fix: fixed tests and logic to find the fastest endpoint

[gentlementlegen]

Resolves #32

[github-actions]

Unlisted binaries (1)

Filename	binaries
.github/workflows/publish-package.yml	publish

[gentlementlegen]

Seems the latest run that didn't have the issue was https://github.com/ubiquity/rpc-handler/actions/runs/10122809912
Will have a look.

[Keyrxng]

good luck, it's got hands

[gentlementlegen]

@Keyrxng I solved the testing issue by mocking the post from axios because we do not need to hit the real endpoints anyway. Also I removed the tests where you were testing JsonRpcProviders because you are testing their API not ours, so I don't think that was really relevant. Everything is green now.

[Keyrxng]

Also I removed the tests where you were testing JsonRpcProviders because you are testing their API not ours

I agree the first/top two tests were testing the provider API endpoints but only with read ops but the following write ops were all targeting the local anvil instance.

  const mods: HandlerConstructorConfig = {
      runtimeRpcs: rpcList.map((rpc) => rpc.url),
      networkRpcs: rpcList,
      autoStorage: false,
      cacheRefreshCycles: 1,
      networkName: "anvil",
      networkId: "31337",
      rpcTimeout: 700,
      proxySettings: {
        retryCount: 1,
        retryDelay: 10,
        logTier: "verbose",
        logger: new PrettyLogs(),
        strictLogs: false,
      },
    };

---

The provider racing was the original functionality of this package but it's now in control of all read and write operations so I think we should have tests for this

[gentlementlegen]

@Keyrxng I did my best to add back the tests with mods and naming them coherently, let me know what you think.

[Keyrxng]

rpc-handler.test.ts contains tests related to the original functionality of the package, so things like proper init values and it returning the fastest provider.

call-handler.test.ts should contain tests that are testing the proxy and therefor all of the retry and reselection logic

As far as I can see all of the call-handler.test tests are now testing with expect(provider).toBeTruthy(); which is only testing for the successful returning of a provider on init

Checkout this jest run as I merged proxify. No leaking worker and lots of logs showing that the retry logic was being hit and was working. This latest run only shows provider initialized logs and doesn't appear to be hitting the retry logic

[gentlementlegen]

@Keyrxng Right ok I get it. But then I should try to entirely mock the RPC provider logic because we do not want real endpoints to be hit.

[Keyrxng]

@Keyrxng Right ok I get it. But then I should try to entirely mock the RPC provider logic because we do not want real endpoints to be hit.

Well no actually because anvil is dumb and will respond successfully more or less no matter what you send to it, so in the specific case that our nonceBitMap read op was failing and it was only for certain providers anvil will always give you a great light under that scenario

Maybe I am mistaken @gentlementlegen and you can do that it may be possible so let me correct above and just say that the above is my opinion and that's why I took the approach that I did but you can mock everything if we are still able to effectively test the call-handler aspect I guess

[gentlementlegen]

@Keyrxng Took another different approach: I removed the mocks and always use Anvil instead, tell me what you think.

[Keyrxng]

Everything looks good as far as I can tell. Test logs showing retry logic and no leaky worker error