On 8/14/06, Albert Jin <albert.jin@gmail.com> wrote:
>
>
> Hi,
>
> I worked on a diff algorithm a lot recently (checkout the development code
> at http://xindiff.cvs.sourceforge.net/). I tried different
> implementations on this two files, (packed as follows)
>
> http://xindiff.sourceforge.net/diff_t5.zip (t5.txt: 467354
> lines, t6.txt: 334422 lines)
>
> XinDiff takes about 3 minutes and finds out that the LCS count is 22926 and
> output html diff view, while the one using libsvn_diff
> (http://subversion.tigris.org/tools_contrib.html#diff)
> takes 30 minutes and does seems to get through. I suspect that there could
> be a dead loop.
>
> Note that XinDiff does perfect LCS calculation and works as GNU diff with
> --minimal option.
It's certainly possible that there's a bug in our diff implementation,
but it seems more likely that it's just spending an awfully long time
trying to find the minimal diff. I know we've seen problems like that
before, where very large files take a disturbingly long time to diff.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Aug 14 19:09:55 2006