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>Local universal quantification:</term>
65 <para><XRef LinkEnd="universal-quantification"></para>
70 <term>Extistentially quantification in data types:</term>
72 <para><XRef LinkEnd="existential-quantification"></para>
77 <term>Scoped type variables:</term>
79 <para>Scoped type variables enable the programmer to
80 supply type signatures for some nested declarations,
81 where this would not be legal in Haskell 98. Details in
82 <XRef LinkEnd="scoped-type-variables">.</para>
90 <term>Pattern guards</term>
92 <para>Instead of being a boolean expression, a guard is a list
93 of qualifiers, exactly as in a list comprehension. See <XRef
94 LinkEnd="pattern-guards">.</para>
99 <term>Foreign calling:</term>
101 <para>Just what it sounds like. We provide
102 <Emphasis>lots</Emphasis> of rope that you can dangle around
103 your neck. Please see <XRef LinkEnd="ffi">.</para>
110 <para>Pragmas are special instructions to the compiler placed
111 in the source file. The pragmas GHC supports are described in
112 <XRef LinkEnd="pragmas">.</para>
117 <term>Rewrite rules:</term>
119 <para>The programmer can specify rewrite rules as part of the
120 source program (in a pragma). GHC applies these rewrite rules
121 wherever it can. Details in <XRef
122 LinkEnd="rewrite-rules">.</para>
127 <term>Generic classes:</term>
129 <para>Generic class declarations allow you to define a class
130 whose methods say how to work over an arbitrary data type.
131 Then it's really easy to make any new type into an instance of
132 the class. This generalises the rather ad-hoc "deriving"
133 feature of Haskell 98. Details in <XRef
134 LinkEnd="generic-classes">.</para>
140 Before you get too carried away working at the lowest level (e.g.,
141 sloshing <Literal>MutableByteArray#</Literal>s around your
142 program), you may wish to check if there are libraries that provide a
143 “Haskellised veneer” over the features you want. See
144 <xref linkend="book-hslibs">.
147 <sect1 id="options-language">
148 <title>Language options</title>
150 <indexterm><primary>language</primary><secondary>option</secondary>
152 <indexterm><primary>options</primary><secondary>language</secondary>
154 <indexterm><primary>extensions</primary><secondary>options controlling</secondary>
157 <para> These flags control what variation of the language are
158 permitted. Leaving out all of them gives you standard Haskell
164 <term><option>-fglasgow-exts</option>:</term>
165 <indexterm><primary><option>-fglasgow-exts</option></primary></indexterm>
167 <para>This simultaneously enables all of the extensions to
168 Haskell 98 described in <xref
169 linkend="ghc-language-features">, except where otherwise
175 <term><option>-fno-monomorphism-restriction</option>:</term>
176 <indexterm><primary><option>-fno-monomorphism-restriction</option></primary></indexterm>
178 <para> Switch off the Haskell 98 monomorphism restriction.
179 Independent of the <Option>-fglasgow-exts</Option>
185 <term><option>-fallow-overlapping-instances</option></term>
186 <term><option>-fallow-undecidable-instances</option></term>
187 <term><option>-fcontext-stack</option></term>
188 <indexterm><primary><option>-fallow-overlapping-instances</option></primary></indexterm>
189 <indexterm><primary><option>-fallow-undecidable-instances</option></primary></indexterm>
190 <indexterm><primary><option>-fcontext-stack</option></primary></indexterm>
192 <para> See <XRef LinkEnd="instance-decls">. Only relevant
193 if you also use <option>-fglasgow-exts</option>.</para>
198 <term><option>-finline-phase</option></term>
199 <indexterm><primary><option>-finline-phase</option></primary></indexterm>
201 <para>See <XRef LinkEnd="rewrite-rules">. Only relevant if
202 you also use <Option>-fglasgow-exts</Option>.</para>
207 <term><option>-fgenerics</option></term>
208 <indexterm><primary><option>-fgenerics</option></primary></indexterm>
210 <para>See <XRef LinkEnd="generic-classes">. Independent of
211 <Option>-fglasgow-exts</Option>.</para>
216 <term><option>-fno-implicit-prelude</option></term>
218 <para><indexterm><primary>-fno-implicit-prelude
219 option</primary></indexterm> GHC normally imports
