Cache language autocomplete inputs/endpoints #9112
Conversation
RayBB
added
Needs: Review
|
Note: This will cache for all autocompletion endpoints |
cdrini
removed
the
Needs: Review
cdrini
changed the title
|
This might be non-ideal in certain cases ; e.g. when new works are created. |
cdrini
changed the title
| @@ -96,6 +96,7 @@ class languages_autocomplete(delegate.page): | |||
| def GET(self): | |||
| i = web.input(q="", limit=5) | |||
| i.limit = safeint(i.limit, 5) | |||
| web.header("Cache-Control", "max-age=%d" % (24 * 3600)) | |||
There was a problem hiding this comment.
There is little value in caching this ; it's very unlikely a user will type the same thing twice! In general, it's not super likely a user will type the same thing many times. I imagine a veeeeeery tiny percentage of user searches will be cache hits; most will be misses. I think we might want to roll this back, and cache it only in the case we care about.
There was a problem hiding this comment.
@cdrini this comes from a librarian commenting of the lag that happens when searching languages. I don't know what exactly they're editing that they see it often enough to comment.
How about after the deploy goes how to we check how many requests this endpoint is getting on average? Also we can follow up with that librarian to see what their use case it.
There was a problem hiding this comment.
Btw this caching works across pages so if someone edits one book to add a language the next time they do that on any book it'll use the cache.
|
@cdrini actually, now I realize the title is wrong and there's a bit of a misunderstanding. This is only for the language auto complete endpoint. Doesn't effect works or authors search. |
RayBB
changed the title
|
Ah my bad! The JS change affects all endpoints, but it looks like we don't cache this endpoint in any way, so the timestamp cache-bust serves no purpose and can be removed. And yep it looks like the python changes is only for the languages endpoint. My bad! |
Closes #9111
Technical
Removes
timestampbecause it is not used and would mess up caching.Testing
Compare the speed of language on prod and testing for a work like
As far as I can tell, it is drastically faster.
Screenshot
Stakeholders