Refactor tests that might be unnecessary

[surister]

in test_http.py, for example tests that use:

crate-python/tests/client/test_http.py

Line 563 in 371e125

class SharedStateRequestHandler(BaseHTTPRequestHandler):

Do not seem to be very useful, we setup an HTTP server that parses things from the headers and puts the parsed stuff to a dictionary, then we check that that dictionary has what we expected.

Why are we testing this on a threaded HTTP server, we could literally just get the headers from a request and check that it's what we'd expect, without any overhead of setting up a thread and an http server. Other tests that also start an http server have the same issue, we could test the exact same thing without all of this machinery.

[amotl]

Hi. I've stumbled upon this area/topics of the test suite in the past as well. Maybe colleagues who have been involved earlier can remember and report about some of the rationale here? Maybe that was meant to validate the HTTP transport a bit more in detail, instead of becoming "full" integration tests against CrateDB?

without any overhead of setting up a thread and an http server

Maybe particular details about handling multiple requests in parallel properly have been a concern of those tests, which wouldn't be possible when using mock objects only?

/cc @matriv, @seut, @mfussenegger

[mfussenegger]

Why are we testing this on a threaded HTTP server, we could literally just get the headers from a request and check that it's what we'd expect, without any overhead of setting up a thread and an http server.

That would also imply testing a lower level method so you'd actually get the request?
The way it is covers way more logic.