220 <filename>Prelude.hi</filename> files for you. If you'd
221 rather it didn't, then give it a
222 <option>-fno-implicit-prelude</option> option. The idea
223 is that you can then import a Prelude of your own. (But
224 don't call it <literal>Prelude</literal>; the Haskell
225 module namespace is flat, and you must not conflict with
226 any Prelude module.)</para>
228 <para>Even though you have not imported the Prelude, all
229 the built-in syntax still refers to the built-in Haskell
230 Prelude types and values, as specified by the Haskell
231 Report. For example, the type <literal>[Int]</literal>
232 still means <literal>Prelude.[] Int</literal>; tuples
233 continue to refer to the standard Prelude tuples; the
234 translation for list comprehensions continues to use
235 <literal>Prelude.map</literal> etc.</para>
237 <para> With one group of exceptions! You may want to
238 define your own numeric class hierarchy. It completely
239 defeats that purpose if the literal "1" means
240 "<literal>Prelude.fromInteger 1</literal>", which is what
241 the Haskell Report specifies. So the
242 <option>-fno-implicit-prelude</option> flag causes the
243 following pieces of built-in syntax to refer to whatever
244 is in scope, not the Prelude versions:</para>
248 <para>Integer and fractional literals mean
249 "<literal>fromInteger 1</literal>" and
250 "<literal>fromRational 3.2</literal>", not the
251 Prelude-qualified versions; both in expressions and in
256 <para>Negation (e.g. "<literal>- (f x)</literal>")
257 means "<literal>negate (f x)</literal>" (not
258 <literal>Prelude.negate</literal>).</para>
262 <para>In an n+k pattern, the standard Prelude
263 <literal>Ord</literal> class is used for comparison,
264 but the necessary subtraction uses whatever
265 "<literal>(-)</literal>" is in scope (not
266 "<literal>Prelude.(-)</literal>").</para>
276 <Sect1 id="primitives">
277 <Title>Unboxed types and primitive operations
279 <IndexTerm><Primary>PrelGHC module</Primary></IndexTerm>
282 This module defines all the types which are primitive in Glasgow
283 Haskell, and the operations provided for them.
286 <Sect2 id="glasgow-unboxed">
291 <IndexTerm><Primary>Unboxed types (Glasgow extension)</Primary></IndexTerm>
294 <para>Most types in GHC are <firstterm>boxed</firstterm>, which means
295 that values of that type are represented by a pointer to a heap
296 object. The representation of a Haskell <literal>Int</literal>, for
297 example, is a two-word heap object. An <firstterm>unboxed</firstterm>
298 type, however, is represented by the value itself, no pointers or heap
299 allocation are involved.
303 Unboxed types correspond to the “raw machine” types you
304 would use in C: <Literal>Int#</Literal> (long int),
305 <Literal>Double#</Literal> (double), <Literal>Addr#</Literal>
306 (void *), etc. The <Emphasis>primitive operations</Emphasis>
307 (PrimOps) on these types are what you might expect; e.g.,
308 <Literal>(+#)</Literal> is addition on
309 <Literal>Int#</Literal>s, and is the machine-addition that we all
310 know and love—usually one instruction.
314 Primitive (unboxed) types cannot be defined in Haskell, and are
315 therefore built into the language and compiler. Primitive types are
316 always unlifted; that is, a value of a primitive type cannot be
317 bottom. We use the convention that primitive types, values, and
318 operations have a <Literal>#</Literal> suffix.
322 Primitive values are often represented by a simple bit-pattern, such
323 as <Literal>Int#</Literal>, <Literal>Float#</Literal>,
324 <Literal>Double#</Literal>. But this is not necessarily the case:
325 a primitive value might be represented by a pointer to a
326 heap-allocated object. Examples include
327 <Literal>Array#</Literal>, the type of primitive arrays. A
328 primitive array is heap-allocated because it is too big a value to fit
329 in a register, and would be too expensive to copy around; in a sense,
330 it is accidental that it is represented by a pointer. If a pointer
331 represents a primitive value, then it really does point to that value:
332 no unevaluated thunks, no indirections…nothing can be at the
333 other end of the pointer than the primitive value.
