Write some tests

[NicolasMICAUX]

Is your feature request related to a problem? Please describe.
Test are crucial to ensure the performance is good.

Describe the solution you think of
I think we must implement 2 types of tests:

• Classical tests to ensure there isn't bugs in the code.
• Performance tests, where we check if/and how fast the algo finds the query in an object.

Roadmap

• Make some "basic" tests
• Create performance tests (time before match, number of "bad" results returned vs the results expected)
• Create some adversarial examples, where we try to make the lib fail.

Importance & Difficulty
Importance: 4/5, tests are crucial.
Difficulty: 2/5, tests may be very basic, or not :)

Any contributor willing to do this is welcome :)

[NicolasMICAUX]

Here are some examples of tests that we could write:

Handling weird attributes/methods

• Should we look into obj.__doc__, probably yes ; but into obj.__new__? Not really. There are a lot of similar methods out there. I someone has the time to understand which are relevant for searching in.

Catching errors & exceptions

• Some weird objects may raise errors/exceptions. We need to catch them to avoid crashing the searching process. Already done a few try catch for OSError etc.

Build a "real-life" perf benchmarking dataset

• Find real object, methods, variables of frequently used libraries (tqdm, pandas, rsa, numpy, pytorch, ...) https://hugovk.github.io/top-pypi-packages, figure out real life scenarios of people trying to find a parameter/property/method.
• Reproduce the structure of the object/method/variable (by copy-pasting the code), but remove the imports etc, to avoid having to import libs when testing)
• See how much time (in avg), and how much of the real-life scenarios are successfully completed by searchin.