Add type caching to documentation parser

[blablabla1234678]

I did not want to process the entire documentation because it can be big. It is faster this way, but with the current approach type caching might be needed.