I’m writing tests that check that an external service is providing inventory data (on their test server) that I can check out with, and also cancel. This testing is in the travel/hotel world, and I need to place test reservations, then cancel them.
I take the following steps :
1) search for inventory(a hotel room) 90 days in the future
2) get the first result, and do a test checkout with it
3) cancel the order and confirm I get a cancellation number
4) confirm the appropriate database entries are performed
This testing touches wide ranging parts of the system, and I don’t (currently) have mechanisms to isolate each of them. I therefore don’t consider this unit testing. However, would this be called functional testing, or something else?
As a follow up, it will probably help to isolate and write tests for each of the subsystems I’m interacting with. Where would you guys start in the isolation process?
Each of those steps you listed is a good candidate for an isolation:
Stringing them together as you do now is only a function of each of them working individually anyhow.
You’re currently testing for functionality (good!) but you didn’t mention robustness.
I would make a suite of data for each function (as I have listed) that would try to break it and probe corner cases. Try to book in the past, try to over-write POST data, try to over book a room, try to simultaneously book the same room twice, etc. This is all stored as input parameters in a test you have written just for one isolate of your application.
The different isolates will have different data that is useful/meaningful to test with, but with each change/commit/build, you should be able to run the test data against each function and have the results be valid (IE you return a query, check out a room, cancel the reservation, or get the cancellation #)