#768: Update environment to python 3.12

[Ed (mo-gill)]

Closes #768

[Ed (mo-gill)]

In commit f21b5c4 I changed array to asarray as per the suggestion from this output generated from running the tests. Doing this fixed ~30 failing tests.

    def make_masked(data, shape, missing_value, dtype):
        logger = logging.getLogger(__name__)
        if dtype in [numpy.dtype('int32'), numpy.dtype('int64')]:
            logger.debug('Received data type "{}", converting to float32 to avoid issues later')
            dtype = numpy.dtype('float32')
    
>       data_array = numpy.array(data, dtype=dtype, copy=False)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
E       ValueError: Unable to avoid copy while creating an array as requested.
E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.

mip_convert/mip_convert/variable.py:98: ValueError

[Ed (mo-gill)]

When fixing unit tests, this doctest was failing:

___________________ [doctest] mip_convert.model_date.CdDate ____________________
019 
020     Dates are generated based optionally on the year, month, day, hour,
021     minute and second.
022 
023     >>> time1 = CdDate(1970, 1, 1, 0, 0)
024     >>> time2 = CdDate(1970, 1, 1, 0, 0, 0)
025 
026     They can be compared for identity:
027 
028     >>> time1 == time2
Expected:
    True
Got:
    np.True_

In commit bde4ce9 i added some casting from numpy bools to python bools to fix this.

Explanation: In numpy 2.0.0 bools went from being represented as True to np.True_, hence i added the casting to match the docstring.

More info on the numpy boolean change here numpy/numpy#26774

Edit:

I had a similar issue here:

_________________________ [doctest] mip_convert.plugins.base.data.processors.eos_insitu _________________________
942     rho: :class:`numpy.ndarray`
943         In-situ density (kg/m3)
944 
945     Examples
946     --------
947     Rounding is necessary to reproduce the following worked examples:
948 
949     * From the NEMO `eos_insitu` docstring
950 
951         >>> round(eos_insitu(40., 40., 10000.), 5)
Expected:
    1060.93299
Got:
    np.float64(1060.93299)

In this case i tried casting the result of eos_insitu to a float to match the expected type but this broke other tests (test_calc_rho_mean and test_zostoga_zero). It seems they were expecting numpy data types rather than floats (and whilst the original implementation was returning a numpy data type, the string representation of those types in numpy <2.0.0 was not explicitly declaring that (instead representing it as a float, which is why the doctest passed previous to numpy being updated to >2.0)

Hence i've just adjusted the docstring to expect a np.float64 instead in commit 91cd7ed
More info: https://numpy.org/neps/nep-0051-scalar-representation.html

[Matthew Mizielinski (matthew-mizielinski)]

Following tests failed for me;

FAILED mip_convert/mip_convert/tests/test_functional/test_functional_cmip7/test_cmip7_fx_rootd_ti_u_hxy_lnd.py::TestCmip7_fx_rootd_ti_u_hxy_lnd::test_cmip7_fx_rootd
FAILED mip_convert/mip_convert/tests/test_functional/test_functional_cmip7/test_cmip7_fx_slthick_ti_sl_hxy_lnd.py::TestCmip7_fx_slthick_ti_sl_hxy_lnd::test_cmip7_efx_slthick

both due to presence of branded_variable attribute -- I think they just need updated reference data.

[Ed (mo-gill)]

Following tests failed for me;

FAILED mip_convert/mip_convert/tests/test_functional/test_functional_cmip7/test_cmip7_fx_rootd_ti_u_hxy_lnd.py::TestCmip7_fx_rootd_ti_u_hxy_lnd::test_cmip7_fx_rootd
FAILED mip_convert/mip_convert/tests/test_functional/test_functional_cmip7/test_cmip7_fx_slthick_ti_sl_hxy_lnd.py::TestCmip7_fx_slthick_ti_sl_hxy_lnd::test_cmip7_efx_slthick

both due to presence of branded_variable attribute -- I think they just need updated reference data.

Have now fixed these and added the reference data so they're passing in 54a80de

[Matthew Mizielinski (matthew-mizielinski)]

LGTM and all tests pass