After studying the code, I've realized that eclipsefp includes a full Antlr parser for Haskell source files, and a full AST object hierarchy, that are not nuch used. Apart from verification that parsing succeeds, we only seem to use the language models for code completion. We don't even use the parsing for syntax highlighting, which explains why for example a single quote is not hightlighted properly. So I'm going to remove all that code. We use scion and the GHC API to parse and typecheck files, so we can use that to get a list of names available for code completion.
If anybody can think of a reason to keep that code active, please shout!