There should be a dump that contains all redirects, one redirect per line (similar to the entity dumps).
In JSON, such a line would probably look like {"from":"Q123","to":"Q456"}.
In n-triples, it would be wd:Q123 owl:sameAs wd:Q456.
This would allow consumers of our JSON dumps to resolve redirects.
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Allow creating redirect only dumps with dumpRdf | mediawiki/extensions/Wikibase | master | +39 -2 |
| Status | Subtype | Assigned | Task | |
|---|---|---|---|---|
| · · · | ||||
| Open | None | T98039 Include incoming redirects in JSON representation of entities (tracking) | ||
| Open | None | T88991 improve Wikidata dumps [tracking] | ||
| Open | None | T46581 Partial dumps | ||
| Stalled | None | T94019 Generate RDF from JSON | ||
| Open | None | T98320 [Task] Create dump of entity redirects (JSON or n-triples) | ||
| · · · |
a separate would avoid breaking the assumption of "one line, one entitiy" for the current dumps
Reworded the task in accordance to what we discussed this week.
This is actionable now.
I'm not sure we really need a JSON dump of redirects. We could directly generate an ntriples dump instead, perhaps even using the existing RDF dump script.
One reason to provide such a dump is for WikiDataToolkit to be able to handle redirects.
Change 354470 had a related patch set uploaded (by Hoo man; owner: Hoo man):
[mediawiki/extensions/Wikibase@master] Allow creating redirect only dumps with dumpRdf
Change 354470 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] Allow creating redirect only dumps with dumpRdf
Hm… doing these redirect dumps separate from the current dumps will mean that the complete dumps might be inconsistent (as both aspects are dumped at separate times).
Maybe we should change dumpJson.php so that it emits both the redirect and the "usual" JSON side-by-side. Or we just include it in the original JSON?