Page MenuHomePhabricator

Implement new Content Translation data collection for Section Translation
Open, HighPublic

Description

We want to instrument the section translation workflow using the new content_translation_event schema. For instructions and documentation, see the parent task (T281973).

Doing focused QA is important for this task, since issues in the events we capture will result in unreliable/misleading data.

Event Timeline

Pginer-WMF raised the priority of this task from Medium to High.Apr 30 2021, 9:58 AM
nshahquinn-wmf renamed this task from Implement instrumentation support for Section Translation to Implement new Content Translation data collection for Section Translation.May 5 2021, 12:32 PM

Change 685760 had a related patch set uploaded (by Santhosh; author: Santhosh):

[mediawiki/extensions/ContentTranslation@master] Section translation Eventlogging instrumentation

https://gerrit.wikimedia.org/r/685760

The schema need some changes to accommodate anonymous users too. Section translation will support anonymous translation(T278942). Waiting for @nshahquinn-wmf for that.

Change 685760 merged by jenkins-bot:

[mediawiki/extensions/ContentTranslation@master] Section translation Eventlogging instrumentation

https://gerrit.wikimedia.org/r/685760

The schema need some changes to accommodate anonymous users too. Section translation will support anonymous translation(T278942). Waiting for @nshahquinn-wmf for that.

I've uploaded a patch for that. Now it just needs to be code reviewed and merged; that should happen within a day or two.

Change 685760 merged by jenkins-bot:

[mediawiki/extensions/ContentTranslation@master] Section translation Eventlogging instrumentation

https://gerrit.wikimedia.org/r/685760

@santhosh I can tell this is a rough initial implementation, but I did notice that you're using the local edit count for user_global_edit_count. I still really want us to use the global edit count. I know this will require some extra network activity, which is not ideal, but it's just a single API call (to the globaluserinfo module) which returns about 1 KB of JSON. I just tested the current SX dashboard, and the total transferred size was about 1,000 KB so this price seems very small.

@nshahquinn-wmf Performance issue is not related to how much data the API fetch, but the network dependent API request time, how many such API requests are happening in parallel(which the user just opened the application) that uses up users's bandwidth in parellel.

I get your requirement for global edit count. I will explore ways to have this in application with zero impact on userfacing features.

I get your requirement for global edit count. I will explore ways to have this in application with zero impact on userfacing features.

Thank you, I appreciate that!

Change 697919 had a related patch set uploaded (by Santhosh; author: Santhosh):

[mediawiki/extensions/ContentTranslation@master] SX Eventlogging: Fetch global edit count for the user

https://gerrit.wikimedia.org/r/697919

Change 697919 merged by jenkins-bot:

[mediawiki/extensions/ContentTranslation@master] SX Eventlogging: Fetch global edit count for the user

https://gerrit.wikimedia.org/r/697919