337 There are some restrictions on the use of primitive types, the main
338 one being that you can't pass a primitive value to a polymorphic
339 function or store one in a polymorphic data type. This rules out
340 things like <Literal>[Int#]</Literal> (i.e. lists of primitive
341 integers). The reason for this restriction is that polymorphic
342 arguments and constructor fields are assumed to be pointers: if an
343 unboxed integer is stored in one of these, the garbage collector would
344 attempt to follow it, leading to unpredictable space leaks. Or a
345 <Function>seq</Function> operation on the polymorphic component may
346 attempt to dereference the pointer, with disastrous results. Even
347 worse, the unboxed value might be larger than a pointer
348 (<Literal>Double#</Literal> for instance).
352 Nevertheless, A numerically-intensive program using unboxed types can
353 go a <Emphasis>lot</Emphasis> faster than its “standard”
354 counterpart—we saw a threefold speedup on one example.
359 <Sect2 id="unboxed-tuples">
360 <Title>Unboxed Tuples
364 Unboxed tuples aren't really exported by <Literal>PrelGHC</Literal>,
365 they're available by default with <Option>-fglasgow-exts</Option>. An
366 unboxed tuple looks like this:
378 where <Literal>e_1..e_n</Literal> are expressions of any
379 type (primitive or non-primitive). The type of an unboxed tuple looks
384 Unboxed tuples are used for functions that need to return multiple
385 values, but they avoid the heap allocation normally associated with
386 using fully-fledged tuples. When an unboxed tuple is returned, the
387 components are put directly into registers or on the stack; the
388 unboxed tuple itself does not have a composite representation. Many
389 of the primitive operations listed in this section return unboxed
394 There are some pretty stringent restrictions on the use of unboxed tuples:
403 Unboxed tuple types are subject to the same restrictions as
404 other unboxed types; i.e. they may not be stored in polymorphic data
405 structures or passed to polymorphic functions.
412 Unboxed tuples may only be constructed as the direct result of
413 a function, and may only be deconstructed with a <Literal>case</Literal> expression.
414 eg. the following are valid:
418 f x y = (# x+1, y-1 #)
419 g x = case f x x of { (# a, b #) -> a + b }
423 but the following are invalid:
437 No variable can have an unboxed tuple type. This is illegal:
441 f :: (# Int, Int #) -> (# Int, Int #)
446 because <VarName>x</VarName> has an unboxed tuple type.
456 Note: we may relax some of these restrictions in the future.
460 The <Literal>IO</Literal> and <Literal>ST</Literal> monads use unboxed
461 tuples to avoid unnecessary allocation during sequences of operations.
467 <Title>Character and numeric types</Title>
469 <IndexTerm><Primary>character types, primitive</Primary></IndexTerm>
470 <IndexTerm><Primary>numeric types, primitive</Primary></IndexTerm>
471 <IndexTerm><Primary>integer types, primitive</Primary></IndexTerm>
472 <IndexTerm><Primary>floating point types, primitive</Primary></IndexTerm>
474 There are the following obvious primitive types:
488 <IndexTerm><Primary><literal>Char#</literal></Primary></IndexTerm>
489 <IndexTerm><Primary><literal>Int#</literal></Primary></IndexTerm>
490 <IndexTerm><Primary><literal>Word#</literal></Primary></IndexTerm>
491 <IndexTerm><Primary><literal>Addr#</literal></Primary></IndexTerm>
492 <IndexTerm><Primary><literal>Float#</literal></Primary></IndexTerm>
493 <IndexTerm><Primary><literal>Double#</literal></Primary></IndexTerm>
494 <IndexTerm><Primary><literal>Int64#</literal></Primary></IndexTerm>
495 <IndexTerm><Primary><literal>Word64#</literal></Primary></IndexTerm>
498 If you really want to know their exact equivalents in C, see
499 <Filename>ghc/includes/StgTypes.h</Filename> in the GHC source tree.
