The initial blog post launching the survey is here: https://thephd.dev/the-big-array-size-survey-for-c.
The results are discussed in paper N3440 in brief in a later post here: https://thephd.dev/the-big-array-size-survey-for-c.
The main.py file parses a (not provided) file from AllCounted to get out information. The CSV -- along with the initial graphics in this repository -- are generated from that AllCounted information. The AllCounted data contains identifying information, and so cannot be released. The CSV is entirely anonymous and contains no identifying information.
The CSV contains a single header at the start identifying what each column represents in the data. Everything except the Geographic Distributions Map and the City Word Cloud can be replicated from the CSV data.