Provide DeerLab through Anaconda Repo

[laenan8466]

Hey,
first congrats moving to python. 😍

From my experience, many researchers use Anaconda (https://www.anaconda.com/) to organize their python packages. So it could be helpful to provide DeerLab via Anaconda as well.
https://docs.anaconda.com/anaconda-cloud/user-guide/getting-started/

If not an option for you feel free to close this issue, I would appreciate consideration, as it would simplify the workflow, especially for python-n00bs. 😉

Cheers,
Dennis

[stestoll]

Makes sense to me - unless it's too much additional trouble.

[luisfabib]

Agreed, this would make sense.

I am not too familiar with Anaconda and the conda manager, so I will need to look at how conda manages the BLAS/LAPACK binaries of numpy/scipy on each OS.

[mtessmer]

Last I checked conda defaults to mkl numpy/scipy on all operating systems. Not sure about cvxopt but I know cvxopt is available.

[stestoll]

I think pip can be used from inside conda, so it should not be necessary to provide DeerLab via Anaconda.

[laenan8466]

I would agree (in principle), but it is not recommended to use pip in a conda environment. But yes, it is possible to use pip with Anaconda.
https://www.anaconda.com/blog/using-pip-in-a-conda-environment

[luisfabib]

For now, DeerLab will only be provided through pip.

Maybe in a future release, this will change, but for now, this would introduce too many maintenance issues in the short-term future which should be focused on more critical development of the software.

[luisfabib]

After some preliminary trials to distribute DeerLab through conda and the Anaconda repo, it seems that the current pip setup is incompatible with the conda system.

Making this compatible would require either a major hack to make it conda-compatible, or remake the whole pip installation. The issue is re-opened as we still consider this, but cannot implement it at the current time.

[laenan8466]

For reference, how to use DeerLab with Anaconda (for everyone else who stumbles over this problem). This is from my memory, so @luisfabib can comment if this way causes problems as well.

Installation

Open your git console.
Choose a local directory, where the files should be stored.

cd /local/directory/
git clone https://github.com/JeschkeLab/DeerLab.git
cd DeerLab
git checkout v0.12.2

Replace the v0.12.2 with the current latest version found on the releases page.

Open the Anaconda Prompt as admin and navigate to the directory, then activate your environment (optional)
Installation will be made with symlink, so updates in the folder will be availible after a kernel restart automatically

cd /local/directory/DeerLab
# activate installation-env   # change to your environement, optional
pip install -e .            # -e statement will establish a symlink

Update

Open a git console and go to the local directory, git pull and checkout new version.

cd /local/directory/DeerLab
git pull
git checkout $NEW_VERSION_TAG

Replace $NEW_VERSION_TAG with the newest version from releases page.

edit: replace git clone command by https statement

[luisfabib]

Thanks for the instructions! I have updated the documentation installation instructions to include this workaround for Anaconda users (see #155).

[laenan8466]

Just realizing I made a minor error:
The git clone call should use the https link, as most users wouldn‘t have a git account with stored ssh-key.
git clone https://github.com/JeschkeLab/DeerLab.git
should do the magic.

[stestoll]

Closed by #157