You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Probably, repository is also important, for instances tracking multiple repos.
And I can't find the link, but I remember talking about how a history_fingerprint filter would be similar to the GET /api/history endpoint. (Maybe that's worth a separate ticket?)
Something to keep in mind as we consider new filters: we have to maintain a DB index for each combination of filters that we let users request, or else the request will return a 500 on Arrow-size data. This will add some disk space to our DB, a bit of extra write time, and more code to maintain. Maybe this is okay. But we should do some due diligence to figure out if it's worth it, versus requiring client-side filtering.
(IMO that index caveat would not apply to a commit_hash filter, since like run_id, the number of results associated with one commit should not grow unbounded over months and months.)
The text was updated successfully, but these errors were encountered:
Today the GET /api/benchmark-results/ endpoint has the following filters:
run_id
earliest_timestamp
latest_timestamp
run_reason
Others have been requested in the past:
#786 (comment)
#786 (comment)
#746 (comment)
#746 (comment)
Probably,
repository
is also important, for instances tracking multiple repos.And I can't find the link, but I remember talking about how a
history_fingerprint
filter would be similar to the GET /api/history endpoint. (Maybe that's worth a separate ticket?)Something to keep in mind as we consider new filters: we have to maintain a DB index for each combination of filters that we let users request, or else the request will return a 500 on Arrow-size data. This will add some disk space to our DB, a bit of extra write time, and more code to maintain. Maybe this is okay. But we should do some due diligence to figure out if it's worth it, versus requiring client-side filtering.
(IMO that index caveat would not apply to a
commit_hash
filter, since likerun_id
, the number of results associated with one commit should not grow unbounded over months and months.)The text was updated successfully, but these errors were encountered: