2 <indexterm><primary>language, GHC</primary></indexterm>
3 <indexterm><primary>extensions, GHC</primary></indexterm>
4 As with all known Haskell systems, GHC implements some extensions to
5 the language. To use them, you'll need to give a <option>-fglasgow-exts</option>
6 <indexterm><primary>-fglasgow-exts option</primary></indexterm> option.
10 Virtually all of the Glasgow extensions serve to give you access to
11 the underlying facilities with which we implement Haskell. Thus, you
12 can get at the Raw Iron, if you are willing to write some non-standard
13 code at a more primitive level. You need not be “stuck” on
14 performance because of the implementation costs of Haskell's
15 “high-level” features—you can always code “under” them. In an extreme case, you can write all your time-critical code in C, and then just glue it together with Haskell!
19 Executive summary of our extensions:
25 <term>Unboxed types and primitive operations:</Term>
27 <para>You can get right down to the raw machine types and
28 operations; included in this are “primitive
29 arrays” (direct access to Big Wads of Bytes). Please
30 see <XRef LinkEnd="glasgow-unboxed"> and following.</para>
35 <term>Type system extensions:</term>
37 <para> GHC supports a large number of extensions to Haskell's
38 type system. Specifically:</para>
42 <term>Multi-parameter type classes:</term>
44 <para><xref LinkEnd="multi-param-type-classes"></para>
49 <term>Functional dependencies:</term>
51 <para><xref LinkEnd="functional-dependencies"></para>
56 <term>Implicit parameters:</term>
58 <para><xref LinkEnd="implicit-parameters"></para>
63 <term>Linear implicit parameters:</term>
65 <para><xref LinkEnd="linear-implicit-parameters"></para>
70 <term>Local universal quantification:</term>
72 <para><xref LinkEnd="universal-quantification"></para>
77 <term>Extistentially quantification in data types:</term>
79 <para><xref LinkEnd="existential-quantification"></para>
84 <term>Scoped type variables:</term>
86 <para>Scoped type variables enable the programmer to
87 supply type signatures for some nested declarations,
88 where this would not be legal in Haskell 98. Details in
89 <xref LinkEnd="scoped-type-variables">.</para>
97 <term>Pattern guards</term>
99 <para>Instead of being a boolean expression, a guard is a list
100 of qualifiers, exactly as in a list comprehension. See <xref
101 LinkEnd="pattern-guards">.</para>
106 <term>Data types with no constructors</term>
108 <para>See <xref LinkEnd="nullary-types">.</para>
113 <term>Parallel list comprehensions</term>
115 <para>An extension to the list comprehension syntax to support
116 <literal>zipWith</literal>-like functionality. See <xref
117 linkend="parallel-list-comprehensions">.</para>
122 <term>Foreign calling:</term>
124 <para>Just what it sounds like. We provide
125 <emphasis>lots</emphasis> of rope that you can dangle around
126 your neck. Please see <xref LinkEnd="ffi">.</para>
133 <para>Pragmas are special instructions to the compiler placed
134 in the source file. The pragmas GHC supports are described in
135 <xref LinkEnd="pragmas">.</para>
140 <term>Rewrite rules:</term>
142 <para>The programmer can specify rewrite rules as part of the
143 source program (in a pragma). GHC applies these rewrite rules
144 wherever it can. Details in <xref
145 LinkEnd="rewrite-rules">.</para>
150 <term>Generic classes:</term>
152 <para>(Note: support for generic classes is currently broken
155 <para>Generic class declarations allow you to define a class
156 whose methods say how to work over an arbitrary data type.
157 Then it's really easy to make any new type into an instance of
158 the class. This generalises the rather ad-hoc "deriving"
159 feature of Haskell 98. Details in <xref
160 LinkEnd="generic-classes">.</para>
166 Before you get too carried away working at the lowest level (e.g.,
167 sloshing <literal>MutableByteArray#</literal>s around your
168 program), you may wish to check if there are libraries that provide a
169 “Haskellised veneer” over the features you want. See
170 <xref linkend="book-hslibs">.
173 <sect1 id="options-language">
174 <title>Language options</title>
176 <indexterm><primary>language</primary><secondary>option</secondary>
178 <indexterm><primary>options</primary><secondary>language</secondary>
180 <indexterm><primary>extensions</primary><secondary>options controlling</secondary>
183 <para> These flags control what variation of the language are
184 permitted. Leaving out all of them gives you standard Haskell
190 <term><option>-fglasgow-exts</option>:</term>
191 <indexterm><primary><option>-fglasgow-exts</option></primary></indexterm>
193 <para>This simultaneously enables all of the extensions to
194 Haskell 98 described in <xref
195 linkend="ghc-language-features">, except where otherwise
201 <term><option>-fno-monomorphism-restriction</option>:</term>
202 <indexterm><primary><option>-fno-monomorphism-restriction</option></primary></indexterm>
204 <para> Switch off the Haskell 98 monomorphism restriction.
205 Independent of the <option>-fglasgow-exts</option>
211 <term><option>-fallow-overlapping-instances</option></term>
212 <term><option>-fallow-undecidable-instances</option></term>
213 <term><option>-fcontext-stack</option></term>
214 <indexterm><primary><option>-fallow-overlapping-instances</option></primary></indexterm>
215 <indexterm><primary><option>-fallow-undecidable-instances</option></primary></indexterm>
216 <indexterm><primary><option>-fcontext-stack</option></primary></indexterm>
218 <para> See <xref LinkEnd="instance-decls">. Only relevant
219 if you also use <option>-fglasgow-exts</option>.</para>
224 <term><option>-finline-phase</option></term>
225 <indexterm><primary><option>-finline-phase</option></primary></indexterm>
227 <para>See <xref LinkEnd="rewrite-rules">. Only relevant if
228 you also use <option>-fglasgow-exts</option>.</para>
233 <term><option>-fgenerics</option></term>
234 <indexterm><primary><option>-fgenerics</option></primary></indexterm>
236 <para>See <xref LinkEnd="generic-classes">. Independent of
237 <option>-fglasgow-exts</option>.</para>
242 <term><option>-fno-implicit-prelude</option></term>
244 <para><indexterm><primary>-fno-implicit-prelude
245 option</primary></indexterm> GHC normally imports
246 <filename>Prelude.hi</filename> files for you. If you'd
247 rather it didn't, then give it a
248 <option>-fno-implicit-prelude</option> option. The idea
249 is that you can then import a Prelude of your own. (But
250 don't call it <literal>Prelude</literal>; the Haskell
251 module namespace is flat, and you must not conflict with
252 any Prelude module.)</para>
254 <para>Even though you have not imported the Prelude, all
255 the built-in syntax still refers to the built-in Haskell
256 Prelude types and values, as specified by the Haskell
257 Report. For example, the type <literal>[Int]</literal>
258 still means <literal>Prelude.[] Int</literal>; tuples
259 continue to refer to the standard Prelude tuples; the
260 translation for list comprehensions continues to use
261 <literal>Prelude.map</literal> etc.</para>
263 <para> With one group of exceptions! You may want to
264 define your own numeric class hierarchy. It completely
265 defeats that purpose if the literal "1" means
266 "<literal>Prelude.fromInteger 1</literal>", which is what
267 the Haskell Report specifies. So the
268 <option>-fno-implicit-prelude</option> flag causes the
269 following pieces of built-in syntax to refer to <emphasis>whatever
270 is in scope</emphasis>, not the Prelude versions:</para>
274 <para>Integer and fractional literals mean
275 "<literal>fromInteger 1</literal>" and
276 "<literal>fromRational 3.2</literal>", not the
277 Prelude-qualified versions; both in expressions and in
282 <para>Negation (e.g. "<literal>- (f x)</literal>")
283 means "<literal>negate (f x)</literal>" (not
284 <literal>Prelude.negate</literal>).</para>
288 <para>In an n+k pattern, the standard Prelude
289 <literal>Ord</literal> class is still used for comparison,
290 but the necessary subtraction uses whatever
291 "<literal>(-)</literal>" is in scope (not
292 "<literal>Prelude.(-)</literal>").</para>
296 <para>Note: Negative literals, such as <literal>-3</literal>, are
297 specified by (a careful reading of) the Haskell Report as
298 meaning <literal>Prelude.negate (Prelude.fromInteger 3)</literal>.
299 However, GHC deviates from this slightly, and treats them as meaning
300 <literal>fromInteger (-3)</literal>. One particular effect of this
301 slightly-non-standard reading is that there is no difficulty with
302 the literal <literal>-2147483648</literal> at type <literal>Int</literal>;
303 it means <literal>fromInteger (-2147483648)</literal>. The strict interpretation
304 would be <literal>negate (fromInteger 2147483648)</literal>,
305 and the call to <literal>fromInteger</literal> would overflow
306 (at type <literal>Int</literal>, remember).
315 <!-- UNBOXED TYPES AND PRIMITIVE OPERATIONS -->
318 <sect1 id="glasgow-ST-monad">
319 <title>Primitive state-transformer monad</title>
322 <indexterm><primary>state transformers (Glasgow extensions)</primary></indexterm>
323 <indexterm><primary>ST monad (Glasgow extension)</primary></indexterm>
327 This monad underlies our implementation of arrays, mutable and
328 immutable, and our implementation of I/O, including “C calls”.
332 The <literal>ST</literal> library, which provides access to the
333 <function>ST</function> monad, is described in <xref
339 <sect1 id="glasgow-prim-arrays">
340 <title>Primitive arrays, mutable and otherwise
344 <indexterm><primary>primitive arrays (Glasgow extension)</primary></indexterm>
345 <indexterm><primary>arrays, primitive (Glasgow extension)</primary></indexterm>
349 GHC knows about quite a few flavours of Large Swathes of Bytes.
353 First, GHC distinguishes between primitive arrays of (boxed) Haskell
354 objects (type <literal>Array# obj</literal>) and primitive arrays of bytes (type
355 <literal>ByteArray#</literal>).
359 Second, it distinguishes between…
363 <term>Immutable:</term>
366 Arrays that do not change (as with “standard” Haskell arrays); you
367 can only read from them. Obviously, they do not need the care and
368 attention of the state-transformer monad.
373 <term>Mutable:</term>
376 Arrays that may be changed or “mutated.” All the operations on them
377 live within the state-transformer monad and the updates happen
378 <emphasis>in-place</emphasis>.
