On Thu, Sep 15, 2011 at 7:51 AM, Markus Läll
Intuitively it seems to be exactly the same as the type families' aproach, and looks quite clear too.
Not exact, no--as written, it's strictly more powerful. Your fundeps go in both directions, whereas the type families didn't (though could easily be extended to do so, if desired, at the cost of some extra verbosity). The main argument in favor of type families here is the ~ equality constraint which, as you've found, works in combination with fundeps as well. :] You can actually simulate it with fundeps alone, but you probably don't want to. All else equal I personally find type families easier to work with, but in this case the difference is minimal. For bidirectional constraints and simple transformations, fundeps are probably a bit nicer, so in hindsight I think yours is the better idea here. - C.