Recommend conda instead of pip
Recommend conda instead of pip
Proposal
from the review in https://github.com/openjournals/joss-reviews/issues/2316
Installation: Does installation proceed as outlined in the documentation?
You list pip
as the preferred installation method, although dantro
depends on packages like numpy, matplotlib, pandas and many more, which are known to be critical when installed via pip
and often cause problems (particularly on Windows).
Ideas for an implementation
I strongly recommend to avoid pip
and rather provide a conda package or an environment.yml
file to install the dependencies, e.g. add the following contents in the environments.yml
file in the root of the repository:
name: dantro
dependencies:
- numpy >=1.17.4
- xarray >=0.15.1
- dask >=2.10.1
- toolz >=0.10.0 # Needed for dask.delayed
- distributed >=2.10.0 # Needed for dask's distributed scheduler
- scipy >=1.4.1 # Used as a netcdf4 storage engine for xarray
- sympy >=1.5.1
- h5py >=2.10.0
- networkx >=2.2
- ruamel.yaml >=0.16.10
- matplotlib >=3.1.3
- pip
- pip:
- paramspace>=2.5.0
- .