jayvdb added a comment.
In https://phabricator.wikimedia.org/T85336#960707, @Nemo_bis wrote:
In https://phabricator.wikimedia.org/T85336#958980, @jayvdb wrote:
TWN tools can't export all message contributors into a new JSON message file? (i.e. editors of the page on TWN)
The full editors list is not a list of "all message contributors", but of all *potential* message contributors. Including, say, a vandal who got reverted. If https://phabricator.wikimedia.org/T4994 was fixed, then Translate could be smarter (there was some research progress, so it might happen).
It is _always_ better to over-attribute than under-attribute :/
proper reverts are able to be eliminated with very minimal fuss; that isnt a research problem. And on a small wiki like TWN, and since it works on discrete units of information, that will eliminate 90+% of problematic attributions.
Add a default enabled block button checkbox "exclude user from translation attribution", and another 8% is solved. ;-) And nobody cares about the last 2% of false positives/
Or, more bluntly, do we need to migrate the old translation metadata data from the python files to the new JSON files?
Yes. However out of 6 cases, you found 1 (or perhaps 2) where the older editor contributed some *words* of the current sentence... it's not a concrete attribution issue except for the pre-TWN translators.
My count is 4 of 6 if I am being generous, or 5 of 6 if I am not. :/ I am only really excluding case (1), as that problem occurred before the history of the TWN page history. That TheBITLINK doesnt need to be attributed for the es translation doesnt feel right given their translation was used by pywikibot for a year an a quarter.
Note that the "current sentence" for this message is two words long in English and similar length in most other languages. If one word of a translators work remains in the current message, that is 50% of the translation.
Anyway, it sounds like the pywikibot team needs to manually add the list of contributors from TWN to the JSON in order for the attribution to not regress further. i.e. the attribution removals in Xqt's patch to the python need to be reversed into additions to the JSON. Or, it might be more efficient to obtain the TWN contributor list from the TWN wiki for each message page. Luckily we have a tool which can do that .. ;-) I can't help but think we'd be re-inventing the TWN tools.
TASK DETAIL https://phabricator.wikimedia.org/T85336
REPLY HANDLER ACTIONS Reply to comment or attach files, or !close, !claim, !unsubscribe or !assign <username>.
EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/
To: siebrand, jayvdb Cc: Aklapper, valhallasw, Nikerabbit, siebrand, jayvdb, Ladsgroup, Nemo_bis, Xqt, Gryllida, Shizhao, Arrbee, pywikipedia-bugs