Add labels in needsdiagnosis endpoint

Question

Add labels in needsdiagnosis endpoint

magsout opened this issue 5 years ago · comments

One thing
that would be very interesting is to pull out the firefox issues from
the needsdiagnosis dashboard. Maybe one line (current approach) for all
issues, and another line for only Firefox issues.

I think we can add labels like we do for needstriage endpoint on webcompat https://webcompat.com/api/issues/category/needstriage ?

Mike Taylor · Answer 1 · Mon May 06 2019 22:55:52 GMT+0800 (China Standard Time)

How many labels?

For me, I think it would be good to know engine-gecko vs everything else. But I can see the main browsers being listed being an interesting data point (maybe?). I just wonder about noise.

Guillaume Démésy · Answer 2 · Mon May 06 2019 23:16:23 GMT+0800 (China Standard Time)

@miketaylr all? for now we just handle firefox and we will see in the future if we want to handle others browsers.

Karl Dubost · Answer 3 · Tue May 07 2019 09:37:11 GMT+0800 (China Standard Time)

@magsout @miketaylr could you clarify a bit. currently needsdiagnosis endpoint doesn't count anything, we give a list of

timestamp
raw number of issues at this time in the needsdiagnosis milestone.

Do you suggest we record the raw number of issues for a specific label?
I don't even know if it's possible. Let's check

The issues endpoint doesn't provide this information even in the HTTP headers.
https://api.github.com/repos/webcompat/web-bugs/issues?milestone=3&labels=engine-gecko
The milestone endpoint doesn't have the notion of parameters for search
https://api.github.com/repos/webcompat/web-bugs/milestones/3

So basically we can't do it except if we count ourselves (go through the pagination), that requires a bit more code and failure handling.

Mike Taylor · Answer 4 · Tue May 07 2019 22:46:39 GMT+0800 (China Standard Time)

Do you suggest we record the raw number of issues for a specific label?

Yeah, that would be ideal, if possible.

Karl Dubost · Answer 5 · Thu May 16 2019 13:45:24 GMT+0800 (China Standard Time)

I'm trying to figure out this one.

The repository issues events list is limited in pagination. For example, as of today, https://api.github.com/repositories/17914657/issues/events?page=1333 the oldest event is 3 months old. 2019-02-25T19:27:58Z All events prior to this are lost.
On the other hand, the events list for all individual issues seems to be one or two pages long. Example https://api.github.com/repositories/17914657/issues/20000/events
Webhook: There is a notion of issues events for webhook https://developer.github.com/v3/activity/events/types/#issuesevent with keywords milestoned and demilestoned. We can imagine that when the issue enter in milestone needsdiagnosis, the label for browser-* or engine-gecko have been already set. So we could have a webhook listening to these event of coming in needsdiagnosis and going out of needsdiagnosis, then fetch the issue description to know its labels. Not necessary very cool and relies on successful fetch at each webhook POST.

ok maybe the solution is to have a single request every x hours to https://api.github.com/search/issues?q=is:open+is:issue+milestone:needsdiagnosis+label:engine-gecko+repo:webcompat/web-bugs This will return a paginated list of issues but also total_count. Today 95 issues with engine-gecko. Maybe this is the simplest. Timestamp in the DB with the count.