On Wednesday, September 14, 2016 at 7:12:31 AM UTC-4, Neal Becker wrote: > > I'd expect it to be called with a large array as input, so hopefully it's > performant enough as is, leaving the type of the argument determined at > runtime:
My issue is more with how it is used. The whole constellation type seems oriented towards the "vectorized" style of constantly constructing new arrays [c.C[x+1] for x in xsyms] (aside: it seems like OOP overkill to define a whole type etcetera just to implement this simple one-line function), whereas you can usually get better performance by combining this kind of trivial array construction with the subsequent computation on the data.