First a little update on the search behaviour on Yarn. We see that the operations seem to swing a little more than they did before the update, but that’s probably because the number is so much bigger now.
We see that the average lays a bit below 40 000 searches a day, that’s mostly because the feature is way more discoverable now, and has a lot of visibility. I do think that because the detail page isn’t finished yet, it’ll be hard to keep this amount of usage.

This week I worked mostly on the detail page (no PR open for that yet), in which I used most of the metadata currently in the Algolia index. I’m still unsure which method will be used to fetch the README and the CHANGELOG, a lot of change that I’ll add that to the Algolia index as non-searchable, but retrievable attribute, which then can be parsed clientside.
Some extra data will have to be retrieved from GitHub at runtime, because they have a limit of 60 requests per minute, we can’t ask for extra metadata (activity, stars …) at indexing time, it would delay the indexing by too much. I hope to find a way that uses the current logged in user to do those requests, since a logged in user can do basically as much requests as needed.
Something else small I worked on is the integration of localisation into the docsearch instance of Yarn (#395, #397, #130, #132 ). What needed to happen is the following:
- add logic to the docsearch crawler to fetch a language, default en (#130)
- add configuration to Yarn to be able to fetch only the current language (#395 and #397)
- enable the other languages in docsearch (#132)
One of the funniest bugs I managed to fix this week was #2026, in which someone asked to fix their bug with icons. It didn’t seem very obvious since the searchbox of instantsearch.js (in contrast to react-instantsearch) doesn’t have icons. At first I didn’t understand what the bug actually was, so I asked to make a jsfiddle. Then I looked at their website and recognised the search box of the bug report (a site completely in Chinese by the way). I inspected their configuration and saw this

This is pretty recognisable as font awesome, which they didn’t include in their page. In the end the bug didn’t have to do with Algolia, but I love how with just a few moments of digging through what the actual problem is, you can solve real things for people.
This week I also filled in two slots during support, which mostly means dispatching into the correct mailbox and answering questions you can.