In English, the reporter is suggesting that the Monad instance of list be rewritten so that it generates a fairer distribution of tuples (in the example above, all tuples generated are of the form `(x, 1)`). The fairness of these types of nondeterminism monads is indeed something that has been investigated in the past (e.g., http://okmij.org/ftp/Computation/LogicT.pdf). Unfortunately, the semantics of the List monad as defined in Haskell are fairly set in stone at this
#13783: 更合理的instance Monad [] -------------------------------------+------------------------------------- Reporter: zaoqi | Owner: (none) Type: feature request | Status: closed Priority: normal | Milestone: Component: Core Libraries | Version: 8.0.1 Resolution: wontfix | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Incorrect result | Unknown/Multiple at runtime | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by zaoqi): Replying to [comment:1 ezyang]: point, and can't be changed without massively breaking existing code. The best course of action is probably to define your own monad with the necessary fairness (do consider using CPS; it's a lot more efficient!) What's CPS? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13783#comment:2 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler