Re: [PATCH 1/2] blame: large-scale performance rewrite
- From
David Kastrup <dak@gnu.org>
- Date
- Apr 26, 2014, 17:22 UTC
- Message-ID
- <878uqs80pt.fsf@fencepost.gnu.org>
- In-Reply-To
- <CAJo=hJsrFb+FdBDoHiYOPeB0nEQhD+9dBXTK84bDSj1MgBBdQA@mail.gmail.com>
Shawn Pearce <spearce@spearce.org> writes:
Show 9 quoted lines
> Right, and JGit blame still is missing the -M and -C options, as I > have not implemented those yet. I got basic blame and reverse blame > working a few years ago and then stopped working on the code for a > while. Now we have interest in improving the latency for $DAY_JOB, so > I've been poking at the code again for the last week or so. > > But that -M and -C thing is still not implemented, and I know its > going to be tricky to get right with the way the scapegoating is > passed along.
Actually, for -M/-C it would be saner to rip open the whole xdiff blackbox which internally goes to quite some effort in order to produce the linear ordering of a diff, and then -M has to simulate dropping that linear ordering requirement by doing a host of parallel diffs for each chunk.
-- David Kastrup