Feature Request: Fake Twilio Client

[qdii]

Issue Summary

Properly testing client code against Twilio python library isn't easy. None of the solutions I have found so far allow to run tests with good coverage that are reproducible and hermetic:

• This blog post suggests mocking the entire client, but this methodology fails to catch problems such as calling client.messages.send() with invalid parameters or function names.
• This feature request suggests to use localhost instead of api.twilio.com, but this means reverse-engineering the protocol used to talk to api.twilio.com and maintaining a server running just for tests.
• The option to use responses python package to mock POST and GET requests to api.twilio.com is also very brittle, as the tests would break every time the client is changed. It also takes reverse engineering to know what the client expects.
• Using test credentials is a bad idea: first because this means the tests are not hermetic and reproducible, then because it hammers Twilio API endpoint every time a test is run, and last because it makes the test flaky by adding a dependency on internet connectivity and the endpoint being online.

To solve that, could you please provide a library that allows users to test the integration with the Client instance?

[tiwarishubham635]

hi! have you tried using this example?

[qdii]

@tiwarishubham635 Yeah, I think this could be used to create such a fake client.

However to make it a viable solution, we'd need to know the details of the API calls made by the real client and api.twilio.com.

Also, I'm not sure how often this API changes, or if such changes are going to be backward-compatible.