Specify dependency package versions in environment.yml about pysteps HOT 5 CLOSED

We need to revise the environment.yml before the next release.

For example, now jsmin and attrdict are listed, but these two packages are not available in the anaconda distribution and nees to be installed with pip. Hence, the current version of the environment.yml cannot be used with conda env create -f environment.yml.

from pysteps.

I was testing pysteps in a minimal environment, with out the optional dependencies listed below:

• opencv-python_ (for the Lucas-Kanade optical flow)
• pillow_ (for importing gif data)
• pyproj_ (for cartographic transformations and needed to import data)

The package functionality is quite minimal without these dependencies and the examples scripts provided can not run with out them.

Since many of them are already in the environment.yml, wouldn't be better to consider them as dependencies instead as optional packages? By doing so, the full functionality is available after the installation.

This is not the case with **dask, toolz and pyfftw **, they are indeed optional, since they help to improve the computational efficiency.

from pysteps.

OK, good idea, I've now listed opencv, pillow and pyproj as main dependencies (commit f157da9).

from pysteps.

Going back to the @pkars suggestion about the version numbers. The environment.yml file is used to create a the pysteps environment. When doing so, the anaconda installer will resolve the dependencies of the different packages listed in the environment.yml file and it will it install the latest versions available in anaconda channels that are configured.

from pysteps.

@aperezhortal 's comment seems to close this issue.

from pysteps.