I looked at the code for a bit, and it's just like most academic code I've ever seen/written: incredibly dense and cryptic. What is the likelihood that he has

I looked at the code for a bit, and it's just like most academic code I've ever seen/written: incredibly dense and cryptic. What is the likelihood that he has a transcription error somewhere? The average number of characters in the main file accel_ortho_wb.py is 52 characters per line (5 of the lines have over 300 characters). Aside from the question of whether the code is semantically/notionally correct, what about numerical precision issues? Python floats are 64-bit IEEE doubles. I've read through enough "floating point representations are fraught with peril" articles in the past to be default-skeptical of code written by people who haven't already run into--and addressed--computational issues specifically with them. I've never used numpy or done any hard mathematical analysis either in python or anything else. Does numpy somehow change the float type in python to use some other representation than doubles? How much precision is in a double? Is there "enough" for these astronomically large(/small) calculations? I guess a 52-bit mantissa can represent 2^52 = 4,503,599,627,370,496 ~= 10^15.65356. Is that ~"15 and a half digits" of precision? How does error propagate through all the calculations? edit: I don't want to just poke holes in this person's research. I'm crossing my fingers and hoping it's all correct both for their sake and for mankind's.