#2806: Require bang-patterns for unlifted bindings
---------------------------------+------------------------------------------
Reporter: simonpj | Owner:
Type: feature request | Status: new
Priority: normal | Milestone: 6.12.1
Component: Compiler | Version: 6.8.3
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Os: Unknown/Multiple
Architecture: Unknown/Multiple |
---------------------------------+------------------------------------------
Comment (by simonpj):
We're getting ourselves confused here. In the rest of this comment I'll
write down a more formal account of what I think The Rules should be. If
we can agree them, then we can implement them and put them in the user
manual.
First, let's agree that the semantics of bang patterns is defined here:
* http://www.haskell.org/ghc/docs/latest/html/users_guide/bang-
patterns.html
This ticket is concerned with patterns that bind variables of unlifted
type (e.g. `Int#`). I propose the following rule:
* A pattern `p` is '''strict down to''' `v` iff `v` does not appear
under a lazy-pattern tilde "`~`".
* A variable `v` of unlifted type may be bound by a pattern only in the
following situations:
* In a lambda or `case`, only by a pattern that is strict down to `v`.
* In a `let` or `where`, only by a binding `!p = e` where `p` is
strict down to `v`.
Rationale: variables of unlifted type cannot be bound to a thunk, so the
pattern must eagerly match all the way down to `v`.
Examples. Suppose `v::Int#`. Then here are some legal bindings:
{{{
\v -> e
\(x, I# v) -> e
let !v = e in b
let !(x, I# v) = e in b
\(x, !(I# v)) -> e -- The inner ! is legal but not required
\(x, I# !v) -> e -- The ! on v is legal but not required
}}}
Here are some illegal ones
{{{
\~(I# v) -> e -- Under a tilde
let I# v = e in b -- No bang on the pattern binding
let !(x, ~(I# v)) = e in b -- Under a tilde
}}}
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2806#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
