I am the owner of a medium-sized nodejs application that can be found on GitHub under the link here. This application heavily relies on various 3rd party APIs to function.
Essentially, the purpose of this app is to make calls to different Salesforce APIs, both REST and SOAP, depending on input parameters and responses from previous API calls. This dynamic nature means that each request may follow a unique execution path and involve a varying number of API calls. It's possible for the app to make over 300 API calls in a single request, thanks to asynchronous job processing.
I'm currently facing challenges in creating effective tests (using mocha, jest, etc) for this application due to the complexity of the API calls and their conditional logic.
Most test examples available online focus on simple endpoints with predictable responses that can be easily tested using assertions like assert(response)ToBe(200). However, my application deals with a multitude of diverse APIs that yield a wide range of responses.
One approach could be to individually mock all these endpoints through unit tests, but this seems labor-intensive and wouldn't provide comprehensive end-to-end integration testing.
Alternatively, I could simulate the entire response of a complete job and verify that the code processing it functions correctly. Yet, my primary interest lies in testing the data transformation code responsible for consolidating results from numerous API calls.
Complicating matters further, much of the API-handling code is encapsulated in closures, making isolated testing challenging. Testing only the outer function without mocking inner function responses would be unfeasible.
Has anyone successfully implemented automated integration tests for applications heavily reliant on transforming 3rd party data? If so, could you share any insights, best practices, or lessons learned?