Do not export unnecessary conda env vars, use conda run

[rly]

Fix #230

[rly]

Tests pass when run locally except for the error TypeError: float() argument must be a string or a real number, not 'coo_matrix' reported in #231.

[kushalkolar]

I wonder if this coo_matrix is due to some recent change with caiman.

[kushalkolar]

Because we didn't have it before.

[kushalkolar]

This would assume that conda is findable in the shell env that that is launched by the subprocess and I don't think that's going to be the case if the conda binary isn't in the PATH env variable.

[kushalkolar]

Is "conda run" the best way to do this? If so then the fully path to the conda executable should be used.

[rly]

An alternative is to activate the conda environment and then run python, but conda activate within a shell script can be tricky.

[rly]

I would go with conda run

[kushalkolar]

Are we sure this is no longer necessary?

[kushalkolar]

I mean the CONDA_PYTHON_EXE

[rly]

I believe these env variables will not do anything if the conda environment is not activated and you call the default python below.

[rly]

This works locally on my Mac M1 system.
Steps:

1. create conda environment
2. run through one of the notebooks like mcorr_cnmf.ipynb up to the df.iloc[0].caiman.run() step
3. the cell runs successfully
4. deactivate conda environment
5. call the runfile bash script without any conda environment activated (not even base). the shell script runs successfully!

relevant env vars without any conda env activated are:

CONDA_CHANGEPS1=no
CONDA_EXE=/Users/rly/mambaforge/bin/conda
_CE_M=
_CE_CONDA=
CONDA_PYTHON_EXE=/Users/rly/mambaforge/bin/python
CONDA_SHLVL=0

[ethanbb]

This is working for me (on Linux).