383 <term>“Static” (in C land):</term>
386 A C routine may pass an <literal>Addr#</literal> pointer back into Haskell land. There
387 are then primitive operations with which you may merrily grab values
388 over in C land, by indexing off the “static” pointer.
393 <term>“Stable” pointers:</term>
396 If, for some reason, you wish to hand a Haskell pointer (i.e.,
397 <emphasis>not</emphasis> an unboxed value) to a C routine, you first make the
398 pointer “stable,” so that the garbage collector won't forget that it
399 exists. That is, GHC provides a safe way to pass Haskell pointers to
404 Please see <xref LinkEnd="sec-stable-pointers"> for more details.
409 <term>“Foreign objects”:</term>
412 A “foreign object” is a safe way to pass an external object (a
413 C-allocated pointer, say) to Haskell and have Haskell do the Right
414 Thing when it no longer references the object. So, for example, C
415 could pass a large bitmap over to Haskell and say “please free this
416 memory when you're done with it.”
420 Please see <xref LinkEnd="sec-ForeignObj"> for more details.
428 The libraries documentatation gives more details on all these
429 “primitive array” types and the operations on them.
435 <sect1 id="nullary-types">
436 <title>Data types with no constructors</title>
438 <para>With the <option>-fglasgow-exts</option> flag, GHC lets you declare
439 a data type with no constructors. For example:</para>
442 data T a -- T :: * -> *
444 <para>Syntactically, the declaration lacks the "= constrs" part. The
445 type can be parameterised, but only over ordinary types, of kind *; since
446 Haskell does not have kind signatures, you cannot parameterise over higher-kinded
449 <para>Such data types have only one value, namely bottom.
450 Nevertheless, they can be useful when defining "phantom types".</para>
453 <sect1 id="pattern-guards">
454 <title>Pattern guards</title>
457 <indexterm><primary>Pattern guards (Glasgow extension)</primary></indexterm>
458 The discussion that follows is an abbreviated version of Simon Peyton Jones's original <ULink URL="http://research.microsoft.com/~simonpj/Haskell/guards.html">proposal</ULink>. (Note that the proposal was written before pattern guards were implemented, so refers to them as unimplemented.)
462 Suppose we have an abstract data type of finite maps, with a
466 lookup :: FiniteMap -> Int -> Maybe Int
469 The lookup returns <function>Nothing</function> if the supplied key is not in the domain of the mapping, and <function>(Just v)</function> otherwise,
470 where <VarName>v</VarName> is the value that the key maps to. Now consider the following definition:
474 clunky env var1 var2 | ok1 && ok2 = val1 + val2
475 | otherwise = var1 + var2
486 The auxiliary functions are
490 maybeToBool :: Maybe a -> Bool
491 maybeToBool (Just x) = True
492 maybeToBool Nothing = False
494 expectJust :: Maybe a -> a
495 expectJust (Just x) = x
496 expectJust Nothing = error "Unexpected Nothing"
500 What is <function>clunky</function> doing? The guard <literal>ok1 &&
501 ok2</literal> checks that both lookups succeed, using
502 <function>maybeToBool</function> to convert the <function>Maybe</function>
503 types to booleans. The (lazily evaluated) <function>expectJust</function>
504 calls extract the values from the results of the lookups, and binds the
505 returned values to <VarName>val1</VarName> and <VarName>val2</VarName>
506 respectively. If either lookup fails, then clunky takes the
507 <literal>otherwise</literal> case and returns the sum of its arguments.
511 This is certainly legal Haskell, but it is a tremendously verbose and
512 un-obvious way to achieve the desired effect. Arguably, a more direct way
513 to write clunky would be to use case expressions:
517 clunky env var1 var1 = case lookup env var1 of
519 Just val1 -> case lookup env var2 of
521 Just val2 -> val1 + val2
527 This is a bit shorter, but hardly better. Of course, we can rewrite any set
528 of pattern-matching, guarded equations as case expressions; that is
529 precisely what the compiler does when compiling equations! The reason that
530 Haskell provides guarded equations is because they allow us to write down
531 the cases we want to consider, one at a time, independently of each other.
532 This structure is hidden in the case version. Two of the right-hand sides
533 are really the same (<function>fail</function>), and the whole expression
534 tends to become more and more indented.
538 Here is how I would write clunky:
543 | Just val1 <- lookup env var1
544 , Just val2 <- lookup env var2
546 ...other equations for clunky...
550 The semantics should be clear enough. The qualifers are matched in order.
551 For a <literal><-</literal> qualifier, which I call a pattern guard, the
552 right hand side is evaluated and matched against the pattern on the left.
553 If the match fails then the whole guard fails and the next equation is
554 tried. If it succeeds, then the appropriate binding takes place, and the
555 next qualifier is matched, in the augmented environment. Unlike list
556 comprehensions, however, the type of the expression to the right of the
557 <literal><-</literal> is the same as the type of the pattern to its
558 left. The bindings introduced by pattern guards scope over all the
559 remaining guard qualifiers, and over the right hand side of the equation.
563 Just as with list comprehensions, boolean expressions can be freely mixed
564 with among the pattern guards. For example:
575 Haskell's current guards therefore emerge as a special case, in which the
576 qualifier list has just one element, a boolean expression.
580 <sect1 id="parallel-list-comprehensions">
581 <title>Parallel List Comprehensions</title>
582 <indexterm><primary>list comprehensions</primary><secondary>parallel</secondary>
584 <indexterm><primary>parallel list comprehensions</primary>
587 <para>Parallel list comprehensions are a natural extension to list
588 comprehensions. List comprehensions can be thought of as a nice
589 syntax for writing maps and filters. Parallel comprehensions
590 extend this to include the zipWith family.</para>
592 <para>A parallel list comprehension has multiple independent
593 branches of qualifier lists, each separated by a `|' symbol. For
594 example, the following zips together two lists:</para>
597 [ (x, y) | x <- xs | y <- ys ]
600 <para>The behavior of parallel list comprehensions follows that of
601 zip, in that the resulting list will have the same length as the
602 shortest branch.</para>
604 <para>We can define parallel list comprehensions by translation to
605 regular comprehensions. Here's the basic idea:</para>
607 <para>Given a parallel comprehension of the form: </para>
610 [ e | p1 <- e11, p2 <- e12, ...
611 | q1 <- e21, q2 <- e22, ...
616 <para>This will be translated to: </para>
619 [ e | ((p1,p2), (q1,q2), ...) <- zipN [(p1,p2) | p1 <- e11, p2 <- e12, ...]
620 [(q1,q2) | q1 <- e21, q2 <- e22, ...]
625 <para>where `zipN' is the appropriate zip for the given number of
630 <sect1 id="multi-param-type-classes">
631 <title>Multi-parameter type classes
635 This section documents GHC's implementation of multi-parameter type
636 classes. There's lots of background in the paper <ULink
637 URL="http://research.microsoft.com/~simonpj/multi.ps.gz" >Type
638 classes: exploring the design space</ULink > (Simon Peyton Jones, Mark
643 I'd like to thank people who reported shorcomings in the GHC 3.02
644 implementation. Our default decisions were all conservative ones, and
645 the experience of these heroic pioneers has given useful concrete
646 examples to support several generalisations. (These appear below as
647 design choices not implemented in 3.02.)
651 I've discussed these notes with Mark Jones, and I believe that Hugs
652 will migrate towards the same design choices as I outline here.
653 Thanks to him, and to many others who have offered very useful
661 There are the following restrictions on the form of a qualified
668 forall tv1..tvn (c1, ...,cn) => type
674 (Here, I write the "foralls" explicitly, although the Haskell source
675 language omits them; in Haskell 1.4, all the free type variables of an
676 explicit source-language type signature are universally quantified,
677 except for the class type variables in a class declaration. However,
678 in GHC, you can give the foralls if you want. See <xref LinkEnd="universal-quantification">).
687 <emphasis>Each universally quantified type variable
688 <literal>tvi</literal> must be mentioned (i.e. appear free) in <literal>type</literal></emphasis>.
690 The reason for this is that a value with a type that does not obey
691 this restriction could not be used without introducing
692 ambiguity. Here, for example, is an illegal type:
696 forall a. Eq a => Int
700 When a value with this type was used, the constraint <literal>Eq tv</literal>
701 would be introduced where <literal>tv</literal> is a fresh type variable, and
702 (in the dictionary-translation implementation) the value would be
703 applied to a dictionary for <literal>Eq tv</literal>. The difficulty is that we
704 can never know which instance of <literal>Eq</literal> to use because we never
705 get any more information about <literal>tv</literal>.
712 <emphasis>Every constraint <literal>ci</literal> must mention at least one of the
713 universally quantified type variables <literal>tvi</literal></emphasis>.
715 For example, this type is OK because <literal>C a b</literal> mentions the
716 universally quantified type variable <literal>b</literal>:
720 forall a. C a b => burble
724 The next type is illegal because the constraint <literal>Eq b</literal> does not
725 mention <literal>a</literal>:
729 forall a. Eq b => burble
733 The reason for this restriction is milder than the other one. The
734 excluded types are never useful or necessary (because the offending
735 context doesn't need to be witnessed at this point; it can be floated
736 out). Furthermore, floating them out increases sharing. Lastly,
737 excluding them is a conservative choice; it leaves a patch of
738 territory free in case we need it later.
748 These restrictions apply to all types, whether declared in a type signature
753 Unlike Haskell 1.4, constraints in types do <emphasis>not</emphasis> have to be of
754 the form <emphasis>(class type-variables)</emphasis>. Thus, these type signatures
761 f :: Eq (m a) => [m a] -> [m a]
768 This choice recovers principal types, a property that Haskell 1.4 does not have.
774 <title>Class declarations</title>
782 <emphasis>Multi-parameter type classes are permitted</emphasis>. For example:
786 class Collection c a where
787 union :: c a -> c a -> c a
798 <emphasis>The class hierarchy must be acyclic</emphasis>. However, the definition
799 of "acyclic" involves only the superclass relationships. For example,
805 op :: D b => a -> b -> b
808 class C a => D a where { ... }
812 Here, <literal>C</literal> is a superclass of <literal>D</literal>, but it's OK for a
813 class operation <literal>op</literal> of <literal>C</literal> to mention <literal>D</literal>. (It
814 would not be OK for <literal>D</literal> to be a superclass of <literal>C</literal>.)
