Conda Packaging

[GiulioRossetti]

To prepare a conda package for cdlib a few dependencies without a dedicated conda distribution need to be packaged as well.

Here's the list of satisfied/not satisfied dependencies so far (channels: conda-forge, giuliorossetti):

• [x] numpy
• [x] future
• [x] matplotlib
• [x] scikit-learn
• [x] tqdm
• [x] networkx
• [x] demon
• [x] python-louvain
• [x] nf1
• [x] scipy
• [x] pulp
• [x] seaborn
• [x] pandas
• [x] eva_lcd
• [x] karateclub
• [x] bimlpa
• [x] markov_clustering
• [x] python-igraph
• [x] leidenalg
• [x] angel_cd
• [x] pooch
• [x] dynetx
• [x] thresholdclustering
• [x] pyclustering
• [x] cython
• [x] python-Levenshtein
• [x] graph-tool
• [x] gmpy2
• [x] chinese_whispers
• [x] networkit
• [x] infomap
• [x] wurlitzer

Optional dependencies:

• [ ] GraphRicciCurvature
• [ ] shuffle_graph
• [ ] similarity-index-of-label-graph
• [ ] ASLPAw
• [ ] pycombo

[GiulioRossetti]

Starting from v0.2.3 CDlib the official conda package is available and installable with:

conda config --add channels giuliorossetti
conda config --add channels conda-forge
conda install cdlib

Optional dependencies still need to be installed separately using pip.

[bramson]

Considering your motivation for having these CD algorithms available to the widest possible research audience, having a working conda install is a great idea.
However, I tried this, and it failed with the following (uninformative to me) messages:

UnsatisfiableError: The following specifications were found to be incompatible with each other:
Output in format: Requested package -> Available versions

Any idea why this failed and how to correct it?

At the moment, I'm only planning on running the fuzzy communities algorithm, so I haven't installed any of the optional dependencies. Do I need to?

[bramson]

Ah, well, I just read in the docs, that your implementation of both fuzzy CD algorithms don't support weighted networks (although the Fuzzy-Rough Community Detection method does), so I can't use your package for my current needs anyway. But I still want to install it for future use, and perhaps I'll use a non-fuzzy method that does support weighted networks...of which I could find none in your documentation (even in cases where the algorithm does support edge weights). Do none of your implementations support weighted networks, or is this a problem with the docs?