The algorithm.py calculates optimal number of representatives in each row for several numbers of rows (+ size of icons and gap between the rows). These numbers are used as parameters for the chart.
It may be slow for big parliaments, but it is needed just once for any number (e.g., 200 representatives took about 1 hour, due to the grid search - further optimization possible, my trial using steepest descent algorithm did not converge many times).