821 <emphasis>There are no restrictions on the context in a class declaration
822 (which introduces superclasses), except that the class hierarchy must
823 be acyclic</emphasis>. So these class declarations are OK:
827 class Functor (m k) => FiniteMap m k where
830 class (Monad m, Monad (t m)) => Transform t m where
831 lift :: m a -> (t m) a
840 <emphasis>In the signature of a class operation, every constraint
841 must mention at least one type variable that is not a class type
848 class Collection c a where
849 mapC :: Collection c b => (a->b) -> c a -> c b
853 is OK because the constraint <literal>(Collection a b)</literal> mentions
854 <literal>b</literal>, even though it also mentions the class variable
855 <literal>a</literal>. On the other hand:
860 op :: Eq a => (a,b) -> (a,b)
864 is not OK because the constraint <literal>(Eq a)</literal> mentions on the class
865 type variable <literal>a</literal>, but not <literal>b</literal>. However, any such
866 example is easily fixed by moving the offending context up to the
871 class Eq a => C a where
876 A yet more relaxed rule would allow the context of a class-op signature
877 to mention only class type variables. However, that conflicts with
878 Rule 1(b) for types above.
885 <emphasis>The type of each class operation must mention <emphasis>all</emphasis> of
886 the class type variables</emphasis>. For example:
892 insert :: s -> a -> s
896 is not OK, because the type of <literal>empty</literal> doesn't mention
897 <literal>a</literal>. This rule is a consequence of Rule 1(a), above, for
898 types, and has the same motivation.
900 Sometimes, offending class declarations exhibit misunderstandings. For
901 example, <literal>Coll</literal> might be rewritten
907 insert :: s a -> a -> s a
911 which makes the connection between the type of a collection of
912 <literal>a</literal>'s (namely <literal>(s a)</literal>) and the element type <literal>a</literal>.
913 Occasionally this really doesn't work, in which case you can split the
921 class CollE s => Coll s a where
922 insert :: s -> a -> s
935 <sect2 id="instance-decls">
936 <title>Instance declarations</title>
944 <emphasis>Instance declarations may not overlap</emphasis>. The two instance
949 instance context1 => C type1 where ...
950 instance context2 => C type2 where ...
954 "overlap" if <literal>type1</literal> and <literal>type2</literal> unify
956 However, if you give the command line option
957 <option>-fallow-overlapping-instances</option><indexterm><primary>-fallow-overlapping-instances
958 option</primary></indexterm> then two overlapping instance declarations are permitted
966 EITHER <literal>type1</literal> and <literal>type2</literal> do not unify
972 OR <literal>type2</literal> is a substitution instance of <literal>type1</literal>
973 (but not identical to <literal>type1</literal>)
986 Notice that these rules
993 make it clear which instance decl to use
994 (pick the most specific one that matches)
1001 do not mention the contexts <literal>context1</literal>, <literal>context2</literal>
1002 Reason: you can pick which instance decl
1003 "matches" based on the type.
1010 Regrettably, GHC doesn't guarantee to detect overlapping instance
1011 declarations if they appear in different modules. GHC can "see" the
1012 instance declarations in the transitive closure of all the modules
1013 imported by the one being compiled, so it can "see" all instance decls
1014 when it is compiling <literal>Main</literal>. However, it currently chooses not
1015 to look at ones that can't possibly be of use in the module currently
1016 being compiled, in the interests of efficiency. (Perhaps we should
1017 change that decision, at least for <literal>Main</literal>.)
1024 <emphasis>There are no restrictions on the type in an instance
1025 <emphasis>head</emphasis>, except that at least one must not be a type variable</emphasis>.
1026 The instance "head" is the bit after the "=>" in an instance decl. For
1027 example, these are OK:
1031 instance C Int a where ...
1033 instance D (Int, Int) where ...
1035 instance E [[a]] where ...
1039 Note that instance heads <emphasis>may</emphasis> contain repeated type variables.
1040 For example, this is OK:
1044 instance Stateful (ST s) (MutVar s) where ...
1048 The "at least one not a type variable" restriction is to ensure that
1049 context reduction terminates: each reduction step removes one type
1050 constructor. For example, the following would make the type checker
1051 loop if it wasn't excluded:
1055 instance C a => C a where ...
1059 There are two situations in which the rule is a bit of a pain. First,
1060 if one allows overlapping instance declarations then it's quite
1061 convenient to have a "default instance" declaration that applies if
1062 something more specific does not:
1071 Second, sometimes you might want to use the following to get the
1072 effect of a "class synonym":
1076 class (C1 a, C2 a, C3 a) => C a where { }
1078 instance (C1 a, C2 a, C3 a) => C a where { }
1082 This allows you to write shorter signatures:
1094 f :: (C1 a, C2 a, C3 a) => ...
1098 I'm on the lookout for a simple rule that preserves decidability while
1099 allowing these idioms. The experimental flag
1100 <option>-fallow-undecidable-instances</option><indexterm><primary>-fallow-undecidable-instances
1101 option</primary></indexterm> lifts this restriction, allowing all the types in an
1102 instance head to be type variables.
1109 <emphasis>Unlike Haskell 1.4, instance heads may use type
1110 synonyms</emphasis>. As always, using a type synonym is just shorthand for
1111 writing the RHS of the type synonym definition. For example:
1115 type Point = (Int,Int)
1116 instance C Point where ...
1117 instance C [Point] where ...
1121 is legal. However, if you added
1125 instance C (Int,Int) where ...
1129 as well, then the compiler will complain about the overlapping
1130 (actually, identical) instance declarations. As always, type synonyms
1131 must be fully applied. You cannot, for example, write:
1136 instance Monad P where ...
1140 This design decision is independent of all the others, and easily
1141 reversed, but it makes sense to me.
1148 <emphasis>The types in an instance-declaration <emphasis>context</emphasis> must all
1149 be type variables</emphasis>. Thus
1153 instance C a b => Eq (a,b) where ...
1161 instance C Int b => Foo b where ...
1165 is not OK. Again, the intent here is to make sure that context
1166 reduction terminates.
1168 Voluminous correspondence on the Haskell mailing list has convinced me
1169 that it's worth experimenting with a more liberal rule. If you use
1170 the flag <option>-fallow-undecidable-instances</option> can use arbitrary
1171 types in an instance context. Termination is ensured by having a
1172 fixed-depth recursion stack. If you exceed the stack depth you get a
1173 sort of backtrace, and the opportunity to increase the stack depth
1174 with <option>-fcontext-stack</option><emphasis>N</emphasis>.
1187 <sect1 id="implicit-parameters">
1188 <title>Implicit parameters
1191 <para> Implicit paramters are implemented as described in
1192 "Implicit parameters: dynamic scoping with static types",
1193 J Lewis, MB Shields, E Meijer, J Launchbury,
1194 27th ACM Symposium on Principles of Programming Languages (POPL'00),
1199 There should be more documentation, but there isn't (yet). Yell if you need it.
1203 <para> You can't have an implicit parameter in the context of a class or instance
1204 declaration. For example, both these declarations are illegal:
1206 class (?x::Int) => C a where ...
1207 instance (?x::a) => Foo [a] where ...
1209 Reason: exactly which implicit parameter you pick up depends on exactly where
1210 you invoke a function. But the ``invocation'' of instance declarations is done
1211 behind the scenes by the compiler, so it's hard to figure out exactly where it is done.
1212 Easiest thing is to outlaw the offending types.</para>
1219 <sect1 id="linear-implicit-parameters">
1220 <title>Linear implicit parameters
1223 Linear implicit parameters are an idea developed by Koen Claessen,
1224 Mark Shields, and Simon PJ. They address the long-standing
1225 problem that monads seem over-kill for certain sorts of problem, notably:
1228 <listitem> <para> distributing a supply of unique names </para> </listitem>
1229 <listitem> <para> distributing a suppply of random numbers </para> </listitem>
1230 <listitem> <para> distributing an oracle (as in QuickCheck) </para> </listitem>
1234 Linear implicit parameters are just like ordinary implicit parameters,
1235 except that they are "linear" -- that is, they cannot be copied, and
1236 must be explicitly "split" instead. Linear implicit parameters are
1237 written '<literal>%x</literal>' instead of '<literal>?x</literal>'.
1238 (The '/' in the '%' suggests the split!)
1243 data NameSupply = ...
1245 splitNS :: NameSupply -> (NameSupply, NameSupply)
1246 newName :: NameSupply -> Name
1248 instance PrelSplit.Splittable NameSupply where
1252 f :: (%ns :: NameSupply) => Env -> Expr -> Expr
1253 f env (Lam x e) = Lam x' (f env e)
1256 env' = extend env x x'
1257 ...more equations for f...
1259 Notice that the implicit parameter %ns is consumed
1261 <listitem> <para> once by the call to <literal>newName</literal> </para> </listitem>
1262 <listitem> <para> once by the recursive call to <literal>f</literal> </para></listitem>
1266 So the translation done by the type checker makes
1267 the parameter explicit:
1269 f :: NameSupply -> Env -> Expr -> Expr
1270 f ns env (Lam x e) = Lam x' (f ns1 env e)
1272 (ns1,ns2) = splitNS ns
1274 env = extend env x x'
1276 Notice the call to 'split' introduced by the type checker.
1277 How did it know to use 'splitNS'? Because what it really did
1278 was to introduce a call to the overloaded function 'split',
1281 class Splittable a where
1284 The instance for <literal>Splittable NameSupply</literal> tells GHC how to implement
1285 split for name supplies. But we can simply write
1291 g :: (Splittable a, %ns :: a) => b -> (b,a,a)
1293 The <literal>Splittable</literal> class is built into GHC. It's defined in <literal>PrelSplit</literal>,
1294 and exported by <literal>GlaExts</literal>.
