feat: mock transport for tests

[shramee]

Testing functionality based on RPC response

Tests require setting up an RPC server with appropriate state to test different cases. This makes tests slow and setup tedious.

This PR adds a mock transport mode to mock certain requests.

Features

Flow

JsonRpcTransport::send_request is implemented to,

1. Tries to find mocked response for request_json in self.mocked_requests.
2. Tries to find mocked response for request method mock in self.mocked_methods.
3. Otherwise, Falls back to http_transport if available.
4. Else throws an error (all requests must be resolved).

Optional HttpTransport fallback

Adding an http_transport allows falling back to HttpTransport for requests not mocked.

Mocking

Responses can be mocked for

1. specific request json string (high priority) or
2. all requests if w specific JsonRpcMethod.

Dev Experience

Helpful prints are added to stdout to help build mocks for the requests received and responses from fallback HTTP.

Use this code to mock this request

mock_transport.mock_request(
    "{\"id\":1,\"jsonrpc\":\"2.0\",\"method\":\"starknet_getBlo...}".into(),
    "{\"jsonrpc\":\"2.0\",\"result\":{\"block_hash\":\"0x5cc6b9...}".into()
);

[shramee]

Raw strings for cleaner JSON...

Use this code to mock this request

mock_transport.mock_request(
    r#"{"id":1,"jsonrpc":"2.0","method":"starknet_getBlockW...}"#.into(),
    r#"{"jsonrpc":"2.0","result":{"block_hash":"0x5b586405c...}"#.into()
);

[xJonathanLEI]

It's been... a while. While I agree that a mock transport would be helpful for testing stuff in downstream applications, I don't quite agree with the approach taken here.

In the context of starknet-rs, a "transport" is strictly a way of passing requests and responses back and forth. For example, an HTTP transport uses HTTP connections. I would imagine a mock transport to always use mock data for responses, instead of wrapping an HTTP transport as fallback in this PR.

I can see how being able to fallback to another transport is useful - sometimes you might want to just mock a few requests but use real data for the rest. However, I think the best way to implement this would be something like a DualProvider, where you can configure routing rules for requests.

On top of that, the PR seems to be incomplete. There are spots that have empty docs sections; there are println statements in non-test code; there are commented out lines. I don't think it's ready for review yet.

So I'm closing this one here. I do believe there's a place for something like this (probably not string-based though), but this one submitted likely isn't the one.

[shramee]

Why strings

It's easier to have it string based coz the response usually is string. And exactly the same serde steps can be taken to parse the mocked string responses.

Ease of strings

The requests can be trivially mocked by just copying over the expected response from, say, a 'prepared' devnet. Making it real easy to write tests that run in a fraction of what setting up a devnet/sequencer in CI would entail.

Using typed responses

We could indeed pass the request we want to mock through serde first and then export the rust variables. Or even write response type instantiations from scratch for tests. This would be more type safe but would force the peeps to to learn about the types used in the the library for every request.

Next?

Yeah, this is quite out of date, was a proof of concept to begin with. DualProvider sounds interesting, would be nice if the learning overhead for writing tests is as low as possible. Looking forward to see what we want to do about this. We might even abandon this issue if not many people need fast tests with starknet-rs.