Caching on heavily-used API calls: srch/api/profiles

[milaaraujo]

Fixes #3149

[plotsbot]

	1 Message
📖	@milaaraujo Thank you for your pull request! I’m here to help with some tips and recommendations. Please take a look at the list provided and help us review and accept your contribution! And don’t be discouraged if you see errors – we’re here to help.

Generated by 🚫 Danger

[jywarren]

Great. I'll roll this out today and we can see how it does!

[jywarren]

Strangely i don't see a lot of tracking in Skylight: https://oss.skylight.io/app/applications/GZDPChmcfm1Q/recent/6h/endpoints/Srch::Search%20%5BGET%5D%20srch%2Fprofiles

Not sure if this is because these calls aren't monitored?

[milaaraujo]

Depending on how it does we can implement the caching for the other endpoints...

[jywarren]

https://publiclab.org/api/srch/profiles?query=jywa&sort_by=recent&field=username&_=1549385453136 is about 1.9 seconds right now.

[jywarren]

https://publiclab.org/api/srch/profiles?query=sidha&sort_by=recent&field=username&_=1549385453136 is 2.4 seconds

[jywarren]

https://publiclab.org/api/srch/profiles?query=camila&sort_by=recent&field=username&_=1549385453136 1.8 seconds

https://publiclab.org/api/srch/profiles?query=mila&sort_by=recent&field=username&_=1549385453136 2.48 seconds

[jywarren]

Wow, ok so i did a hard cache flush and reloaded this:

https://publiclab.org/api/srch/profiles?query=camila&sort_by=recent&field=username&_=1549385453136

Initially 6 seconds, but on repeating, 578ms

https://publiclab.org/api/srch/profiles?query=sidha&sort_by=recent&field=username

2 seconds, then on refresh 553 ms

https://publiclab.org/api/srch/profiles?query=jywa&sort_by=recent&field=username

2 seconds then 80 ms

so, i think this worked pretty well. Let me try actually using it in a comment...

[jywarren]

And a longer interactive session here:

You can see how typing @warren 3 times, the later ones show much much lower load times; the initial ones are at 2-4 seconds, the later at 68-631ms, although one was 2.4s

This is a big improvement, but I do want to note how many requests go out -- i think we can change the settings on the typeahead to refine this too. See a gif:

Also note it's sending lots of requests where query is empty. We should adjust so it only sends when there are at least 2.

[jywarren]

Copying this into #4670