Use generators in the parser

[pR0Ps]

A lot of the data the parser gives back is gone through with for x in data loops.

Making the parser yield items instead of having to initially compute all of them and store them in memory would probably be more efficient.

[uniphil]

Awesome. I would love to see a comparisson of the mem usage before/after this change! Maybe I'll have some time to try it next weekend, if no one gets to it first.