I have just pushed a new version of the TitleValue patch to Gerrit: https://gerrit.wikimedia.org/r/106517.
I have also updated the RDF to reflect the latest changes: https://www.mediawiki.org/wiki/Requests_for_comment/TitleValue.
Please have a look. I have tried to address several issues with the previous proposal, and reduce the complexity of the proposal. I have also tried to adjust the service interfaces to make migration easier.
Any feedback would be very welcome!
-- daniel
On 02/26/2014 12:23 PM, Daniel Kinzler wrote:
I have just pushed a new version of the TitleValue patch to Gerrit: https://gerrit.wikimedia.org/r/106517.
I have also updated the RDF to reflect the latest changes: https://www.mediawiki.org/wiki/Requests_for_comment/TitleValue.
Please have a look. I have tried to address several issues with the previous proposal, and reduce the complexity of the proposal. I have also tried to adjust the service interfaces to make migration easier.
Any feedback would be very welcome!
-- daniel
Update: https://gerrit.wikimedia.org/r/#/c/106517/ has now been merged. To quote the commit message:
The intention is to improve testability and reusability; these advantages will however only become apparent with further refactoring of the respective special pages. This will be done in follow-up changes.
More work will be needed to migrate essential functionality from Title and Linker classes into the respective service classes, MediaWikiTitleCodec and MediaWikiPageLinkRenderer.
On 03/31/2014 06:14 PM, Sumana Harihareswara wrote:
On 02/26/2014 12:23 PM, Daniel Kinzler wrote:
I have just pushed a new version of the TitleValue patch to Gerrit: https://gerrit.wikimedia.org/r/106517.
I have also updated the RDF to reflect the latest changes: https://www.mediawiki.org/wiki/Requests_for_comment/TitleValue.
Please have a look. I have tried to address several issues with the previous proposal, and reduce the complexity of the proposal. I have also tried to adjust the service interfaces to make migration easier.
Any feedback would be very welcome!
-- daniel
Update: https://gerrit.wikimedia.org/r/#/c/106517/ has now been merged. To quote the commit message:
The intention is to improve testability and reusability; these advantages will however only become apparent with further refactoring of the respective special pages. This will be done in follow-up changes.
More work will be needed to migrate essential functionality from Title and Linker classes into the respective service classes, MediaWikiTitleCodec and MediaWikiPageLinkRenderer.
I asked Daniel what we need to do next on TitleValue and he said:
Sorry for not getting back to you on this sooner. It's a bit tricky to say what the next steps would be. There is several directions we could take this. But none of them fit naturally with my Wikidata work, so I can only try to squeeze in bits and pieces every now and then. Would be great to get more people in on this.
Anyway, here is what could be done:
- Using TitleValue in more places. There's a refactoring of CategoryLinks [1]
that I still need to update and rebase, and more special pages and API modules could follow this approach. This would improve testability, in particular.
- Contimue wrapping up lose ends left by the TitleValue refactor. The code for
splitting a title into its components needs to be refactored and tidied, in particular. This also means writing lots of tests for the current behavior of Title, so we don't change it.
- factor more things out of the Title class. TitleValue is just the start. We
could factor out the logic for permission checks into a separate "service"; Also, we could start defining value objects for different kinds of links (internal, interwiki, external, etc) on top of/in addition to TitleValue. Note that the conflation of "title" and "link target" is the cause of quite a bit of confustion.
- Refactor other "nasty" parts of MediaWiki. Some prime candidates are Linker,
User, and WikiPage. Actually, ContentHandler could also use an overhaul: Content should be a plain value interface, all "smart" aspects should be factored into services.
So much to code, so little time :)
This kind of lots-of-little-refactorings work feels like a good candidate for a sprint at the Zurich hackathon, where people who don't know this codebase very well can help out. Daniel, do you agree?
wikitech-l@lists.wikimedia.org