Changing strings to tuples has far-reaching effects (PySNMP optimisation might constitute a fork?)

Today I spent almost exclusively on getting the tuple-based OID code worked through PySNMP, TwistedSNMP and Cinemon. The changes were far more extensive than I was hoping. They are mostly integrated, but it's going to take a few more hours to track down all of the string-assumptive code.

Anyway, the OID class now seems fairly reasonable, but I haven't yet got a handle on how much of a speedup it will allow, the change in TwistedSNMP's test suite is tiny, but then that suite isn't what was showing 33% for the OID isaprefix() method. We'll see when the conversion is complete, which will likely be Thursday, as tomorrow we have an all-day meeting scheduled.

Biggest problem as I see it now is that client code will have to change. Anything that assumed a string is going to have to switch to assuming a tuple... that's not great :( .


Comments are closed.


Pingbacks are closed.