1299 <listitem> <para> '<literal>?x</literal>' and '<literal>%x</literal>'
1300 are entirely distinct implicit parameters: you
1301 can use them together and they won't intefere with each other. </para>
1304 <listitem> <para> You can bind linear implicit parameters in 'with' clauses. </para> </listitem>
1306 <listitem> <para>You cannot have implicit parameters (whether linear or not)
1307 in the context of a class or instance declaration. </para></listitem>
1311 <sect2><title>Warnings</title>
1314 The monomorphism restriction is even more important than usual.
1315 Consider the example above:
1317 f :: (%ns :: NameSupply) => Env -> Expr -> Expr
1318 f env (Lam x e) = Lam x' (f env e)
1321 env' = extend env x x'
1323 If we replaced the two occurrences of x' by (newName %ns), which is
1324 usually a harmless thing to do, we get:
1326 f :: (%ns :: NameSupply) => Env -> Expr -> Expr
1327 f env (Lam x e) = Lam (newName %ns) (f env e)
1329 env' = extend env x (newName %ns)
1331 But now the name supply is consumed in <emphasis>three</emphasis> places
1332 (the two calls to newName,and the recursive call to f), so
1333 the result is utterly different. Urk! We don't even have
1337 Well, this is an experimental change. With implicit
1338 parameters we have already lost beta reduction anyway, and
1339 (as John Launchbury puts it) we can't sensibly reason about
1340 Haskell programs without knowing their typing.
1347 <sect1 id="functional-dependencies">
1348 <title>Functional dependencies
1351 <para> Functional dependencies are implemented as described by Mark Jones
1352 in "Type Classes with Functional Dependencies", Mark P. Jones,
1353 In Proceedings of the 9th European Symposium on Programming,
1354 ESOP 2000, Berlin, Germany, March 2000, Springer-Verlag LNCS 1782.
1358 There should be more documentation, but there isn't (yet). Yell if you need it.
1363 <sect1 id="universal-quantification">
1364 <title>Explicit universal quantification
1368 GHC's type system supports explicit universal quantification in
1369 constructor fields and function arguments. This is useful for things
1370 like defining <literal>runST</literal> from the state-thread world.
1371 GHC's syntax for this now agrees with Hugs's, namely:
1377 forall a b. (Ord a, Eq b) => a -> b -> a
1383 The context is, of course, optional. You can't use <literal>forall</literal> as
1384 a type variable any more!
1388 Haskell type signatures are implicitly quantified. The <literal>forall</literal>
1389 allows us to say exactly what this means. For example:
1407 g :: forall b. (b -> b)
1413 The two are treated identically.
1417 <title>Universally-quantified data type fields
1421 In a <literal>data</literal> or <literal>newtype</literal> declaration one can quantify
1422 the types of the constructor arguments. Here are several examples:
1428 data T a = T1 (forall b. b -> b -> b) a
1430 data MonadT m = MkMonad { return :: forall a. a -> m a,
1431 bind :: forall a b. m a -> (a -> m b) -> m b
1434 newtype Swizzle = MkSwizzle (Ord a => [a] -> [a])
1440 The constructors now have so-called <emphasis>rank 2</emphasis> polymorphic
1441 types, in which there is a for-all in the argument types.:
1447 T1 :: forall a. (forall b. b -> b -> b) -> a -> T a
1448 MkMonad :: forall m. (forall a. a -> m a)
1449 -> (forall a b. m a -> (a -> m b) -> m b)
1451 MkSwizzle :: (Ord a => [a] -> [a]) -> Swizzle
1457 Notice that you don't need to use a <literal>forall</literal> if there's an
1458 explicit context. For example in the first argument of the
1459 constructor <function>MkSwizzle</function>, an implicit "<literal>forall a.</literal>" is
1460 prefixed to the argument type. The implicit <literal>forall</literal>
1461 quantifies all type variables that are not already in scope, and are
1462 mentioned in the type quantified over.
1466 As for type signatures, implicit quantification happens for non-overloaded
1467 types too. So if you write this:
1470 data T a = MkT (Either a b) (b -> b)
1473 it's just as if you had written this:
1476 data T a = MkT (forall b. Either a b) (forall b. b -> b)
1479 That is, since the type variable <literal>b</literal> isn't in scope, it's
1480 implicitly universally quantified. (Arguably, it would be better
1481 to <emphasis>require</emphasis> explicit quantification on constructor arguments
1482 where that is what is wanted. Feedback welcomed.)
1488 <title>Construction </title>
1491 You construct values of types <literal>T1, MonadT, Swizzle</literal> by applying
1492 the constructor to suitable values, just as usual. For example,
1498 (T1 (\xy->x) 3) :: T Int
1500 (MkSwizzle sort) :: Swizzle
1501 (MkSwizzle reverse) :: Swizzle
1508 MkMonad r b) :: MonadT Maybe
1514 The type of the argument can, as usual, be more general than the type
1515 required, as <literal>(MkSwizzle reverse)</literal> shows. (<function>reverse</function>
1516 does not need the <literal>Ord</literal> constraint.)
1522 <title>Pattern matching</title>
1525 When you use pattern matching, the bound variables may now have
1526 polymorphic types. For example:
1532 f :: T a -> a -> (a, Char)
1533 f (T1 f k) x = (f k x, f 'c' 'd')
1535 g :: (Ord a, Ord b) => Swizzle -> [a] -> (a -> b) -> [b]
1536 g (MkSwizzle s) xs f = s (map f (s xs))
1538 h :: MonadT m -> [m a] -> m [a]
1539 h m [] = return m []
1540 h m (x:xs) = bind m x $ \y ->
1541 bind m (h m xs) $ \ys ->
1548 In the function <function>h</function> we use the record selectors <literal>return</literal>
1549 and <literal>bind</literal> to extract the polymorphic bind and return functions
1550 from the <literal>MonadT</literal> data structure, rather than using pattern
1555 You cannot pattern-match against an argument that is polymorphic.
1559 newtype TIM s a = TIM (ST s (Maybe a))
1561 runTIM :: (forall s. TIM s a) -> Maybe a
1562 runTIM (TIM m) = runST m
1568 Here the pattern-match fails, because you can't pattern-match against
1569 an argument of type <literal>(forall s. TIM s a)</literal>. Instead you
1570 must bind the variable and pattern match in the right hand side:
1573 runTIM :: (forall s. TIM s a) -> Maybe a
1574 runTIM tm = case tm of { TIM m -> runST m }
1577 The <literal>tm</literal> on the right hand side is (invisibly) instantiated, like
1578 any polymorphic value at its occurrence site, and now you can pattern-match
1585 <title>The partial-application restriction</title>
1588 There is really only one way in which data structures with polymorphic
1589 components might surprise you: you must not partially apply them.
1590 For example, this is illegal:
1596 map MkSwizzle [sort, reverse]
1602 The restriction is this: <emphasis>every subexpression of the program must
1603 have a type that has no for-alls, except that in a function
1604 application (f e1…en) the partial applications are not subject to
1605 this rule</emphasis>. The restriction makes type inference feasible.
1609 In the illegal example, the sub-expression <literal>MkSwizzle</literal> has the
1610 polymorphic type <literal>(Ord b => [b] -> [b]) -> Swizzle</literal> and is not
1611 a sub-expression of an enclosing application. On the other hand, this
1618 map (T1 (\a b -> a)) [1,2,3]
1624 even though it involves a partial application of <function>T1</function>, because
1625 the sub-expression <literal>T1 (\a b -> a)</literal> has type <literal>Int -> T
1632 <title>Type signatures
1636 Once you have data constructors with universally-quantified fields, or
1637 constants such as <Constant>runST</Constant> that have rank-2 types, it isn't long
1638 before you discover that you need more! Consider:
1644 mkTs f x y = [T1 f x, T1 f y]
1650 <function>mkTs</function> is a fuction that constructs some values of type
1651 <literal>T</literal>, using some pieces passed to it. The trouble is that since
1652 <literal>f</literal> is a function argument, Haskell assumes that it is
1653 monomorphic, so we'll get a type error when applying <function>T1</function> to
1654 it. This is a rather silly example, but the problem really bites in
1655 practice. Lots of people trip over the fact that you can't make
1656 "wrappers functions" for <Constant>runST</Constant> for exactly the same reason.
1657 In short, it is impossible to build abstractions around functions with
1662 The solution is fairly clear. We provide the ability to give a rank-2
1663 type signature for <emphasis>ordinary</emphasis> functions (not only data
1664 constructors), thus:
1670 mkTs :: (forall b. b -> b -> b) -> a -> [T a]
1671 mkTs f x y = [T1 f x, T1 f y]
1677 This type signature tells the compiler to attribute <literal>f</literal> with
1678 the polymorphic type <literal>(forall b. b -> b -> b)</literal> when type
1679 checking the body of <function>mkTs</function>, so now the application of
1680 <function>T1</function> is fine.
1684 There are two restrictions:
1693 You can only define a rank 2 type, specified by the following
1698 rank2type ::= [forall tyvars .] [context =>] funty
1699 funty ::= ([forall tyvars .] [context =>] ty) -> funty
1701 ty ::= ...current Haskell monotype syntax...
1705 Informally, the universal quantification must all be right at the beginning,
1706 or at the top level of a function argument.
1713 There is a restriction on the definition of a function whose
1714 type signature is a rank-2 type: the polymorphic arguments must be
1715 matched on the left hand side of the "<literal>=</literal>" sign. You can't
1716 define <function>mkTs</function> like this:
1720 mkTs :: (forall b. b -> b -> b) -> a -> [T a]
1721 mkTs = \ f x y -> [T1 f x, T1 f y]
1726 The same partial-application rule applies to ordinary functions with
1727 rank-2 types as applied to data constructors.
1740 <title>Type synonyms and hoisting
1744 GHC also allows you to write a <literal>forall</literal> in a type synonym, thus:
1746 type Discard a = forall b. a -> b -> a
1751 However, it is often convenient to use these sort of synonyms at the right hand
1752 end of an arrow, thus:
1754 type Discard a = forall b. a -> b -> a
1756 g :: Int -> Discard Int
1759 Simply expanding the type synonym would give
1761 g :: Int -> (forall b. Int -> b -> Int)
1763 but GHC "hoists" the <literal>forall</literal> to give the isomorphic type
1765 g :: forall b. Int -> Int -> b -> Int
1767 In general, the rule is this: <emphasis>to determine the type specified by any explicit
1768 user-written type (e.g. in a type signature), GHC expands type synonyms and then repeatedly
1769 performs the transformation:</emphasis>
1771 <emphasis>type1</emphasis> -> forall a. <emphasis>type2</emphasis>
1773 forall a. <emphasis>type1</emphasis> -> <emphasis>type2</emphasis>
1775 (In fact, GHC tries to retain as much synonym information as possible for use in
1776 error messages, but that is a usability issue.) This rule applies, of course, whether
1777 or not the <literal>forall</literal> comes from a synonym. For example, here is another
1778 valid way to write <literal>g</literal>'s type signature:
1780 g :: Int -> Int -> forall b. b -> Int
1787 <sect1 id="existential-quantification">
1788 <title>Existentially quantified data constructors
1792 The idea of using existential quantification in data type declarations
1793 was suggested by Laufer (I believe, thought doubtless someone will
1794 correct me), and implemented in Hope+. It's been in Lennart
1795 Augustsson's <Command>hbc</Command> Haskell compiler for several years, and
1796 proved very useful. Here's the idea. Consider the declaration:
1802 data Foo = forall a. MkFoo a (a -> Bool)
1809 The data type <literal>Foo</literal> has two constructors with types:
1815 MkFoo :: forall a. a -> (a -> Bool) -> Foo
1822 Notice that the type variable <literal>a</literal> in the type of <function>MkFoo</function>
1823 does not appear in the data type itself, which is plain <literal>Foo</literal>.
