Hugh Perkins wrote:
Cant decide whether this list is appropriate for questions related to generics usage, or only to generics design?
I think your question is quite appropriate -- actually, it is a good question. It asks essentially for a monadic gmap, which seems quite important. This operation is also the `inverse' of one of our comparison tests: listify. The latter collects values from the leaves of a data type in a list. That list is a `state' of a value, so to speak. The state can be transported over the network, and, on the other end, can be infused into a prototype object (assuming it has the same structure). The monadic gmap subsumes two traversals: generic map and reduction. So, your question looks even better. I wonder if we should add monadic gmap to our test suite? I have just committed the monadic gmap implementation and the test for one generic library, the one I'm more familiar with. The test is in the file http://darcs.haskell.org/generics/comparison/SmashA/Deserialize.hs We start with the data type of a company C [D "Research" (E (P "Laemmel" "Amsterdam") (S 8000.0)) [PU (E (P "Joost" "Amsterdam") (S 1000.0)), PU (E (P "Marlow" "Cambridge") (S 2000.0))], D "Strategy" (E (P "Blair" "London") (S 100000.0)) []] (see http://darcs.haskell.org/generics/comparison/CompanyDatatypes.hs) We can serialize it: test1 = serialize genCom obtaining a list ["Research","Laemmel","Amsterdam","8000.0","Joost","Amsterdam", "1000.0","Marlow","Cambridge","2000.0","Strategy","Blair", "London","100000.0"] OTH, we can take the list retro = ["Metaphysics", "Kant","Koeningsberg","800.0", "Hume","Edinburgh","100.0", "Marlowe","Cambridge","200.0", "Ruling","Thatcher","London","50000.0"] and use it to `upgrade' the company genCom: test2 = deserialize genCom retro giving us *Deserialize> test2 C [D "Metaphysics" (E (P "Kant" "Koeningsberg") (S 800.0)) [PU (E (P "Hume" "Edinburgh") (S 100.0)), PU (E (P "Marlowe" "Cambridge") (S 200.0))], D "Ruling" (E (P "Thatcher" "London") (S 50000.0)) []]