Ivan Miljenovic wrote:
How I envisage it happening is that a parser would be used to find all "functions" in the given code, treat these as nodes in the graph and then use directed edges to indicate which functions call other functions. This resultant graph can then be analysed in various ways suitable to the context (e.g. find that a library module can be split into two since there are two completely separate trees present in the graph that don't interact at all, or if a function is only ever called by one other function then it can be subsumed into it).
It just occurred to me that this idea is more general than the control or data flow analysis that are the focus of similar ideas I've seen before. For example, you could trace type usage through the code (which would likely be a subset of the control flow for Haskell, but an interesting subset nonetheless). :-) -- Tommy M. McGuire mcguire@crsr.net