proposal: Allow providing default values for missing series in binary operations

[juliusv]

See prometheus/prometheus#13625

[MichaHoffmann]

LGTM!

[juliusv]

FYI, in the implementation and practical experimentation of this, I noticed that it doesn't make sense to disallow filling in the "many" side of a many-to-one / one-to-many match, even though we cannot make up (fill in) multiple missing series due to the unknown differentiating labels. We can still always fill in a single missing series that just has the match group's labels though. This can still be useful for people who want to use group_x(<labels>) in a 1:1 situation, but where they just want to include (<labels>) from one of the sides (usually, an info metric on the right side).

I've updated the proposal accordingly.

[bwplotka]

I like this. From UX this is something I have been missing often and fill looks like a simple enough solution 👍🏽

cc @realschwa

[juliusv]

Going to merge this for now so I can properly link to it. That doesn't mean it can't change of course.