DRAFT: Refactoring RPCWithMemoTest so to make easier to re-implement in Go

[edmondop]

This PR comes as a consequence of trying to implement indeedeng/iwf-golang-sdk#64. After completing the model, I had to migrate the tests and it was very hard to make sense of the existing one.

Overall, I propose the following improvements:

• Break down each step that performs multiple assertions into different tests, so to have independent and isolated failures
• Avoid Thread.sleep in tests and delegate eventuay consistency matching logic to a library, in this case awaitility
• Use Hamcrest matchers and their combinatorial logic to create other matchers
• Make better usage of JUnit behavior, in particular that each tests will be run on a separate instance of the test class, to do more initialization in the @BeforeEach rather than at the beginning of each test
• Use generics to avoid repetition for the tests that set and get the attributes and keyword properties
• Added thread-safety in gray areas for singleton

TODO:

• Port the readonly test
• Fix the two tests that fail io.iworkflow.integ.RpcWithMemoTest#testRpcFunc0OutputIsCorrect, io.iworkflow.integ.RpcWithMemoTest#testRpcFunc1WorkflowResultIsCorrect both failing for this workflow should have one or zero state output for using this API, found 2, after filtered NULL: 0

Further improvement down the line:

• Use dynamic generation of tests using the @TestFactory JUnit annotation so to automatically generate the testcases for Func1, Func0 and Proc1/Proc0
• Move integration tests from a package to a separate source folder, as per Java conventions

[longquanzheng]

looks like the tests are failing.

[longquanzheng]

Thanks for fixing that

[edmondop]

looks like the tests are failing.

Yes, that's why is in Draft. I was looking for help to understand why the two tests are failing :)

[longquanzheng]

I believe this rpc doesn't mutate the TEST_SEARCH_ATTRIBUTE_INT and the int doens't work for string