MAINT: Pytest refactoring

[MBounouar]

Refactored the asset classes to take advantage of pytest fixtures

re-enabled a test example that was originally skipped (it works on linux)

[stefan-jansen]

There are varying (small) numbers of tests failing with pandas 1.1, such as some of the example strategies.

[MBounouar]

There are varying (small) numbers of tests failing with pandas 1.1, such as some of the example strategies.

I can confirm on my machine (Debian 10, python3.8.5) that indeed pandas-1.1.5 is failing the test_examples.py , it wasn't such a bad idea to re-enable the tests_examples after all... (pandas-1.2.4 passes the test)
I haven't found the issue, the error message is not helpful, but I strongly suspect at the moment that pandas-1.1.5 can't properly load the data.
The tests are passed sucessfuly for pandas>=1.2.3

The issue is not clear to me yet but the class dataframe_cache(MutableMapping) can't load the data

[stefan-jansen]

I think the example tests fail because there need to be data for each pandas version (see (example_data)[https://github.com/stefan-jansen/zipline-reloaded/tree/main/tests/resources] files - data and rebuild_. This should probably only be required per minor version.

[MBounouar]

Indeed I completely overlooked that (facepalm) .... pandas could be indeed considered mature enough for considering only minor numbering changes. BTW rebuild_example_data hasn't been deleted from the repo when you added the missing extension.
and the rebuild script won't work because the zipline_git_root changed

[stefan-jansen]

Both Windows and macOS fail because of Py3.9 dependencies (Windows - HDF5, macOS due to c-blosc). We can skip Windows for 3.9; there's a homebrew formula for c-blosc, perhaps that does the trick, will check tomorrow. I'll merge anyway as this is unrelated.