In some cases the algorithm seems to mark whole paragraphs as changed even though just a word was added/removed.
e.g. see
Description
Description
Details
Details
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Improve diff quality to match TableDiffFormatter | mediawiki/extensions/TwoColConflict | master | +30 -63 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T72163 Edit conflicts (tracking) | |||
| Resolved | None | T139601 Better solutions to edit conflicts (#1) | |||
| Resolved | WMDE-Fisch | T161784 Review diff formater algorithm. |
Event Timeline
Comment Actions
Change 347589 had a related patch set uploaded (by WMDE-Fisch):
[mediawiki/extensions/TwoColConflict@master] Improve diff quality to match TableDiffFormatter
Comment Actions
Some custom modifications on how the diff coming from the PHP diff engine is interpreted made the result in the TwoColumnConflict extension look worse then in the original table based diff view.
The above patch fixes that and should improve the results immensely and probably addresses the issue shown above.
A comparison between the PHP and C++ diff engine was not done for now. We might want to create a new ticket for that case.
Comment Actions
Change 347589 merged by jenkins-bot:
[mediawiki/extensions/TwoColConflict@master] Improve diff quality to match TableDiffFormatter