503 Literals for these types may be written as follows:
512 'a'# a Char#; for weird characters, use e.g. '\o<octal>'#
513 "a"# an Addr# (a `char *'); only characters '\0'..'\255' allowed
516 <IndexTerm><Primary>literals, primitive</Primary></IndexTerm>
517 <IndexTerm><Primary>constants, primitive</Primary></IndexTerm>
518 <IndexTerm><Primary>numbers, primitive</Primary></IndexTerm>
524 <Title>Comparison operations</Title>
527 <IndexTerm><Primary>comparisons, primitive</Primary></IndexTerm>
528 <IndexTerm><Primary>operators, comparison</Primary></IndexTerm>
534 {>,>=,==,/=,<,<=}# :: Int# -> Int# -> Bool
536 {gt,ge,eq,ne,lt,le}Char# :: Char# -> Char# -> Bool
537 -- ditto for Word# and Addr#
540 <IndexTerm><Primary><literal>>#</literal></Primary></IndexTerm>
541 <IndexTerm><Primary><literal>>=#</literal></Primary></IndexTerm>
542 <IndexTerm><Primary><literal>==#</literal></Primary></IndexTerm>
543 <IndexTerm><Primary><literal>/=#</literal></Primary></IndexTerm>
544 <IndexTerm><Primary><literal><#</literal></Primary></IndexTerm>
545 <IndexTerm><Primary><literal><=#</literal></Primary></IndexTerm>
546 <IndexTerm><Primary><literal>gt{Char,Word,Addr}#</literal></Primary></IndexTerm>
547 <IndexTerm><Primary><literal>ge{Char,Word,Addr}#</literal></Primary></IndexTerm>
548 <IndexTerm><Primary><literal>eq{Char,Word,Addr}#</literal></Primary></IndexTerm>
549 <IndexTerm><Primary><literal>ne{Char,Word,Addr}#</literal></Primary></IndexTerm>
550 <IndexTerm><Primary><literal>lt{Char,Word,Addr}#</literal></Primary></IndexTerm>
551 <IndexTerm><Primary><literal>le{Char,Word,Addr}#</literal></Primary></IndexTerm>
557 <Title>Primitive-character operations</Title>
560 <IndexTerm><Primary>characters, primitive operations</Primary></IndexTerm>
561 <IndexTerm><Primary>operators, primitive character</Primary></IndexTerm>
567 ord# :: Char# -> Int#
568 chr# :: Int# -> Char#
571 <IndexTerm><Primary><literal>ord#</literal></Primary></IndexTerm>
572 <IndexTerm><Primary><literal>chr#</literal></Primary></IndexTerm>
578 <Title>Primitive-<Literal>Int</Literal> operations</Title>
581 <IndexTerm><Primary>integers, primitive operations</Primary></IndexTerm>
582 <IndexTerm><Primary>operators, primitive integer</Primary></IndexTerm>
588 {+,-,*,quotInt,remInt,gcdInt}# :: Int# -> Int# -> Int#
589 negateInt# :: Int# -> Int#
591 iShiftL#, iShiftRA#, iShiftRL# :: Int# -> Int# -> Int#
592 -- shift left, right arithmetic, right logical
594 addIntC#, subIntC#, mulIntC# :: Int# -> Int# -> (# Int#, Int# #)
595 -- add, subtract, multiply with carry
598 <IndexTerm><Primary><literal>+#</literal></Primary></IndexTerm>
599 <IndexTerm><Primary><literal>-#</literal></Primary></IndexTerm>
600 <IndexTerm><Primary><literal>*#</literal></Primary></IndexTerm>
601 <IndexTerm><Primary><literal>quotInt#</literal></Primary></IndexTerm>
602 <IndexTerm><Primary><literal>remInt#</literal></Primary></IndexTerm>
603 <IndexTerm><Primary><literal>gcdInt#</literal></Primary></IndexTerm>
604 <IndexTerm><Primary><literal>iShiftL#</literal></Primary></IndexTerm>
605 <IndexTerm><Primary><literal>iShiftRA#</literal></Primary></IndexTerm>
606 <IndexTerm><Primary><literal>iShiftRL#</literal></Primary></IndexTerm>
607 <IndexTerm><Primary><literal>addIntC#</literal></Primary></IndexTerm>
608 <IndexTerm><Primary><literal>subIntC#</literal></Primary></IndexTerm>
609 <IndexTerm><Primary><literal>mulIntC#</literal></Primary></IndexTerm>
610 <IndexTerm><Primary>shift operations, integer</Primary></IndexTerm>
614 <Emphasis>Note:</Emphasis> No error/overflow checking!
