Hello S\'ebastien,
I am a Haskell newbie, but have been interested in Haskell (and generally speaking ML-derivates) for some time. I am currently evaluating different languages for implementing an application which will have to manipulate large graphs representing the structure of programs and their evolution.
Maybe you should consider using a combination of Haskell and C++. I bet there are good graph libraries in C++ available so that would save a lot of work. Using Haskell for the front end is a lot more fun than C++ or Java, I think. As you say you are a Haskell newbie I must warn you that linking to a C++ library is not trivial and that's not a Haskell problem but more a compiler-options-problem. What I suggest is exactly what I am doing at the moment; I am building a tool to edit Bayesian networks and to apply inference algorithms to them. The GUI is written in Haskell (wxHaskell) and the inference is a C++ library I am reusing. The choice for C++ was not speed but reuse. I have never come across a problem for which Haskell was too slow, but maybe that depends on my choice of problems. The tool I am writing now runs smoothly on an ancient (700Mhz) computer. Good luck, Arjan