This program (src/venmo_degree.py) calculates the rolling median degree of nodes representing users of the Venmo payment service within a 60-second window of the latest transaction. For the run.sh demonstration, the input is venmo_input/venmo-trans.txt and the output, venmo_output/output.txt.
Python 3.5 or later is required for the deque object's insert method. The program checks for this requirement on startup and will not proceed further if it is not met. There are no other requirements.
I aim for the sweet spot between performance and memory-efficiency. Without rigorous benchmarking, I estimate that every step of the program is carried out in roughly constant time (O(1)), with a modest memory requirement. The largely repetitive degrees are efficiently stored in a frequency table (implemented as a humble list), which yields automatic sorting. Memory footprint is further reduced by storing edges as sorted tuples instead of frozensets.
Only a transaction with a valid payor, payee, and timestamp is processed. The payor and the payee cannot be the same person in a transaction.