620 <Title>Primitive-<Literal>Double</Literal> and <Literal>Float</Literal> operations</Title>
623 <IndexTerm><Primary>floating point numbers, primitive</Primary></IndexTerm>
624 <IndexTerm><Primary>operators, primitive floating point</Primary></IndexTerm>
630 {+,-,*,/}## :: Double# -> Double# -> Double#
631 {<,<=,==,/=,>=,>}## :: Double# -> Double# -> Bool
632 negateDouble# :: Double# -> Double#
633 double2Int# :: Double# -> Int#
634 int2Double# :: Int# -> Double#
636 {plus,minux,times,divide}Float# :: Float# -> Float# -> Float#
637 {gt,ge,eq,ne,lt,le}Float# :: Float# -> Float# -> Bool
638 negateFloat# :: Float# -> Float#
639 float2Int# :: Float# -> Int#
640 int2Float# :: Int# -> Float#
646 <IndexTerm><Primary><literal>+##</literal></Primary></IndexTerm>
647 <IndexTerm><Primary><literal>-##</literal></Primary></IndexTerm>
648 <IndexTerm><Primary><literal>*##</literal></Primary></IndexTerm>
649 <IndexTerm><Primary><literal>/##</literal></Primary></IndexTerm>
650 <IndexTerm><Primary><literal><##</literal></Primary></IndexTerm>
651 <IndexTerm><Primary><literal><=##</literal></Primary></IndexTerm>
652 <IndexTerm><Primary><literal>==##</literal></Primary></IndexTerm>
653 <IndexTerm><Primary><literal>=/##</literal></Primary></IndexTerm>
654 <IndexTerm><Primary><literal>>=##</literal></Primary></IndexTerm>
655 <IndexTerm><Primary><literal>>##</literal></Primary></IndexTerm>
656 <IndexTerm><Primary><literal>negateDouble#</literal></Primary></IndexTerm>
657 <IndexTerm><Primary><literal>double2Int#</literal></Primary></IndexTerm>
658 <IndexTerm><Primary><literal>int2Double#</literal></Primary></IndexTerm>
662 <IndexTerm><Primary><literal>plusFloat#</literal></Primary></IndexTerm>
663 <IndexTerm><Primary><literal>minusFloat#</literal></Primary></IndexTerm>
664 <IndexTerm><Primary><literal>timesFloat#</literal></Primary></IndexTerm>
665 <IndexTerm><Primary><literal>divideFloat#</literal></Primary></IndexTerm>
666 <IndexTerm><Primary><literal>gtFloat#</literal></Primary></IndexTerm>
667 <IndexTerm><Primary><literal>geFloat#</literal></Primary></IndexTerm>
668 <IndexTerm><Primary><literal>eqFloat#</literal></Primary></IndexTerm>
669 <IndexTerm><Primary><literal>neFloat#</literal></Primary></IndexTerm>
670 <IndexTerm><Primary><literal>ltFloat#</literal></Primary></IndexTerm>
671 <IndexTerm><Primary><literal>leFloat#</literal></Primary></IndexTerm>
672 <IndexTerm><Primary><literal>negateFloat#</literal></Primary></IndexTerm>
673 <IndexTerm><Primary><literal>float2Int#</literal></Primary></IndexTerm>
674 <IndexTerm><Primary><literal>int2Float#</literal></Primary></IndexTerm>
678 And a full complement of trigonometric functions:
684 expDouble# :: Double# -> Double#
685 logDouble# :: Double# -> Double#
686 sqrtDouble# :: Double# -> Double#
687 sinDouble# :: Double# -> Double#
688 cosDouble# :: Double# -> Double#
689 tanDouble# :: Double# -> Double#
690 asinDouble# :: Double# -> Double#
691 acosDouble# :: Double# -> Double#
692 atanDouble# :: Double# -> Double#
693 sinhDouble# :: Double# -> Double#
694 coshDouble# :: Double# -> Double#
695 tanhDouble# :: Double# -> Double#
696 powerDouble# :: Double# -> Double# -> Double#
699 <IndexTerm><Primary>trigonometric functions, primitive</Primary></IndexTerm>
703 similarly for <Literal>Float#</Literal>.
707 There are two coercion functions for <Literal>Float#</Literal>/<Literal>Double#</Literal>:
713 float2Double# :: Float# -> Double#
714 double2Float# :: Double# -> Float#
717 <IndexTerm><Primary><literal>float2Double#</literal></Primary></IndexTerm>
718 <IndexTerm><Primary><literal>double2Float#</literal></Primary></IndexTerm>
722 The primitive version of <Function>decodeDouble</Function>
723 (<Function>encodeDouble</Function> is implemented as an external C
730 decodeDouble# :: Double# -> PrelNum.ReturnIntAndGMP
733 <IndexTerm><Primary><literal>encodeDouble#</literal></Primary></IndexTerm>
734 <IndexTerm><Primary><literal>decodeDouble#</literal></Primary></IndexTerm>
738 (And the same for <Literal>Float#</Literal>s.)
