To compile the tester (tested on Ubuntu 20.04, gcc version 9.3.0):
$ make tester

The tester run the clustering algorithms on the instance specified by command line args:
$ ./tester -ds=data/real/4area.ds -k=10 -z=0.05 -colorless=tmp/colorless -colored=tmp/colored -fairz=tmp/fairz -colorful=tmp/colorful -coreset=tmp/coreset -overwrite=1

The dataset must be a .ds file (described later), k is the number of clusters and z is the fraction of outliers. For the colored version, each z_a = z. Without -overwrite=1 solutions already computed will be skipped.
The datasets used in the article are available as a compressed file in the data/real folder.
If present, each of the following arguments specifies the output directory for the respective test:
 - colorless: run the baselines on the colorless instance
 - colored: from the colorless, force the fairness and recompute the clustering (using the same centers)
 - fairz: from the colorless, recompute discard extra outliers due to rounding in the color instance; since the number of outliers is ceiled in the colorful version, the colorless version may have less outliers: this fixes it.
 - colorful: run our algorithm
 - coreset: compute the coreset and run all the algorithms on it

For each algorithm, the corresponding solution is saved as a JSON file in the specified directory. This file contains all the information about a solution and can be used to compare the performance of the different algorithms.

The datasets are .ds files, formatted as:
    # Comment lines
    #@meta_key_1:value_1
    #@meta_key_2:value_2
    Name Size Dimensions
    ColorClasses ColorClass1Name(str) ColorClass2Name ...
    PointID X_1 ... X_Dimensions Color_1 .. Color_ColorClasses
    ...
With meta being arbitrary strings (if needed), Name is the dataset name (e.g. adult), Size is the number of points, and Dimensions the dimensions of each point.
ColorClasses is the number of different colors, followed by the name of each color class (e.g. for adult: 2 Sex Race).
Each point is then described as an id, followed by the point coordinates, followed by the point colors (treated as strings). For example, the first lines of adult.ds:

adult 45222 6
2 race sex
0 0.034200950931070354 -1.0622948702184187 1.1287528085714997 0.14288836417960854 -0.2187802566102861 -0.07812006020555637 White Male
1 0.8664168406269813 -1.00743773062647 1.1287528085714997 -0.14673320201323659 -0.2187802566102861 -2.3267380094882686 White Male

If a new dataset is used, it must be first preprocessed to get the required data using:
$ make minmax_dist
$ ./minmax_dist <dataset_path.ds>
In order to create the <dataset_path.ds>.dist file needed by the tester. It contains statistics about min and max distances among the points.