Add a mock clock

[schomatis]

Context: libp2p/go-libp2p-core#267

It seems all Meters use a single globalSweeper running on the Go clock. I'm not sure how to go about this. For example:

• Should each Meter have its own sweeper now (with its own mock/Go clock)?
• Should we still rely on the globalSweeper but somehow have it track individual clocks (either in another structure like meters []*Meter or inside the Meters themselves)?

@Stebalien help (cc @marten-seemann)

[Stebalien]

So, I created this library because metrics kept showing up in performance profiles.

• Each meter should not have a sweeper, that would kill performance.
• We can't have individual clocks (performance, complexity, size).

The best we could do would be to have a global clock and mock it via an environment variable (on init).

[marten-seemann]

The best we could do would be to have a global clock and mock it via an environment variable (on init).

Not sure how you'd inject a clock via an environment variable, the caller needs to be able to control the (mock) clock.

We could introduce a SetClock(clock) function.

The interesting thing here is that once a sweeper is started, it can't be stopped. That means we can only single mock clock for all test cases. I don't expect this to be a problem though, after all, the tests work right now using a single clock.

[Stebalien]

Not sure how you'd inject a clock via an environment variable, the caller needs to be able to control the (mock) clock.

Uh, sorry, not environment variable. Global variable.

We could introduce a SetClock(clock) function.

That also works, but it would likely still be a global.

[schomatis]

Thanks for the input. I'm going then (unless someone says otherwise) with the global SetClock API meant to be called only once and before registering (using) any meter (leveraging the sweepOnce mechanism). It's on the test to inject the mock clock before using meters. (I'm assuming that separate tests have separate global/static contexts and hence separate sweepers.)

[schomatis]

#18

[schomatis]

(I'm assuming that separate tests have separate global/static contexts and hence separate sweepers.)

@marten-seemann @Stebalien We're sharing the global sweeper across tests so the above isn't correct and I'll need some advice on how to proceed.