743 <Sect2 id="integer-operations">
744 <Title>Operations on/for <Literal>Integers</Literal> (interface to GMP)
748 <IndexTerm><Primary>arbitrary precision integers</Primary></IndexTerm>
749 <IndexTerm><Primary>Integer, operations on</Primary></IndexTerm>
753 We implement <Literal>Integers</Literal> (arbitrary-precision
754 integers) using the GNU multiple-precision (GMP) package (version
759 The data type for <Literal>Integer</Literal> is either a small
760 integer, represented by an <Literal>Int</Literal>, or a large integer
761 represented using the pieces required by GMP's
762 <Literal>MP_INT</Literal> in <Filename>gmp.h</Filename> (see
763 <Filename>gmp.info</Filename> in
764 <Filename>ghc/includes/runtime/gmp</Filename>). It comes out as:
770 data Integer = S# Int# -- small integers
771 | J# Int# ByteArray# -- large integers
774 <IndexTerm><Primary>Integer type</Primary></IndexTerm> The primitive
775 ops to support large <Literal>Integers</Literal> use the
776 “pieces” of the representation, and are as follows:
782 negateInteger# :: Int# -> ByteArray# -> Integer
784 {plus,minus,times}Integer#, gcdInteger#,
785 quotInteger#, remInteger#, divExactInteger#
786 :: Int# -> ByteArray#
787 -> Int# -> ByteArray#
788 -> (# Int#, ByteArray# #)
791 :: Int# -> ByteArray#
792 -> Int# -> ByteArray#
793 -> Int# -- -1 for <; 0 for ==; +1 for >
796 :: Int# -> ByteArray#
798 -> Int# -- -1 for <; 0 for ==; +1 for >
801 :: Int# -> ByteArray#
805 divModInteger#, quotRemInteger#
806 :: Int# -> ByteArray#
807 -> Int# -> ByteArray#
808 -> (# Int#, ByteArray#,
811 integer2Int# :: Int# -> ByteArray# -> Int#
813 int2Integer# :: Int# -> Integer -- NB: no error-checking on these two!
814 word2Integer# :: Word# -> Integer
816 addr2Integer# :: Addr# -> Integer
817 -- the Addr# is taken to be a `char *' string
818 -- to be converted into an Integer.
821 <IndexTerm><Primary><literal>negateInteger#</literal></Primary></IndexTerm>
822 <IndexTerm><Primary><literal>plusInteger#</literal></Primary></IndexTerm>
823 <IndexTerm><Primary><literal>minusInteger#</literal></Primary></IndexTerm>
824 <IndexTerm><Primary><literal>timesInteger#</literal></Primary></IndexTerm>
825 <IndexTerm><Primary><literal>quotInteger#</literal></Primary></IndexTerm>
826 <IndexTerm><Primary><literal>remInteger#</literal></Primary></IndexTerm>
827 <IndexTerm><Primary><literal>gcdInteger#</literal></Primary></IndexTerm>
828 <IndexTerm><Primary><literal>gcdIntegerInt#</literal></Primary></IndexTerm>
829 <IndexTerm><Primary><literal>divExactInteger#</literal></Primary></IndexTerm>
830 <IndexTerm><Primary><literal>cmpInteger#</literal></Primary></IndexTerm>
831 <IndexTerm><Primary><literal>divModInteger#</literal></Primary></IndexTerm>
832 <IndexTerm><Primary><literal>quotRemInteger#</literal></Primary></IndexTerm>
833 <IndexTerm><Primary><literal>integer2Int#</literal></Primary></IndexTerm>
834 <IndexTerm><Primary><literal>int2Integer#</literal></Primary></IndexTerm>
835 <IndexTerm><Primary><literal>word2Integer#</literal></Primary></IndexTerm>
836 <IndexTerm><Primary><literal>addr2Integer#</literal></Primary></IndexTerm>
842 <Title>Words and addresses</Title>
845 <IndexTerm><Primary>word, primitive type</Primary></IndexTerm>
846 <IndexTerm><Primary>address, primitive type</Primary></IndexTerm>
847 <IndexTerm><Primary>unsigned integer, primitive type</Primary></IndexTerm>
848 <IndexTerm><Primary>pointer, primitive type</Primary></IndexTerm>
852 A <Literal>Word#</Literal> is used for bit-twiddling operations.
853 It is the same size as an <Literal>Int#</Literal>, but has no sign
854 nor any arithmetic operations.
