| > The GHC versions are simply more efficient versions. Show and Read are | > completely portable. | | AIIEE! Now they tell me -- I've optimized one of my programs, and it | now spends most of it's time 'read'ing. So the thing is just to | import GHC(read) instead of using the Prelude version? I guess it's | worth a try (although I expect I really need to go to binary IO) No, you get the efficient ones when you import the Prelude. GHC's Read class has two extra methods, readPrec and readListPrec, which read using a parser monad rather than the 'list of successes' that is (sadly) built into the definition of the standard Read class. When defining a new type, and making a Read instance for it, you'll get more efficient reading if you define GHC's Read methods (the normal readsPrec and readList will then be defined in terms of them). There are many examples in GHC.Base.Read The underlying parser technology is Koen Claessens's parallel parser http://www.cs.chalmers.se/~koen/publications.html This new parsing stuff is hardly documented at all (e.g. http://www.haskell.org/ghc/docs/latest/html/base/GHC.Read.html#readPrec) . If anyone would like to remedy this by writing Haddock documentation, it'd be great. (I'd be glad to hand-hold.) Simon