Complexity of standard Python data structures

[Delaney, Timothy C (Timothy)]

> From: Michael Hudson [mailto:mwh at python.net]
> 
> I'm afraid your line of questioning slightly confuses me.  The Python
> language is not specified, in any precise sense of the word, anywhere.
> For one thing, it's evolving.  If some super new implementation of
> hash tables appeared[1] that had somewhat different algorithmic
> behaviour but were considered superior, they might well turn up in
> Python 2.5 (say) and *if* the closest thing Python has to a spec (the
> language reference) described the algorithmic nature of various
> operations (which it doesn't) it would then be changed.

One of the big advantages of this is that is assists greatly in trying things out in different implementations, etc.

For example, PyPy can implement something in the simplest way that *works* (matches the CPython semantics) early in the development cycle.

After getting much more of the system to work, the PyPy developers can optimise those parts that need it.

Likewise, the CPython developers can put something in early which does what is needed, in order to build infrastructure for a new feature. It can later be reworked to give performance improvements, if needed.

Tim Delaney