Expose invocation metering in the SDK.

[dmkozh]

What

Expose invocation metering in the SDK.

Why

Improving testing UX.

#1319

Known limitations

N/A

[leighmcculloch]

💯 This is so needed and awesome to see getting added to the SDK.

I have some thoughts just about the API and how we integrate it. Lmk what you think, and we can DM / chat too to do a little more realtime.

[leighmcculloch]

👏🏻 Love the level of docs we're adding around these functions.

[leighmcculloch]

See the suggestion inline about how to deal with set_config.

[dmkozh]

If set_config is called with turning the value off, when the value is already turned on (should be an easy comparison), panic with an informative message saying that once it is enabled it cannot be disabled.

That makes sense and I've tried to sketch this. But this doesn't pass the semver check, as we're adding a new pub field to a struct, so initialization will be broken. So I'm even more in favor of enabling this by default.

[leighmcculloch]

this doesn't pass the semver check, as we're adding a new pub field to a struct

Darn, I forgot about this when I created this struct. In v23 we should change the struct into private fields with functions to modify the state. I created an issue to do that here:

• Make EnvTestConfig internals private and add functions #1419

So I'm even more in favor of enabling this by default.

👍🏻. We can plan to retroactively introduce a EnvTestConfig flag in v23 too if needed. If folks need to turn it off before then, we can introduce an env var too. I had considered adding an env var for the other env test config item too.

[dmkozh]

Based on the previous discussion I've enabled this by default, PTAL.

One caveat that I forgot about is that besides budget we also reset all the events between the calls. This makes sense in general and may actually highlight issues people might have had (if an event was expected to happen after a wrong contract call).

[leighmcculloch]

I think it's reasonable to treat that like a bug fix. The way it works today is too unspecific as to when events occur.