[PromQL] Percentile parsing

[aurrelhebert]

Proposal

Use case. Why is this important?
When add an incorrect parameter for quantile function, a result is loaded

Nice to have' is not a good use case :)
You can test it with: "percentile(2.0, metric...)

Bug Report

What did you do?
Just execute command below on Prometheus

What did you expect to see?
An error message

What did you see instead? Under which circumstances?
A result (+inf)

Environment

• Prometheus version: 2.3.1

[brian-brazil]

May I ask why this makes a difference to you? Changing this would be a breaking change, as it is the documented behaviour of our quantile functions.

[aurrelhebert]

We are working on PromQL proxies and validate the query with the parser first. We didn't expect this to be a valid request.
That's why we added some checks before on our side.

[brian-brazil]

This is valid, though not particularly sane, PromQL.