In Chocolat we take a similar approach, but the histogram is of the GCDs, not of the indentations. So we just take a sample of 1000 lines, then take the GCD of

In Chocolat we take a similar approach, but the histogram is of the GCDs, not of the indentations. So we just take a sample of 1000 lines, then take the GCD of every pair. The indentation width is simply the most common pair. If it turns out that one bin is not clearly more dominant than the others, then we default to whatever was used last time for that language.