[PATCH 5 of 7] revset: introduce a filterrevs function

[Matt Mackall]

On Mon, 2012-06-04 at 15:25 -0700, Bryan O'Sullivan wrote:
> # HG changeset patch
> # User Bryan O'Sullivan <bryano at fb.com>
> # Date 1338848634 25200
> # Node ID 1fda5b47f842b8b57ee866684aaabcc1052a06e1
> # Parent  115c3e5c708813a9800dfe4938a7c9202a4d00ba
> revset: introduce a filterrevs function
> 
> For large revsets, our pattern of always creating new lists is
> expensive.  This function tries to avoid unnecessary list duplication.
> 
> As an example, here's performance on a kernel repo with "hg
> perfrevrange 2222::233333".
> 
> Before: 0.572 sec
> After:  0.461 sec (20% better)
> 
> diff --git a/mercurial/revset.py b/mercurial/revset.py
> --- a/mercurial/revset.py
> +++ b/mercurial/revset.py
> @@ -84,6 +84,21 @@ def revsubset(subset):
>          return set(subset)
>      return _revinterval(subset)
>  
> +def filterrevs(revs, subset, contig=False):

Huh? I'm not sure whether contig applies to revs or subset.

> +    if not revs:
> +        return revs
> +    subset = revsubset(subset)
> +    try:
> +        if contig or isinstance(revs, (_revinterval, xrange)):
> +            lo, hi = sorted((revs[0], revs[-1]))

I guess it applies to revs.

> +        else:
> +            lo, hi = min(revs), max(revs)
> +        if subset._start <= lo and subset._stop > hi:
> +            return revs
> +    except (AttributeError, TypeError, KeyError):
> +        pass
> +    return filter(subset.__contains__, revs)
> +
>  def _revancestors(repo, revs, followfirst):
>      """Like revlog.ancestors(), but supports followfirst."""
>      cut = followfirst and 1 or None
> @@ -296,8 +311,7 @@ def dagrange(repo, subset, x, y):
>      if subset:
>          r = revsubset(repo)
>          xs = _revsbetween(repo, getset(repo, r, x), getset(repo, r, y))

How do we know xs is a continuous interval? I don't think it is
guaranteed?

> -        s = revsubset(subset)
> -        return [r for r in xs if r in s]
> +        return filterrevs(xs, subset, contig=True)
>      return []
>  
>  def andset(repo, subset, x, y):


-- 
Mathematics is the supreme nostalgia of our time.