Hi Christian, Your suggestion is of course possible. However, it imposes an overhead which I would like to avoid. Consider you have a data structure (e.g. an abstract syntax tree) consisting of nodes belonging to different data types. The difficulty is to write a function that abstracts from the traversal of that tree. I.e. try to write a map and a filter function on such data structures (what's the type of the function argument to map/filter?). In using a discriminated union type as you propose, one solution would be to traverse the data structure once and generate for each node a corresponding "tagged" node (a B becomes an UB B etc.). The result can then be mapped over or filtered by a function of type A -> A or A -> [A]. Finally, the result has again to be untagged to get a structure of the original types. Creating the tagged variant from the beginning is undesirable because this would lead to less type safety in the construction and node-specific manipulation of the data structure. These disadvantages could be avoided when (undiscriminated) union types were allowed. Bernd Christian Lescher wrote:

Hi Bernd,

...

data B = ... data C = ... type A = B | C | D -- A accepts values of either B or C or D (cf. the "Either a" type in the Prelude)

What about the construction of a union type in Haskell like this?:

data B = B1 | B2 data C = C1 | C2 | C3 data D = D1 | D2 data A = UB B | UC C | UD D -- here's the union type

For instance a list of type [A] may look like this: [UB B1,UC C2,UC C1,UD D1]

Christian

_______________________________________________ Haskell mailing list Haskell@haskell.org http://www.haskell.org/mailman/listinfo/haskell