#14084: Strange behavior of GHC by writing the types in GHCi -------------------------------------+------------------------------------- Reporter: vanto | Owner: (none) Type: bug | Status: closed Priority: normal | Milestone: Component: GHCi | Version: 8.0.2 Resolution: invalid | 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 AntC): Replying to [comment:3 vanto]:
However something is not right.\\
Chiefly, that you do not understand Haskell's typeclass system and overloading. GHCi is great for 'quick qnd dirty' experimenting; but I note you haven't actually written a program yet. So to head you off from alleging more bugs that are not bugs ...
1:2 is not a list. Where is the empty list [] ?\\
If you try to put (say) `x = 1:2` in a program, it won't compile. So you can not make any claims about what `1:2` is or is not. What GHCi's `:t 1:2` is telling you is: * if you supply some specific type (call it `a0`); * and you have `instance Num a0` and `instance Num [a0]`; * then you'll have a list type `[a0]`. Note there is no `instance Num a => Num [a]` in the Prelude, but for all GHCi knows, perhaps you're about to load a program that supplies one. Perhaps the reason you're asking for `:t 1:2` is precisely because you're doing that, and you want to check first. It would be really annoying if GHCi didn't play ball.
GHC should not be able to evaluate this expression which is poorly written. And yet it does. It's amazing.\\
No it doesn't evaluate nothing. `:t` is not evaluating (as I said); it's telling you what you need to supply, in order to evaluate. You've tried putting `1:2` and you got an error message; no evaluating. Even if you `:set -FlexibleContexts`, you'll still get a (different) error message. It'll (probably) talk about `No instance for Show [a0] ...`. That's not a bug; and it's not evaluating. As another example of that, put `x = 1:2:[]` in a program and compile it. Then ask `:t x`. Please do not report that as a bug. You'll be learning about default typing.
It can with numbers but it can not with characters.\\
Who says characters are not `Num`eric? Try this at the prompt: `:t 1 + 'c'`. This is also not a bug.
The ML and Miranda compilers, for example, do not manage to evaluate this kind of poorly written list, as well as with numbers or characters.\\
Haskell does not evaluate this kind of poorly written list. (Unless you supply extra instances over what's in the Prelude.) Same same.
Intellectually this should be considered a bug.\\
That's a comment on your intellect. You could go and find a copy of Hugs. That represents Haskell of ~10 years ago. You'll see exactly the same behaviour. You could go and read Wadler&Blott's 1988 paper that formalised the typeclass system (and compared to ML and Miranda). You'll see exactly the same, and why Haskell is a huge improvement.
I'm going to open a new bug ticket soon. And that is not a request for information.
Then you're going to be wasting a lot of people's time. Putting `1:2` at the GHCi prompt is a way to start learning, OK. You've spent 2~3 days at it. Most people get over it in an hour or so. I suggest you get on with some proper programming. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14084#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler