous message) it relies on the
> function returning an iterator, but not a list of requested revisions
> and uses this knowledge (if it was returning list,
> 'itertools.islice(revs(start, 0), count)' would be very inefficient).

I've got bad news for you:

a) Trying to change changelog.revs just for hgweb is not terribly
desirable, especially on stable.
b) I have an irrational hatred of itertools

Just open-code it. Probably with changelog.hasnode.

-- 
Mathematics is the supreme nostalgia of our time.