Don’t you just hate it when you have code in lots of different files that you want to work with, but you only need one function in each file? Recently I was working on an abstract syntax tree which is constructed from XML, and the lead developer suggested that I put a function in each node type called parseXML. I suggested that it would be better to have an loader with a function for each node type — the same system, but all the loader code is in one place, and the nodes don’t need to know anything about the xml that they come from. This is a pretty standard thing in compiler design, I remember creating lots of AST visitors when I was building a compiler.
But there are a few disadvantages: you need to make public setters for object properties, even when you ‘really’ want the properties to be private. And you can’t see the code of the node while you are working on the loader. There is a way around the first thing (a true visitor implementation and pass by reference) — but the second thing is a consequence of our folder-based way of looking at code.
So I was thinking that Tiddlywiki would be an interesting way to edit code — you could call up exactly the functions that you want, in the order you want.
The big problem is getting the functions back into the correct files, and doing just-in-time parsing. I suppose that prettyprinting could be hacked into tinymce — there’s a css stylesheet / javascript for prettyprinting on google code.
