Add re-initialization logic for event filters

[jaxernst]

Here's my proposal to address #836. Filters can timeout or get into a bad state where reinitialization is necessary to continue getting filter changes on the polling interval. This request adds fallback reinitialization, so if 'getFilterChanges' fails repeatedly, it will attempt to recreate the filter.

---

PR-Codex overview

Focus of the PR:

This PR focuses on adding filter reinitialization logic for watchContractEvent and watchEvent when a filter has been uninstalled.

Detailed summary:

• Added filter reinitialization logic for watchContractEvent and watchEvent when a filter has been uninstalled.
• Imported InvalidInputRpcError in watchEvent.ts and watchContractEvent.ts.
• Added error handling and filter reinitialization logic in watchEvent.ts and watchContractEvent.ts if a filter has been set and gets uninstalled.
• Added a test case to verify the filter reinitialization logic in watchEvent.test.ts and watchContractEvent.test.ts.

✨ Ask PR-Codex anything about this PR by commenting with /codex {your question}

[vercel]

@jaxernst is attempting to deploy a commit to the wagmi Team on Vercel.

A member of the Team first needs to authorize it.

[changeset-bot]

🦋 Changeset detected

Latest commit: 732a56d

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[jxom]

Instead of checking if the filter has errored n times, maybe we can check the error for “filter not found”.

Most nodes support this, examples:
https://github.com/ethereum/go-ethereum/blob/master/eth/filters/api.go#L38
https://github.com/paradigmxyz/reth/blob/f41386d28e89dd436feea872178452e5302314a5/crates/rpc/rpc/src/eth/filter.rs#L447

[jaxernst]

Instead of checking if the filter has errored n times, maybe we can check the error for “filter not found”.

Most nodes support this, examples: https://github.com/ethereum/go-ethereum/blob/master/eth/filters/api.go#L38 https://github.com/paradigmxyz/reth/blob/f41386d28e89dd436feea872178452e5302314a5/crates/rpc/rpc/src/eth/filter.rs#L447

This does make more sense. I was hesitant to do this because I wasn't sure if "filter not found" was standardized, so I looked a bit further. Every client I've checked returns this error except for Erigon:

Besu: https://github.com/hyperledger/besu/blob/main/ethereum/api/src/main/java/org/hyperledger/besu/ethereum/api/jsonrpc/internal/response/RpcErrorType.java#L45

Nethermind:
https://github.com/NethermindEth/nethermind/blob/a099f0fe2158b969ca6b81861c5097ecc5f6c0b1/src/Nethermind/Nethermind.JsonRpc/Modules/Eth/EthRpcModule.cs#L553

Erigon:
https://github.com/ledgerwatch/erigon/blob/56934395210c48585ae8d7f77404d7c9b9ba3a30/turbo/jsonrpc/eth_filters.go#L78

Should we still check for this even if not all clients return it?

[jxom]

Thought so. Maybe we can keep the current implementation as a fallback, and check the error first?

[jaxernst]

Although Erigon doesn't throw a 'filter not found', they do all appear to return a -32000 (Invalid input) error code when the filter isn't found. It looks like Viem already catches these as "InvalidInputRpcError", so might we just reinitialize the filter if we catch an InvalidInputRpcError?

[jaxernst]

Implemented this with tests for watchContractEvent and watchEvent

[jxom]

Looks good! Will merge it in soon.