RE: svn commit: r1091262 - /subversion/trunk/subversion/libsvn_wc/wc_db.c

> -----Original Message-----
> From: Johan Corveleyn [mailto:jcorvel_at_gmail.com]
> Sent: dinsdag 12 april 2011 15:27
> To: Hyrum Wright
> Cc: C. Michael Pilato; Bert Huijben; Greg Stein; dev_at_subversion.apache.org
> Subject: Re: svn commit: r1091262 -
> /subversion/trunk/subversion/libsvn_wc/wc_db.c

> Changelists in IDE's are used a lot, for separating work. If you do a
> major refactoring, touching say 2000 files, and you want those to be
> part of some changelist ... Also, when we do a sync-merge of a feature
> branch, we usually let those merge-changes appear in a changelist
> (makes it easy to commit with a nicely prepared commit message). If
> those syncs are more than a week apart, the amount of changed files
> can easily go to 2000-3000.
>
> I can't comment on the technical merits/problems of adapting
> changelists for more optimal wc-ng use (I get the impression that the
> effort currently outweighs the gains), but just wanted to add my 0.02
> €: adding thousands of files to a changelist is not unusual when
> working from an IDE.

I use the same system in AnkhSVN, but then again when I change the changelists in Subversion I don't use any of the recursive operations: I apply the changelist changes as a batch of paths.
(It is very expensive to calculate whether I could use a recursive set on a path and everything below).

For my use case optimizing for batch processing is more important then for depth processing. (The current code optimizes for depth processing)

And as we don't have proper changelist support on directories yet, some operations ignore directories when you use changelists (Diff & revert work this way if I remember correctly), while others apply the same change to all directories within the original filter (commit as a notable example).

        Bert
Received on 2011-04-12 15:41:13 CEST