1824 For example, the following expression is fine:
1830 [MkFoo 3 even, MkFoo 'c' isUpper] :: [Foo]
1836 Here, <literal>(MkFoo 3 even)</literal> packages an integer with a function
1837 <function>even</function> that maps an integer to <literal>Bool</literal>; and <function>MkFoo 'c'
1838 isUpper</function> packages a character with a compatible function. These
1839 two things are each of type <literal>Foo</literal> and can be put in a list.
1843 What can we do with a value of type <literal>Foo</literal>?. In particular,
1844 what happens when we pattern-match on <function>MkFoo</function>?
1850 f (MkFoo val fn) = ???
1856 Since all we know about <literal>val</literal> and <function>fn</function> is that they
1857 are compatible, the only (useful) thing we can do with them is to
1858 apply <function>fn</function> to <literal>val</literal> to get a boolean. For example:
1865 f (MkFoo val fn) = fn val
1871 What this allows us to do is to package heterogenous values
1872 together with a bunch of functions that manipulate them, and then treat
1873 that collection of packages in a uniform manner. You can express
1874 quite a bit of object-oriented-like programming this way.
1877 <sect2 id="existential">
1878 <title>Why existential?
1882 What has this to do with <emphasis>existential</emphasis> quantification?
1883 Simply that <function>MkFoo</function> has the (nearly) isomorphic type
1889 MkFoo :: (exists a . (a, a -> Bool)) -> Foo
1895 But Haskell programmers can safely think of the ordinary
1896 <emphasis>universally</emphasis> quantified type given above, thereby avoiding
1897 adding a new existential quantification construct.
1903 <title>Type classes</title>
1906 An easy extension (implemented in <Command>hbc</Command>) is to allow
1907 arbitrary contexts before the constructor. For example:
1913 data Baz = forall a. Eq a => Baz1 a a
1914 | forall b. Show b => Baz2 b (b -> b)
1920 The two constructors have the types you'd expect:
1926 Baz1 :: forall a. Eq a => a -> a -> Baz
1927 Baz2 :: forall b. Show b => b -> (b -> b) -> Baz
1933 But when pattern matching on <function>Baz1</function> the matched values can be compared
1934 for equality, and when pattern matching on <function>Baz2</function> the first matched
1935 value can be converted to a string (as well as applying the function to it).
1936 So this program is legal:
1943 f (Baz1 p q) | p == q = "Yes"
1945 f (Baz1 v fn) = show (fn v)
1951 Operationally, in a dictionary-passing implementation, the
1952 constructors <function>Baz1</function> and <function>Baz2</function> must store the
1953 dictionaries for <literal>Eq</literal> and <literal>Show</literal> respectively, and
1954 extract it on pattern matching.
1958 Notice the way that the syntax fits smoothly with that used for
1959 universal quantification earlier.
1965 <title>Restrictions</title>
1968 There are several restrictions on the ways in which existentially-quantified
1969 constructors can be use.
1978 When pattern matching, each pattern match introduces a new,
1979 distinct, type for each existential type variable. These types cannot
1980 be unified with any other type, nor can they escape from the scope of
1981 the pattern match. For example, these fragments are incorrect:
1989 Here, the type bound by <function>MkFoo</function> "escapes", because <literal>a</literal>
1990 is the result of <function>f1</function>. One way to see why this is wrong is to
1991 ask what type <function>f1</function> has:
1995 f1 :: Foo -> a -- Weird!
1999 What is this "<literal>a</literal>" in the result type? Clearly we don't mean
2004 f1 :: forall a. Foo -> a -- Wrong!
2008 The original program is just plain wrong. Here's another sort of error
2012 f2 (Baz1 a b) (Baz1 p q) = a==q
2016 It's ok to say <literal>a==b</literal> or <literal>p==q</literal>, but
2017 <literal>a==q</literal> is wrong because it equates the two distinct types arising
2018 from the two <function>Baz1</function> constructors.
2026 You can't pattern-match on an existentially quantified
2027 constructor in a <literal>let</literal> or <literal>where</literal> group of
2028 bindings. So this is illegal:
2032 f3 x = a==b where { Baz1 a b = x }
2036 You can only pattern-match
2037 on an existentially-quantified constructor in a <literal>case</literal> expression or
2038 in the patterns of a function definition.
2040 The reason for this restriction is really an implementation one.
2041 Type-checking binding groups is already a nightmare without
2042 existentials complicating the picture. Also an existential pattern
2043 binding at the top level of a module doesn't make sense, because it's
2044 not clear how to prevent the existentially-quantified type "escaping".
2045 So for now, there's a simple-to-state restriction. We'll see how
2053 You can't use existential quantification for <literal>newtype</literal>
2054 declarations. So this is illegal:
2058 newtype T = forall a. Ord a => MkT a
2062 Reason: a value of type <literal>T</literal> must be represented as a pair
2063 of a dictionary for <literal>Ord t</literal> and a value of type <literal>t</literal>.
2064 That contradicts the idea that <literal>newtype</literal> should have no
2065 concrete representation. You can get just the same efficiency and effect
2066 by using <literal>data</literal> instead of <literal>newtype</literal>. If there is no
2067 overloading involved, then there is more of a case for allowing
2068 an existentially-quantified <literal>newtype</literal>, because the <literal>data</literal>
2069 because the <literal>data</literal> version does carry an implementation cost,
2070 but single-field existentially quantified constructors aren't much
2071 use. So the simple restriction (no existential stuff on <literal>newtype</literal>)
2072 stands, unless there are convincing reasons to change it.
2080 You can't use <literal>deriving</literal> to define instances of a
2081 data type with existentially quantified data constructors.
2083 Reason: in most cases it would not make sense. For example:#
2086 data T = forall a. MkT [a] deriving( Eq )
2089 To derive <literal>Eq</literal> in the standard way we would need to have equality
2090 between the single component of two <function>MkT</function> constructors:
2094 (MkT a) == (MkT b) = ???
2097 But <VarName>a</VarName> and <VarName>b</VarName> have distinct types, and so can't be compared.
2098 It's just about possible to imagine examples in which the derived instance
2099 would make sense, but it seems altogether simpler simply to prohibit such
2100 declarations. Define your own instances!
2112 <sect1 id="sec-assertions">
2114 <indexterm><primary>Assertions</primary></indexterm>
2118 If you want to make use of assertions in your standard Haskell code, you
2119 could define a function like the following:
2125 assert :: Bool -> a -> a
2126 assert False x = error "assertion failed!"
2133 which works, but gives you back a less than useful error message --
2134 an assertion failed, but which and where?
2138 One way out is to define an extended <function>assert</function> function which also
2139 takes a descriptive string to include in the error message and
2140 perhaps combine this with the use of a pre-processor which inserts
2141 the source location where <function>assert</function> was used.
2145 Ghc offers a helping hand here, doing all of this for you. For every
2146 use of <function>assert</function> in the user's source:
2152 kelvinToC :: Double -> Double
2153 kelvinToC k = assert (k >= 0.0) (k+273.15)
2159 Ghc will rewrite this to also include the source location where the
2166 assert pred val ==> assertError "Main.hs|15" pred val
2172 The rewrite is only performed by the compiler when it spots
2173 applications of <function>Exception.assert</function>, so you can still define and
2174 use your own versions of <function>assert</function>, should you so wish. If not,
2175 import <literal>Exception</literal> to make use <function>assert</function> in your code.
2179 To have the compiler ignore uses of assert, use the compiler option
2180 <option>-fignore-asserts</option>. <indexterm><primary>-fignore-asserts option</primary></indexterm> That is,
2181 expressions of the form <literal>assert pred e</literal> will be rewritten to <literal>e</literal>.
2185 Assertion failures can be caught, see the documentation for the
2186 <literal>Exception</literal> library (<xref linkend="sec-Exception">)
2192 <sect1 id="scoped-type-variables">
2193 <title>Scoped Type Variables
2197 A <emphasis>pattern type signature</emphasis> can introduce a <emphasis>scoped type
2198 variable</emphasis>. For example
2204 f (xs::[a]) = ys ++ ys
2213 The pattern <literal>(xs::[a])</literal> includes a type signature for <VarName>xs</VarName>.
2214 This brings the type variable <literal>a</literal> into scope; it scopes over
2215 all the patterns and right hand sides for this equation for <function>f</function>.
2216 In particular, it is in scope at the type signature for <VarName>y</VarName>.
2220 Pattern type signatures are completely orthogonal to ordinary, separate
2221 type signatures. The two can be used independently or together.
2222 At ordinary type signatures, such as that for <VarName>ys</VarName>, any type variables
2223 mentioned in the type signature <emphasis>that are not in scope</emphasis> are
2224 implicitly universally quantified. (If there are no type variables in
2225 scope, all type variables mentioned in the signature are universally
2226 quantified, which is just as in Haskell 98.) In this case, since <VarName>a</VarName>
2227 is in scope, it is not universally quantified, so the type of <VarName>ys</VarName> is
2228 the same as that of <VarName>xs</VarName>. In Haskell 98 it is not possible to declare
2229 a type for <VarName>ys</VarName>; a major benefit of scoped type variables is that
2230 it becomes possible to do so.