857 type Word# -- Same size/etc as Int# but *unsigned*
858 type Addr# -- A pointer from outside the "Haskell world" (from C, probably);
859 -- described under "arrays"
862 <IndexTerm><Primary><literal>Word#</literal></Primary></IndexTerm>
863 <IndexTerm><Primary><literal>Addr#</literal></Primary></IndexTerm>
867 <Literal>Word#</Literal>s and <Literal>Addr#</Literal>s have
868 the usual comparison operations. Other
869 unboxed-<Literal>Word</Literal> ops (bit-twiddling and coercions):
875 {gt,ge,eq,ne,lt,le}Word# :: Word# -> Word# -> Bool
877 and#, or#, xor# :: Word# -> Word# -> Word#
880 quotWord#, remWord# :: Word# -> Word# -> Word#
881 -- word (i.e. unsigned) versions are different from int
882 -- versions, so we have to provide these explicitly.
884 not# :: Word# -> Word#
886 shiftL#, shiftRL# :: Word# -> Int# -> Word#
887 -- shift left, right logical
889 int2Word# :: Int# -> Word# -- just a cast, really
890 word2Int# :: Word# -> Int#
893 <IndexTerm><Primary>bit operations, Word and Addr</Primary></IndexTerm>
894 <IndexTerm><Primary><literal>gtWord#</literal></Primary></IndexTerm>
895 <IndexTerm><Primary><literal>geWord#</literal></Primary></IndexTerm>
896 <IndexTerm><Primary><literal>eqWord#</literal></Primary></IndexTerm>
897 <IndexTerm><Primary><literal>neWord#</literal></Primary></IndexTerm>
898 <IndexTerm><Primary><literal>ltWord#</literal></Primary></IndexTerm>
899 <IndexTerm><Primary><literal>leWord#</literal></Primary></IndexTerm>
900 <IndexTerm><Primary><literal>and#</literal></Primary></IndexTerm>
901 <IndexTerm><Primary><literal>or#</literal></Primary></IndexTerm>
902 <IndexTerm><Primary><literal>xor#</literal></Primary></IndexTerm>
903 <IndexTerm><Primary><literal>not#</literal></Primary></IndexTerm>
904 <IndexTerm><Primary><literal>quotWord#</literal></Primary></IndexTerm>
905 <IndexTerm><Primary><literal>remWord#</literal></Primary></IndexTerm>
906 <IndexTerm><Primary><literal>shiftL#</literal></Primary></IndexTerm>
907 <IndexTerm><Primary><literal>shiftRA#</literal></Primary></IndexTerm>
908 <IndexTerm><Primary><literal>shiftRL#</literal></Primary></IndexTerm>
909 <IndexTerm><Primary><literal>int2Word#</literal></Primary></IndexTerm>
910 <IndexTerm><Primary><literal>word2Int#</literal></Primary></IndexTerm>
914 Unboxed-<Literal>Addr</Literal> ops (C casts, really):
917 {gt,ge,eq,ne,lt,le}Addr# :: Addr# -> Addr# -> Bool
919 int2Addr# :: Int# -> Addr#
920 addr2Int# :: Addr# -> Int#
921 addr2Integer# :: Addr# -> (# Int#, ByteArray# #)
924 <IndexTerm><Primary><literal>gtAddr#</literal></Primary></IndexTerm>
925 <IndexTerm><Primary><literal>geAddr#</literal></Primary></IndexTerm>
926 <IndexTerm><Primary><literal>eqAddr#</literal></Primary></IndexTerm>
927 <IndexTerm><Primary><literal>neAddr#</literal></Primary></IndexTerm>
928 <IndexTerm><Primary><literal>ltAddr#</literal></Primary></IndexTerm>
929 <IndexTerm><Primary><literal>leAddr#</literal></Primary></IndexTerm>
930 <IndexTerm><Primary><literal>int2Addr#</literal></Primary></IndexTerm>
931 <IndexTerm><Primary><literal>addr2Int#</literal></Primary></IndexTerm>
932 <IndexTerm><Primary><literal>addr2Integer#</literal></Primary></IndexTerm>
936 The casts between <Literal>Int#</Literal>,
937 <Literal>Word#</Literal> and <Literal>Addr#</Literal>
938 correspond to null operations at the machine level, but are required
939 to keep the Haskell type checker happy.
