Ah, thanks, folks! I'll just implement my own hashing by generating a string and calling the hash function on that. That's what I was doing in the old version of my code, anyway.
It's just that in the core Data.HashTable, you had to provide a hash function, so the point where I used the hash table was able to call a string conversion function that was defined elsewhere. With the hashtable package, you have to define a Hashable instance in the same package as your datatype definition - which is not where the string conversion is implemented. Calling the string conversion function leads to a cyclic dependency that I wanted to avoid. So I'll have to maybe move it to live with the datatype definition, or duplicate it, or use some other means of hashing. In other words, package A defines datatype A. Package B defines A -> String. Package C creates a HashTable that indexes by As, and also imports package B, so it can build a hash function as the composition of the string conversion and the provided string hashing. But now I need to define the Hashable instance in A, which doesn't have access to package B, since package B also depends on A, and I don't like circular dependencies. - Lyle
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
