decklist_analyser

decklist_analyser takes Magic: the Gathering decklists in CSV form from a specific source and uses them to either check for overlapping cards between lists or group decks based on the cards they contain into distinct clusters detected algorithmically.

Installation

Installing decklist_analyser can be tricky thanks to some of the packages used.

Simple method

Run pip install -r requirements.txt and hope it works

Involved method (Windows)

Start by installing numpy, pandas and regex with whatever package manager you prefer. Next, attempt to install pycairo. If this fails, try the wheels located at https://www.lfd.uci.edu/~gohlke/pythonlibs/. Then, install python-igraph. It's important to do them this way around - python-igraph depends on pycairo and if you are manually downloading the wheels it won't use those, but rather download its own and fail. The above URL has wheels for python-igraph. Next, install louvain. Again, the above URL has the wheels. matplotlib can also be found above.

Involved method (Unix)

Start by installing numpy, pandas and regex with whatever package manager you prefer. Since Unix systems vary wildly, I only detail the broad strokes below. Next, attempt to install pycairo. If this fails, you're probably missing dependencies required. See https://pycairo.readthedocs.io/en/latest/getting_started.html for details. Then, install python-igraph. Again, missing dependencies are likely - check https://igraph.org/python/. Similarly, for louvain, detailed installation instructions can be found from https://louvain-igraph.readthedocs.io/en/latest/install.html. matplotlib installation instructions are at https://matplotlib.org/3.1.1/users/installing.html.

Configuration

In the directory, create the folders logs/ and output/. This prevents the program throwing errors because it doesn't currently make folders for these. In config/, copy CONFIG.example.py to CONFIG.py and configure this as you wish.

Quick rundown of CONFIG

• graphml_location: where output graphs will be saved.
• log_location: where log files will be created. "{}" is the current date.
• decklist_directory: directory containing decklists to check for overlap, when running on --mode overlap.
• aggregate_data_loc: path to CSV containing the data for all decklists, for --mode analysis.
• cluster_percentage: the percentage (interior-total) of weighted edges each node should have.
• deck_percentage: what percentage of a deck should be in cluster X for it to count as a cluster X deck.

Usage

Usage is perfomed through the command line. Here, python3 will be used to avoid ambiguity, but depending on your installation of Python it may be python. Running python3 main.py -h produces the following:

usage: main.py [-h] {overlap,analysis} ...

optional arguments:
  -h, --help          show this help message and exit

mode:
  {overlap,analysis}  Whether the program runs on overlap or analysis: overlap - Discover all files in the decklist_directory specified by CONFIG.py and
                      return the cards that they share. analysis - Cluster decklists

python3 main.py overlap -h has the following output:

usage: main.py overlap [-h] [--all]

optional arguments:
  -h, --help  show this help message and exit
  --all, -a   Check for overlap in all decks of specifed aggregate location.

python3 main.py analysis -h produces:

usage: main.py analysis [-h] [--profile PROFILE PROFILE | --resolution_parameter RESOLUTION_PARAMETER] [--graph] [--no-lands] [--label] [--colour]
                        [--start START]

optional arguments:
  -h, --help            show this help message and exit
  --profile PROFILE PROFILE, -p PROFILE PROFILE
                        Run a profile and create a graph of number of clusters against resolution_parameter. The lower and upper bounds are specified after
                        the parameter is passed. This disables the standard output.
  --resolution_parameter RESOLUTION_PARAMETER, -r RESOLUTION_PARAMETER
                        Set the resolution_parameter used in the program to the value passed.
  --graph, -g           Produce a visual for the clusters and a .graphml file in graphml_location (config).
  --no-lands            Remove all land cards from the decks to produce clusters of only nonland cards.
  --label               Ask the user for names for each cluster detected
  --colour              Ask the user for colours to plot each cluster's wedge in.
  --start START, -s START
                        Where initial card cluster membership is stored.

Example usage

python3 main.py overlap python3 main.py analysis -r 0.9 --label --no-lands --graph

Issues and bug reports

Issues and bug reports, as well as suggestions, can be filed at https://github.com/Jlobblet/decklist_analyser/issues

Contributing

Contribution should be done through the repository at https://github.com/Jlobblet/decklist_analyser.