On 02/25/2013 08:48 PM, Kai Willadsen wrote:
You're talking about your suggestion of doing an input split followed by a double compare? In that case, yes, you should always get the sync point included. I was talking about a different option entirely of simply forcing the algorithm through that point. Since the result should be almost identical (bar some heuristic speed-ups maybe?), I think doing whatever is easiest is the best plan.
I am attaching a patch with SyncPointMyersSequenceMatcher class that does the split & merge as described before. There are also some unit tests included (at last). I have no idea what the UI should look like hence I did not even try to implement anything but the 'backend' stuff. Cheers, Piotr
Attachment:
0001-Add-SyncPointMyersSequenceMatcher-and-matcher-relate.patch
Description: Text Data