2234 Scoped type variables are implemented in both GHC and Hugs. Where the
2235 implementations differ from the specification below, those differences
2240 So much for the basic idea. Here are the details.
2244 <title>What a pattern type signature means</title>
2246 A type variable brought into scope by a pattern type signature is simply
2247 the name for a type. The restriction they express is that all occurrences
2248 of the same name mean the same type. For example:
2250 f :: [Int] -> Int -> Int
2251 f (xs::[a]) (y::a) = (head xs + y) :: a
2253 The pattern type signatures on the left hand side of
2254 <literal>f</literal> express the fact that <literal>xs</literal>
2255 must be a list of things of some type <literal>a</literal>; and that <literal>y</literal>
2256 must have this same type. The type signature on the expression <literal>(head xs)</literal>
2257 specifies that this expression must have the same type <literal>a</literal>.
2258 <emphasis>There is no requirement that the type named by "<literal>a</literal>" is
2259 in fact a type variable</emphasis>. Indeed, in this case, the type named by "<literal>a</literal>" is
2260 <literal>Int</literal>. (This is a slight liberalisation from the original rather complex
2261 rules, which specified that a pattern-bound type variable should be universally quantified.)
2262 For example, all of these are legal:</para>
2265 t (x::a) (y::a) = x+y*2
2267 f (x::a) (y::b) = [x,y] -- a unifies with b
2269 g (x::a) = x + 1::Int -- a unifies with Int
2271 h x = let k (y::a) = [x,y] -- a is free in the
2272 in k x -- environment
2274 k (x::a) True = ... -- a unifies with Int
2275 k (x::Int) False = ...
2278 w (x::a) = x -- a unifies with [b]
2284 <title>Scope and implicit quantification</title>
2292 All the type variables mentioned in a pattern,
2293 that are not already in scope,
2294 are brought into scope by the pattern. We describe this set as
2295 the <emphasis>type variables bound by the pattern</emphasis>.
2298 f (x::a) = let g (y::(a,b)) = fst y
2302 The pattern <literal>(x::a)</literal> brings the type variable
2303 <literal>a</literal> into scope, as well as the term
2304 variable <literal>x</literal>. The pattern <literal>(y::(a,b))</literal>
2305 contains an occurrence of the already-in-scope type variable <literal>a</literal>,
2306 and brings into scope the type variable <literal>b</literal>.
2312 The type variable(s) bound by the pattern have the same scope
2313 as the term variable(s) bound by the pattern. For example:
2316 f (x::a) = <...rhs of f...>
2317 (p::b, q::b) = (1,2)
2318 in <...body of let...>
2320 Here, the type variable <literal>a</literal> scopes over the right hand side of <literal>f</literal>,
2321 just like <literal>x</literal> does; while the type variable <literal>b</literal> scopes over the
2322 body of the <literal>let</literal>, and all the other definitions in the <literal>let</literal>,
2323 just like <literal>p</literal> and <literal>q</literal> do.
2324 Indeed, the newly bound type variables also scope over any ordinary, separate
2325 type signatures in the <literal>let</literal> group.
2332 The type variables bound by the pattern may be
2333 mentioned in ordinary type signatures or pattern
2334 type signatures anywhere within their scope.
2341 In ordinary type signatures, any type variable mentioned in the
2342 signature that is in scope is <emphasis>not</emphasis> universally quantified.
2350 Ordinary type signatures do not bring any new type variables
2351 into scope (except in the type signature itself!). So this is illegal:
2358 It's illegal because <VarName>a</VarName> is not in scope in the body of <function>f</function>,
2359 so the ordinary signature <literal>x::a</literal> is equivalent to <literal>x::forall a.a</literal>;
2360 and that is an incorrect typing.
2367 The pattern type signature is a monotype:
2372 A pattern type signature cannot contain any explicit <literal>forall</literal> quantification.
2376 The type variables bound by a pattern type signature can only be instantiated to monotypes,
2377 not to type schemes.
2381 There is no implicit universal quantification on pattern type signatures (in contrast to
2382 ordinary type signatures).
2392 The type variables in the head of a <literal>class</literal> or <literal>instance</literal> declaration
2393 scope over the methods defined in the <literal>where</literal> part. For example:
2407 (Not implemented in Hugs yet, Dec 98).
2418 <title>Result type signatures</title>
2426 The result type of a function can be given a signature,
2431 f (x::a) :: [a] = [x,x,x]
2435 The final <literal>:: [a]</literal> after all the patterns gives a signature to the
2436 result type. Sometimes this is the only way of naming the type variable
2441 f :: Int -> [a] -> [a]
2442 f n :: ([a] -> [a]) = let g (x::a, y::a) = (y,x)
2443 in \xs -> map g (reverse xs `zip` xs)
2455 Result type signatures are not yet implemented in Hugs.
2461 <title>Where a pattern type signature can occur</title>
2464 A pattern type signature can occur in any pattern. For example:
2469 A pattern type signature can be on an arbitrary sub-pattern, not
2474 f ((x,y)::(a,b)) = (y,x) :: (b,a)
2483 Pattern type signatures, including the result part, can be used
2484 in lambda abstractions:
2487 (\ (x::a, y) :: a -> x)
2494 Pattern type signatures, including the result part, can be used
2495 in <literal>case</literal> expressions:
2499 case e of { (x::a, y) :: a -> x }
2507 To avoid ambiguity, the type after the “<literal>::</literal>” in a result
2508 pattern signature on a lambda or <literal>case</literal> must be atomic (i.e. a single
2509 token or a parenthesised type of some sort). To see why,
2510 consider how one would parse this:
2524 Pattern type signatures can bind existential type variables.
2529 data T = forall a. MkT [a]
2532 f (MkT [t::a]) = MkT t3
2545 Pattern type signatures
2546 can be used in pattern bindings:
2549 f x = let (y, z::a) = x in ...
2550 f1 x = let (y, z::Int) = x in ...
2551 f2 (x::(Int,a)) = let (y, z::a) = x in ...
2552 f3 :: (b->b) = \x -> x
2555 In all such cases, the binding is not generalised over the pattern-bound
2556 type variables. Thus <literal>f3</literal> is monomorphic; <literal>f3</literal>
2557 has type <literal>b -> b</literal> for some type <literal>b</literal>,
2558 and <emphasis>not</emphasis> <literal>forall b. b -> b</literal>.
2559 In contrast, the binding
2564 makes a polymorphic function, but <literal>b</literal> is not in scope anywhere
2565 in <literal>f4</literal>'s scope.
2577 <sect1 id="pragmas">
2578 <title>Pragmas</title>
2580 <indexterm><primary>pragma</primary></indexterm>
2582 <para>GHC supports several pragmas, or instructions to the
2583 compiler placed in the source code. Pragmas don't normally affect
2584 the meaning of the program, but they might affect the efficiency
2585 of the generated code.</para>
2587 <para>Pragmas all take the form
2589 <literal>{-# <replaceable>word</replaceable> ... #-}</literal>
2591 where <replaceable>word</replaceable> indicates the type of
2592 pragma, and is followed optionally by information specific to that
2593 type of pragma. Case is ignored in
2594 <replaceable>word</replaceable>. The various values for
2595 <replaceable>word</replaceable> that GHC understands are described
2596 in the following sections; any pragma encountered with an
2597 unrecognised <replaceable>word</replaceable> is (silently)
2600 <sect2 id="inline-pragma">
2601 <title>INLINE pragma
2603 <indexterm><primary>INLINE pragma</primary></indexterm>
2604 <indexterm><primary>pragma, INLINE</primary></indexterm></title>
2607 GHC (with <option>-O</option>, as always) tries to inline (or “unfold”)
2608 functions/values that are “small enough,” thus avoiding the call
2609 overhead and possibly exposing other more-wonderful optimisations.
2613 You will probably see these unfoldings (in Core syntax) in your
2618 Normally, if GHC decides a function is “too expensive” to inline, it
2619 will not do so, nor will it export that unfolding for other modules to
2624 The sledgehammer you can bring to bear is the
2625 <literal>INLINE</literal><indexterm><primary>INLINE pragma</primary></indexterm> pragma, used thusly:
2628 key_function :: Int -> String -> (Bool, Double)
2630 #ifdef __GLASGOW_HASKELL__
2631 {-# INLINE key_function #-}
2635 (You don't need to do the C pre-processor carry-on unless you're going
2636 to stick the code through HBC—it doesn't like <literal>INLINE</literal> pragmas.)
2640 The major effect of an <literal>INLINE</literal> pragma is to declare a function's
2641 “cost” to be very low. The normal unfolding machinery will then be
2642 very keen to inline it.
2646 An <literal>INLINE</literal> pragma for a function can be put anywhere its type
2647 signature could be put.
2651 <literal>INLINE</literal> pragmas are a particularly good idea for the
2652 <literal>then</literal>/<literal>return</literal> (or <literal>bind</literal>/<literal>unit</literal>) functions in a monad.
