August 2013 - ghc-tickets

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by cactus): I realize I was wrong on my last comment: let's not even discuss exporting data constructors as pattern-only here, otherwise this ticket will lose focus. So let's just come up with some syntax to discern unidirectional vs. bidirectional pattern synonym definitions. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:29> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by igloo): Ah, I see, I had expected that something defined with "pattern" (e.g. `pattern Arrow t1 t2 = App "->" [t1, t2]`) could only be used as a pattern. I think a different keyword, e.g. "synonym" (or perhaps "view", as Simon suggested) would be better in that case. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:28> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by simonpj): Replying to [comment:25 cactus]: > > We have at various times discussed how to allow a module to export a data constructor that can be used as a pattern but not as a constructor (in expressions), because the data type is abstract and you should only use a smart constructor. With pattern synonyms you can (almost) do that > > What if you could just export `pattern P` (where `P` is the name of the data constructor) without having to define `AbsP`? Would that be taking things too far? That would treat pattern synonyms and data constructors differently, which I am keen to avoid. Under what you propose: * If P is a data constructor then `pattern P` would export P uni- directionally (pattern only) * If P is a (bi-directonal) pattern synonym then `pattern P` would export P bi-directionally. I don't like that... a data constructor is precisely a (degenerate) bidirectional pattern. I suppose you could put the directionality in the export list: {{{ module M( pattern P, -- Uni-directional constructor Q -- Bidirectional ) where pattern P x = [x] pattern Q y = [y,y] }}} But I don't want to go this way. If you export P you get all of what P is. Otherwise we'll get into exporting it unidirectionally from one module, bidirectionally from another, then importing both of those, and combining the directinoality of those imports. Too complicated. Maybe `datacon P`, or `view P`, rather than `pattern P` would address Ian's worry? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:27> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by simonpj): Replying to [comment:24 igloo]: > I haven't followed the design, but I would expect exporting "pattern P" to either be an error ("P is a data constructor, not a pattern") or to export P in a way that allows it only to be used as a pattern, not an expression. Well in the design here [wiki:PatternSynonyms], `pattern P` exports P for use both in patterns and expressions. By all means suggest alternative designs; that's what this thread is all about. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:26> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by cactus): Replying to [comment:23 simonpj]: > I suggest we allow the user to control whether a pattern synonym is bidirectional or not. I'm not sure what syntax to use for that. We want to be able to say > * Uni-directional pattern synonym > * Bi-directional pattern synonym > * Bi-directional pattern synonym with explicit user code for the reverse direction > Once a syntax for this is decided, this is trivial to add to my current implementation. > We have at various times discussed how to allow a module to export a data constructor that > can be used as a pattern but not as a constructor (in expressions), because the data type is abstract and you should only use a smart constructor. With pattern synonyms you can (almost) do that What if you could just export `pattern P` (where `P` is the name of the data constructor) without having to define `AbsP`? Would that be taking things too far? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:25> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by igloo): I haven't followed the design, but I would expect exporting "pattern P" to either be an error ("P is a data constructor, not a pattern") or to export P in a way that allows it only to be used as a pattern, not an expression. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:24> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by simonpj): We must treat data constructors uniformly with pattern synonyms. For example, it should definitely be OK to have {{{ module M( pattern P, foo, bar ) where -- The "pattern P" exports data constructor P data T = P | Q foo :: T foo = P bar :: T -> Int bar P = 1 bar Q = 2 module A where import M -- T is not in scope w1 = bar P -- Uses P as an expression w2 = case foo of P -> 1 -- Uses P as a pattern }}} We have at various times discussed how to allow a module to export a data constructor that can be used as a ''pattern'' but not as a ''constructor'' (in expressions), because the data type is abstract and you should only use a smart constructor. With pattern synonyms you can (almost) do that: {{{ module M( T, pattern AbsP, pattern Q ) where data T = P Int | Q pattern AbsP = P }}} But I really wanted `AbsP` to be uni-directional and according to the current spec [wiki:PatternSynonyms] it is "implicitly bidirectional". I suggest we allow the user to control whether a pattern synonym is bidirectional or not. I'm not sure what syntax to use for that. We want to be able to say * Uni-directional pattern synonym * Bi-directional pattern synonym * Bi-directional pattern synonym with explicit user code for the reverse direction -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:23> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #5144: Pattern synonyms
by GHC 22 Aug '13

22 Aug '13

#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by cactus): If your question is whether there's a syntactic way to see if a given pattern synonym import can be used as an expression, then no, not at the moment. So if you do something like {{{ import M(pattern P) x = P }}} then the only error you will get if `P` is not bidirectional is from the type checker: {{{ /tmp/hs/M2.hs:5:5: P used in an expression, but it's a non-bidirectional pattern synonym In the expression: P In an equation for ‛x’: x = P }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/5144#comment:22> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #8039: RTS linker: unloadObj() does not actually unload the code
by GHC 22 Aug '13

22 Aug '13

#8039: RTS linker: unloadObj() does not actually unload the code -------------------------------------+------------------------------------ Reporter: simonmar | Owner: simonmar Type: task | Status: new Priority: high | Milestone: 7.8.1 Component: Runtime System | Version: 7.6.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by Simon Marlow <marlowsd@…>): In [changeset:269c89062dd5b6d2b8bc4d41e5dca0eca2308d02/testsuite]: {{{ #!CommitTicketReference repository="testsuite" revision="269c89062dd5b6d2b8bc4d41e5dca0eca2308d02" Add a test for unloading object files in the linker (#8039) }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8039#comment:12> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #8039: RTS linker: unloadObj() does not actually unload the code
by GHC 22 Aug '13

22 Aug '13

#8039: RTS linker: unloadObj() does not actually unload the code -------------------------------------+------------------------------------ Reporter: simonmar | Owner: simonmar Type: task | Status: new Priority: high | Milestone: 7.8.1 Component: Runtime System | Version: 7.6.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by Simon Marlow <marlowsd@…>): In [changeset:bdfefb3b72a71cd0afca6e7766456c0d97c47c86/ghc]: {{{ #!CommitTicketReference repository="ghc" revision="bdfefb3b72a71cd0afca6e7766456c0d97c47c86" Really unload object code when it is safe to do so (#8039) The next major GC after an unloadObj() will do a traversal of the heap to determine whether the object code can be removed from memory or not. We'll keep doing these until it is safe to remove the object code. In my experiments with GHCi, the objects get unloaded immediately, which is a good sign: we're not accidentally holding on to any references anywhere in the GHC data structures. Changes relative to the patch earlier posted on the ticket: - fix two memory leaks discovered with Valgrind, after testing with tests/rts/linker_unload.c }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8039#comment:11> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0