Sub-second algorithm for pairing and unpairing non-negative integers. Tests execute under one second with integers up to 7MB large in my laptop. Numbers larger than that take more, even minutes.
Please contribute with more pairing and benchmarks, maybe we find an underlying bottleneck we could fix! :)
Example with numbers much larger than C ULONG_MAX for 64bit that is 18446744073709551615.
import time
import cantor
import elegant
if __name__ == '__main__':
# cantor
start = time.time()
assert (87917298379182739871892379817239871982379817239871928371923791827398172397129387192379182379182379182739182739187239817329871923791827981723,
981729837981273971298379812739817239871923791827398172398712983767468675465746172649162486347649289817239) == \
cantor.unpair(cantor.pair(87917298379182739871892379817239871982379817239871928371923791827398172397129387192379182379182379182739182739187239817329871923791827981723,
981729837981273971298379812739817239871923791827398172398712983767468675465746172649162486347649289817239))
end = time.time()
print('cantor\t', '{0:.10f} secs'.format(end - start))
# elegant
start = time.time()
assert (87917298379182739871892379817239871982379817239871928371923791827398172397129387192379182379182379182739182739187239817329871923791827981723,
981729837981273971298379812739817239871923791827398172398712983767468675465746172649162486347649289817239) == \
elegant.unpair(elegant.pair(87917298379182739871892379817239871982379817239871928371923791827398172397129387192379182379182379182739182739187239817329871923791827981723,
981729837981273971298379812739817239871923791827398172398712983767468675465746172649162486347649289817239))
end = time.time()
print('elegant\t', '{0:.10f} secs'.format(end - start))
# Results:
# cantor 0.0001599789 secs
# elegant 0.0000128746 secsCheck test.py for more examples of usage. This package will be made available over pypi once some more pairing algos are added. If you want to publish it, please be kind to let me know before hand. :)
PS.: Python 3.8 includes math.isqrt, supposedly slower than the gmpy2 library, dependency for this project. For higher portability just upgrade python and change the libraries to remove this dependency. Kudos to Anders Kaseorg and Mathmandan for pointing this out this fantastic library.