Open pauldambra opened 1 year ago
oh, interesting! I think it's not event uuids, but person distinct IDs. There's some optimisation converting the 'personID' to a list of distinct IDs (so no joins in the SQL), but this doesn't work when the person has two many distinct IDs
oh, interesting! I think it's not event uuids, but person distinct IDs.
Doh, late Friday reading skills š
here's where a query for a person id is converted to a list of distinct ids to query directly...
but that list can be too large to fit in a query, and should probably be a subquery š¤·
Historic context: this behaviour was copied over from the pre-hogql "live events" page.
We need the properties and distinct id-s of a person to show the nice "(-_-) user@host.com" person display. Getting them with a subquery made the page much slower. The timings have probably changed by now, but we're talking about something like 3sec -> 13sec for loading the event explorer default view.
I'd love it if we didn't have to do this fumbling though... š¤
When viewing a session recording we show users the associated events.
To do that we use a hog event query
A user complained that this was failing for them intermittently. And I noticed the generated SQL is passing a list of event UUIDs that is (sometimes) too large to be passed to the query
This has been reported again
When filtering by a person we lookup the person in django and then add their distinct ids to the query. But the list of distinct ids can be way too long to fit into a query. So CH errors and we can't load events for the recording