I recently I had the opportunity to speed up some Websocket code that was a major bottleneck. The final solution was 60X (!) faster than the first pass, and an interesting exercise in optimizing inner loops.

When you send data via a websocket it is masked with a randomly generated four byte key. This masking provides protection from certain exploits against proxies (details here). The algorithm is super simple; essentially each byte in the payload is XORed with a corresponding byte from the key.

The first version of my XOR mask function looked this:

That's the kind of unfussy and elegant code you want to write for a job interview. If that's the solution that first came to mind, then you know your Python. Alas, this code is as slow as it is beautiful. continue reading…