2653 For example, in GHC's own <literal>UniqueSupply</literal> monad code, we have:
2656 #ifdef __GLASGOW_HASKELL__
2657 {-# INLINE thenUs #-}
2658 {-# INLINE returnUs #-}
2666 <sect2 id="noinline-pragma">
2667 <title>NOINLINE pragma
2670 <indexterm><primary>NOINLINE pragma</primary></indexterm>
2671 <indexterm><primary>pragma</primary><secondary>NOINLINE</secondary></indexterm>
2672 <indexterm><primary>NOTINLINE pragma</primary></indexterm>
2673 <indexterm><primary>pragma</primary><secondary>NOTINLINE</secondary></indexterm>
2676 The <literal>NOINLINE</literal> pragma does exactly what you'd expect:
2677 it stops the named function from being inlined by the compiler. You
2678 shouldn't ever need to do this, unless you're very cautious about code
2682 <para><literal>NOTINLINE</literal> is a synonym for
2683 <literal>NOINLINE</literal> (<literal>NOTINLINE</literal> is specified
2684 by Haskell 98 as the standard way to disable inlining, so it should be
2685 used if you want your code to be portable).</para>
2689 <sect2 id="specialize-pragma">
2690 <title>SPECIALIZE pragma</title>
2692 <indexterm><primary>SPECIALIZE pragma</primary></indexterm>
2693 <indexterm><primary>pragma, SPECIALIZE</primary></indexterm>
2694 <indexterm><primary>overloading, death to</primary></indexterm>
2696 <para>(UK spelling also accepted.) For key overloaded
2697 functions, you can create extra versions (NB: more code space)
2698 specialised to particular types. Thus, if you have an
2699 overloaded function:</para>
2702 hammeredLookup :: Ord key => [(key, value)] -> key -> value
2705 <para>If it is heavily used on lists with
2706 <literal>Widget</literal> keys, you could specialise it as
2710 {-# SPECIALIZE hammeredLookup :: [(Widget, value)] -> Widget -> value #-}
2713 <para>To get very fancy, you can also specify a named function
2714 to use for the specialised value, as in:</para>
2717 {-# RULES hammeredLookup = blah #-}
2720 <para>where <literal>blah</literal> is an implementation of
2721 <literal>hammerdLookup</literal> written specialy for
2722 <literal>Widget</literal> lookups. It's <emphasis>Your
2723 Responsibility</emphasis> to make sure that
2724 <function>blah</function> really behaves as a specialised
2725 version of <function>hammeredLookup</function>!!!</para>
2727 <para>Note we use the <literal>RULE</literal> pragma here to
2728 indicate that <literal>hammeredLookup</literal> applied at a
2729 certain type should be replaced by <literal>blah</literal>. See
2730 <xref linkend="rules"> for more information on
2731 <literal>RULES</literal>.</para>
2733 <para>An example in which using <literal>RULES</literal> for
2734 specialisation will Win Big:
2737 toDouble :: Real a => a -> Double
2738 toDouble = fromRational . toRational
2740 {-# SPECIALIZE toDouble :: Int -> Double = i2d #-}
2741 i2d (I# i) = D# (int2Double# i) -- uses Glasgow prim-op directly
2744 The <function>i2d</function> function is virtually one machine
2745 instruction; the default conversion—via an intermediate
2746 <literal>Rational</literal>—is obscenely expensive by
2749 <para>A <literal>SPECIALIZE</literal> pragma for a function can
2750 be put anywhere its type signature could be put.</para>
2754 <sect2 id="specialize-instance-pragma">
2755 <title>SPECIALIZE instance pragma
2759 <indexterm><primary>SPECIALIZE pragma</primary></indexterm>
2760 <indexterm><primary>overloading, death to</primary></indexterm>
2761 Same idea, except for instance declarations. For example:
2764 instance (Eq a) => Eq (Foo a) where {
2765 {-# SPECIALIZE instance Eq (Foo [(Int, Bar)]) #-}
2769 The pragma must occur inside the <literal>where</literal> part
2770 of the instance declaration.
2773 Compatible with HBC, by the way, except perhaps in the placement
2779 <sect2 id="line-pragma">
2784 <indexterm><primary>LINE pragma</primary></indexterm>
2785 <indexterm><primary>pragma, LINE</primary></indexterm>
2789 This pragma is similar to C's <literal>#line</literal> pragma, and is mainly for use in
2790 automatically generated Haskell code. It lets you specify the line
2791 number and filename of the original code; for example
2797 {-# LINE 42 "Foo.vhs" #-}
2803 if you'd generated the current file from something called <filename>Foo.vhs</filename>
2804 and this line corresponds to line 42 in the original. GHC will adjust
2805 its error messages to refer to the line/file named in the <literal>LINE</literal>
2812 <title>RULES pragma</title>
2815 The RULES pragma lets you specify rewrite rules. It is described in
2816 <xref LinkEnd="rewrite-rules">.
2821 <sect2 id="deprecated-pragma">
2822 <title>DEPRECATED pragma</title>
2825 The DEPRECATED pragma lets you specify that a particular function, class, or type, is deprecated.
2826 There are two forms.
2830 You can deprecate an entire module thus:</para>
2832 module Wibble {-# DEPRECATED "Use Wobble instead" #-} where
2836 When you compile any module that import <literal>Wibble</literal>, GHC will print
2837 the specified message.</para>
2842 You can deprecate a function, class, or type, with the following top-level declaration:
2845 {-# DEPRECATED f, C, T "Don't use these" #-}
2848 When you compile any module that imports and uses any of the specifed entities,
2849 GHC will print the specified message.
2853 <para>You can suppress the warnings with the flag <option>-fno-warn-deprecations</option>.</para>
2859 <sect1 id="rewrite-rules">
2860 <title>Rewrite rules
2862 <indexterm><primary>RULES pagma</primary></indexterm>
2863 <indexterm><primary>pragma, RULES</primary></indexterm>
2864 <indexterm><primary>rewrite rules</primary></indexterm></title>
2867 The programmer can specify rewrite rules as part of the source program
2868 (in a pragma). GHC applies these rewrite rules wherever it can.
2876 "map/map" forall f g xs. map f (map g xs) = map (f.g) xs
2883 <title>Syntax</title>
2886 From a syntactic point of view:
2892 Each rule has a name, enclosed in double quotes. The name itself has
2893 no significance at all. It is only used when reporting how many times the rule fired.
2899 There may be zero or more rules in a <literal>RULES</literal> pragma.
2905 Layout applies in a <literal>RULES</literal> pragma. Currently no new indentation level
2906 is set, so you must lay out your rules starting in the same column as the
2907 enclosing definitions.
2913 Each variable mentioned in a rule must either be in scope (e.g. <function>map</function>),
2914 or bound by the <literal>forall</literal> (e.g. <function>f</function>, <function>g</function>, <function>xs</function>). The variables bound by
2915 the <literal>forall</literal> are called the <emphasis>pattern</emphasis> variables. They are separated
2916 by spaces, just like in a type <literal>forall</literal>.
2922 A pattern variable may optionally have a type signature.
2923 If the type of the pattern variable is polymorphic, it <emphasis>must</emphasis> have a type signature.
2924 For example, here is the <literal>foldr/build</literal> rule:
2927 "fold/build" forall k z (g::forall b. (a->b->b) -> b -> b) .
2928 foldr k z (build g) = g k z
2931 Since <function>g</function> has a polymorphic type, it must have a type signature.
2938 The left hand side of a rule must consist of a top-level variable applied
2939 to arbitrary expressions. For example, this is <emphasis>not</emphasis> OK:
2942 "wrong1" forall e1 e2. case True of { True -> e1; False -> e2 } = e1
2943 "wrong2" forall f. f True = True
2946 In <literal>"wrong1"</literal>, the LHS is not an application; in <literal>"wrong2"</literal>, the LHS has a pattern variable
2953 A rule does not need to be in the same module as (any of) the
2954 variables it mentions, though of course they need to be in scope.
2960 Rules are automatically exported from a module, just as instance declarations are.
2971 <title>Semantics</title>
2974 From a semantic point of view:
2980 Rules are only applied if you use the <option>-O</option> flag.
2986 Rules are regarded as left-to-right rewrite rules.
2987 When GHC finds an expression that is a substitution instance of the LHS
2988 of a rule, it replaces the expression by the (appropriately-substituted) RHS.
2989 By "a substitution instance" we mean that the LHS can be made equal to the
2990 expression by substituting for the pattern variables.
2997 The LHS and RHS of a rule are typechecked, and must have the
3005 GHC makes absolutely no attempt to verify that the LHS and RHS
3006 of a rule have the same meaning. That is undecideable in general, and
3007 infeasible in most interesting cases. The responsibility is entirely the programmer's!
3014 GHC makes no attempt to make sure that the rules are confluent or
3015 terminating. For example:
3018 "loop" forall x,y. f x y = f y x
3021 This rule will cause the compiler to go into an infinite loop.
3028 If more than one rule matches a call, GHC will choose one arbitrarily to apply.
3034 GHC currently uses a very simple, syntactic, matching algorithm
3035 for matching a rule LHS with an expression. It seeks a substitution
3036 which makes the LHS and expression syntactically equal modulo alpha
3037 conversion. The pattern (rule), but not the expression, is eta-expanded if
3038 necessary. (Eta-expanding the epression can lead to laziness bugs.)
3039 But not beta conversion (that's called higher-order matching).
3043 Matching is carried out on GHC's intermediate language, which includes
3044 type abstractions and applications. So a rule only matches if the
3045 types match too. See <xref LinkEnd="rule-spec"> below.
3051 GHC keeps trying to apply the rules as it optimises the program.
3052 For example, consider:
3061 The expression <literal>s (t xs)</literal> does not match the rule <literal>"map/map"</literal>, but GHC
3062 will substitute for <VarName>s</VarName> and <VarName>t</VarName>, giving an expression which does match.
3063 If <VarName>s</VarName> or <VarName>t</VarName> was (a) used more than once, and (b) large or a redex, then it would
3064 not be substituted, and the rule would not fire.
3071 In the earlier phases of compilation, GHC inlines <emphasis>nothing
3072 that appears on the LHS of a rule</emphasis>, because once you have substituted
3073 for something you can't match against it (given the simple minded
3074 matching). So if you write the rule
3077 "map/map" forall f,g. map f . map g = map (f.g)
3080 this <emphasis>won't</emphasis> match the expression <literal>map f (map g xs)</literal>.
3081 It will only match something written with explicit use of ".".
3082 Well, not quite. It <emphasis>will</emphasis> match the expression
3088 where <function>wibble</function> is defined:
3091 wibble f g = map f . map g
3094 because <function>wibble</function> will be inlined (it's small).
3096 Later on in compilation, GHC starts inlining even things on the
3097 LHS of rules, but still leaves the rules enabled. This inlining
3098 policy is controlled by the per-simplification-pass flag <option>-finline-phase</option><emphasis>n</emphasis>.
3105 All rules are implicitly exported from the module, and are therefore
3106 in force in any module that imports the module that defined the rule, directly
3107 or indirectly. (That is, if A imports B, which imports C, then C's rules are
3108 in force when compiling A.) The situation is very similar to that for instance
3120 <title>List fusion</title>
3123 The RULES mechanism is used to implement fusion (deforestation) of common list functions.
3124 If a "good consumer" consumes an intermediate list constructed by a "good producer", the
3125 intermediate list should be eliminated entirely.
