Fix extraction when there are multiple reverts #127
Conversation
|
Comparing the ptwiki datasets before and after this change, using this gist, shows that these are the edits which will not be added to the dataset after the changes:
On the other hand, these are the edits which will be added (and were not in the previous version): |
| if verbose: | ||
| sys.stderr.write("r") | ||
| sys.stderr.flush() | ||
| else: | ||
| # This revision is not a revert. Get the new labels | ||
| # FIXME: what if last_labels contains labels from one of | ||
| # reverted edits? | ||
| new_labels = project_labels - last_labels |
There was a problem hiding this comment.
Since we're storing every single revision in revisions, we could possibly also store the labels and take a second pass looking for revisions that are not reverted.
* Invert the reverted status recursively * Add tests for the new behavior
he7d3r
force-pushed
the
extractor-and-reverts
branch
from
c3b9b31
to
2476d22
Compare
Codecov Report
@@ Coverage Diff @@
## master #127 +/- ##
==========================================
+ Coverage 52.07% 52.21% +0.13%
==========================================
Files 47 47
Lines 1373 1377 +4
==========================================
+ Hits 715 719 +4
Misses 658 658
Continue to review full report at Codecov.
|
|
Merged as part of #129 |
See https://phabricator.wikimedia.org/T252152.