3129 The following are good producers:
3141 Enumerations of <literal>Int</literal> and <literal>Char</literal> (e.g. <literal>['a'..'z']</literal>).
3147 Explicit lists (e.g. <literal>[True, False]</literal>)
3153 The cons constructor (e.g <literal>3:4:[]</literal>)
3159 <function>++</function>
3165 <function>map</function>
3171 <function>filter</function>
3177 <function>iterate</function>, <function>repeat</function>
3183 <function>zip</function>, <function>zipWith</function>
3192 The following are good consumers:
3204 <function>array</function> (on its second argument)
3210 <function>length</function>
3216 <function>++</function> (on its first argument)
3222 <function>foldr</function>
3228 <function>map</function>
3234 <function>filter</function>
3240 <function>concat</function>
3246 <function>unzip</function>, <function>unzip2</function>, <function>unzip3</function>, <function>unzip4</function>
3252 <function>zip</function>, <function>zipWith</function> (but on one argument only; if both are good producers, <function>zip</function>
3253 will fuse with one but not the other)
3259 <function>partition</function>
3265 <function>head</function>
3271 <function>and</function>, <function>or</function>, <function>any</function>, <function>all</function>
3277 <function>sequence_</function>
3283 <function>msum</function>
3289 <function>sortBy</function>
3298 So, for example, the following should generate no intermediate lists:
3301 array (1,10) [(i,i*i) | i <- map (+ 1) [0..9]]
3307 This list could readily be extended; if there are Prelude functions that you use
3308 a lot which are not included, please tell us.
3312 If you want to write your own good consumers or producers, look at the
3313 Prelude definitions of the above functions to see how to do so.
3318 <sect2 id="rule-spec">
3319 <title>Specialisation
3323 Rewrite rules can be used to get the same effect as a feature
3324 present in earlier version of GHC:
3327 {-# SPECIALIZE fromIntegral :: Int8 -> Int16 = int8ToInt16 #-}
3330 This told GHC to use <function>int8ToInt16</function> instead of <function>fromIntegral</function> whenever
3331 the latter was called with type <literal>Int8 -> Int16</literal>. That is, rather than
3332 specialising the original definition of <function>fromIntegral</function> the programmer is
3333 promising that it is safe to use <function>int8ToInt16</function> instead.
3337 This feature is no longer in GHC. But rewrite rules let you do the
3342 "fromIntegral/Int8/Int16" fromIntegral = int8ToInt16
3346 This slightly odd-looking rule instructs GHC to replace <function>fromIntegral</function>
3347 by <function>int8ToInt16</function> <emphasis>whenever the types match</emphasis>. Speaking more operationally,
3348 GHC adds the type and dictionary applications to get the typed rule
3351 forall (d1::Integral Int8) (d2::Num Int16) .
3352 fromIntegral Int8 Int16 d1 d2 = int8ToInt16
3356 this rule does not need to be in the same file as fromIntegral,
3357 unlike the <literal>SPECIALISE</literal> pragmas which currently do (so that they
3358 have an original definition available to specialise).
3364 <title>Controlling what's going on</title>
3372 Use <option>-ddump-rules</option> to see what transformation rules GHC is using.
3378 Use <option>-ddump-simpl-stats</option> to see what rules are being fired.
3379 If you add <option>-dppr-debug</option> you get a more detailed listing.
3385 The defintion of (say) <function>build</function> in <FileName>PrelBase.lhs</FileName> looks llike this:
3388 build :: forall a. (forall b. (a -> b -> b) -> b -> b) -> [a]
3389 {-# INLINE build #-}
3393 Notice the <literal>INLINE</literal>! That prevents <literal>(:)</literal> from being inlined when compiling
3394 <literal>PrelBase</literal>, so that an importing module will “see” the <literal>(:)</literal>, and can
3395 match it on the LHS of a rule. <literal>INLINE</literal> prevents any inlining happening
3396 in the RHS of the <literal>INLINE</literal> thing. I regret the delicacy of this.
3403 In <filename>ghc/lib/std/PrelBase.lhs</filename> look at the rules for <function>map</function> to
3404 see how to write rules that will do fusion and yet give an efficient
3405 program even if fusion doesn't happen. More rules in <filename>PrelList.lhs</filename>.
3417 <sect1 id="generic-classes">
3418 <title>Generic classes</title>
3420 <para>(Note: support for generic classes is currently broken in
3424 The ideas behind this extension are described in detail in "Derivable type classes",
3425 Ralf Hinze and Simon Peyton Jones, Haskell Workshop, Montreal Sept 2000, pp94-105.
3426 An example will give the idea:
3434 fromBin :: [Int] -> (a, [Int])
3436 toBin {| Unit |} Unit = []
3437 toBin {| a :+: b |} (Inl x) = 0 : toBin x
3438 toBin {| a :+: b |} (Inr y) = 1 : toBin y
3439 toBin {| a :*: b |} (x :*: y) = toBin x ++ toBin y
3441 fromBin {| Unit |} bs = (Unit, bs)
3442 fromBin {| a :+: b |} (0:bs) = (Inl x, bs') where (x,bs') = fromBin bs
3443 fromBin {| a :+: b |} (1:bs) = (Inr y, bs') where (y,bs') = fromBin bs
3444 fromBin {| a :*: b |} bs = (x :*: y, bs'') where (x,bs' ) = fromBin bs
3445 (y,bs'') = fromBin bs'
3448 This class declaration explains how <literal>toBin</literal> and <literal>fromBin</literal>
3449 work for arbitrary data types. They do so by giving cases for unit, product, and sum,
3450 which are defined thus in the library module <literal>Generics</literal>:
3454 data a :+: b = Inl a | Inr b
3455 data a :*: b = a :*: b
3458 Now you can make a data type into an instance of Bin like this:
3460 instance (Bin a, Bin b) => Bin (a,b)
3461 instance Bin a => Bin [a]
3463 That is, just leave off the "where" clasuse. Of course, you can put in the
3464 where clause and over-ride whichever methods you please.
3468 <title> Using generics </title>
3469 <para>To use generics you need to</para>
3472 <para>Use the flags <option>-fglasgow-exts</option> (to enable the extra syntax),
3473 <option>-fgenerics</option> (to generate extra per-data-type code),
3474 and <option>-package lang</option> (to make the <literal>Generics</literal> library
3478 <para>Import the module <literal>Generics</literal> from the
3479 <literal>lang</literal> package. This import brings into
3480 scope the data types <literal>Unit</literal>,
3481 <literal>:*:</literal>, and <literal>:+:</literal>. (You
3482 don't need this import if you don't mention these types
3483 explicitly; for example, if you are simply giving instance
3484 declarations.)</para>
3489 <sect2> <title> Changes wrt the paper </title>
3491 Note that the type constructors <literal>:+:</literal> and <literal>:*:</literal>
3492 can be written infix (indeed, you can now use
3493 any operator starting in a colon as an infix type constructor). Also note that
3494 the type constructors are not exactly as in the paper (Unit instead of 1, etc).
3495 Finally, note that the syntax of the type patterns in the class declaration
3496 uses "<literal>{|</literal>" and "<literal>|}</literal>" brackets; curly braces
3497 alone would ambiguous when they appear on right hand sides (an extension we
3498 anticipate wanting).
3502 <sect2> <title>Terminology and restrictions</title>
3504 Terminology. A "generic default method" in a class declaration
3505 is one that is defined using type patterns as above.
3506 A "polymorphic default method" is a default method defined as in Haskell 98.
3507 A "generic class declaration" is a class declaration with at least one
3508 generic default method.
3516 Alas, we do not yet implement the stuff about constructor names and
3523 A generic class can have only one parameter; you can't have a generic
3524 multi-parameter class.
3530 A default method must be defined entirely using type patterns, or entirely
3531 without. So this is illegal:
3534 op :: a -> (a, Bool)
3535 op {| Unit |} Unit = (Unit, True)
3538 However it is perfectly OK for some methods of a generic class to have
3539 generic default methods and others to have polymorphic default methods.
3545 The type variable(s) in the type pattern for a generic method declaration
3546 scope over the right hand side. So this is legal (note the use of the type variable ``p'' in a type signature on the right hand side:
3550 op {| p :*: q |} (x :*: y) = op (x :: p)
3558 The type patterns in a generic default method must take one of the forms:
3564 where "a" and "b" are type variables. Furthermore, all the type patterns for
3565 a single type constructor (<literal>:*:</literal>, say) must be identical; they
3566 must use the same type variables. So this is illegal:
3570 op {| a :+: b |} (Inl x) = True
3571 op {| p :+: q |} (Inr y) = False
3573 The type patterns must be identical, even in equations for different methods of the class.
3574 So this too is illegal:
3578 op1 {| a :*: b |} (x :*: y) = True
3581 op2 {| p :*: q |} (x :*: y) = False
3583 (The reason for this restriction is that we gather all the equations for a particular type consructor
3584 into a single generic instance declaration.)
3590 A generic method declaration must give a case for each of the three type constructors.
3596 The type for a generic method can be built only from:
3598 <listitem> <para> Function arrows </para> </listitem>
3599 <listitem> <para> Type variables </para> </listitem>
3600 <listitem> <para> Tuples </para> </listitem>
3601 <listitem> <para> Arbitrary types not involving type variables </para> </listitem>
3603 Here are some example type signatures for generic methods:
3606 op2 :: Bool -> (a,Bool)
3607 op3 :: [Int] -> a -> a
3610 Here, op1, op2, op3 are OK, but op4 is rejected, because it has a type variable
3614 This restriction is an implementation restriction: we just havn't got around to
3615 implementing the necessary bidirectional maps over arbitrary type constructors.
3616 It would be relatively easy to add specific type constructors, such as Maybe and list,
3617 to the ones that are allowed.</para>
3622 In an instance declaration for a generic class, the idea is that the compiler
3623 will fill in the methods for you, based on the generic templates. However it can only
3628 The instance type is simple (a type constructor applied to type variables, as in Haskell 98).
3633 No constructor of the instance type has unboxed fields.
3637 (Of course, these things can only arise if you are already using GHC extensions.)
3638 However, you can still give an instance declarations for types which break these rules,
3639 provided you give explicit code to override any generic default methods.
3647 The option <option>-ddump-deriv</option> dumps incomprehensible stuff giving details of
3648 what the compiler does with generic declarations.
3653 <sect2> <title> Another example </title>
3655 Just to finish with, here's another example I rather like:
3659 nCons {| Unit |} _ = 1
3660 nCons {| a :*: b |} _ = 1
3661 nCons {| a :+: b |} _ = nCons (bot::a) + nCons (bot::b)
3664 tag {| Unit |} _ = 1
3665 tag {| a :*: b |} _ = 1
3666 tag {| a :+: b |} (Inl x) = tag x
3667 tag {| a :+: b |} (Inr y) = nCons (bot::a) + tag y
3674 ;;; Local Variables: ***
3676 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter" "sect1") ***