943 Operations for indexing off of C pointers
944 (<Literal>Addr#</Literal>s) to snatch values are listed under
945 “arrays”.
951 <Title>Arrays</Title>
954 <IndexTerm><Primary>arrays, primitive</Primary></IndexTerm>
958 The type <Literal>Array# elt</Literal> is the type of primitive,
959 unpointed arrays of values of type <Literal>elt</Literal>.
968 <IndexTerm><Primary><literal>Array#</literal></Primary></IndexTerm>
972 <Literal>Array#</Literal> is more primitive than a Haskell
973 array—indeed, the Haskell <Literal>Array</Literal> interface is
974 implemented using <Literal>Array#</Literal>—in that an
975 <Literal>Array#</Literal> is indexed only by
976 <Literal>Int#</Literal>s, starting at zero. It is also more
977 primitive by virtue of being unboxed. That doesn't mean that it isn't
978 a heap-allocated object—of course, it is. Rather, being unboxed
979 means that it is represented by a pointer to the array itself, and not
980 to a thunk which will evaluate to the array (or to bottom). The
981 components of an <Literal>Array#</Literal> are themselves boxed.
985 The type <Literal>ByteArray#</Literal> is similar to
986 <Literal>Array#</Literal>, except that it contains just a string
987 of (non-pointer) bytes.
996 <IndexTerm><Primary><literal>ByteArray#</literal></Primary></IndexTerm>
1000 Arrays of these types are useful when a Haskell program wishes to
1001 construct a value to pass to a C procedure. It is also possible to use
1002 them to build (say) arrays of unboxed characters for internal use in a
1003 Haskell program. Given these uses, <Literal>ByteArray#</Literal>
1004 is deliberately a bit vague about the type of its components.
1005 Operations are provided to extract values of type
1006 <Literal>Char#</Literal>, <Literal>Int#</Literal>,
1007 <Literal>Float#</Literal>, <Literal>Double#</Literal>, and
1008 <Literal>Addr#</Literal> from arbitrary offsets within a
1009 <Literal>ByteArray#</Literal>. (For type
1010 <Literal>Foo#</Literal>, the $i$th offset gets you the $i$th
1011 <Literal>Foo#</Literal>, not the <Literal>Foo#</Literal> at
1012 byte-position $i$. Mumble.) (If you want a
1013 <Literal>Word#</Literal>, grab an <Literal>Int#</Literal>,
1018 Lastly, we have static byte-arrays, of type
1019 <Literal>Addr#</Literal> [mentioned previously]. (Remember
1020 the duality between arrays and pointers in C.) Arrays of this types
1021 are represented by a pointer to an array in the world outside Haskell,
1022 so this pointer is not followed by the garbage collector. In other
1023 respects they are just like <Literal>ByteArray#</Literal>. They
1024 are only needed in order to pass values from C to Haskell.
1030 <Title>Reading and writing</Title>
1033 Primitive arrays are linear, and indexed starting at zero.
1037 The size and indices of a <Literal>ByteArray#</Literal>, <Literal>Addr#</Literal>, and
1038 <Literal>MutableByteArray#</Literal> are all in bytes. It's up to the program to
1039 calculate the correct byte offset from the start of the array. This
1040 allows a <Literal>ByteArray#</Literal> to contain a mixture of values of different
1041 type, which is often needed when preparing data for and unpicking
1042 results from C. (Umm…not true of indices…WDP 95/09)
1046 <Emphasis>Should we provide some <Literal>sizeOfDouble#</Literal> constants?</Emphasis>
1050 Out-of-range errors on indexing should be caught by the code which
1051 uses the primitive operation; the primitive operations themselves do
1052 <Emphasis>not</Emphasis> check for out-of-range indexes. The intention is that the
1053 primitive ops compile to one machine instruction or thereabouts.
1057 We use the terms “reading” and “writing” to refer to accessing
1058 <Emphasis>mutable</Emphasis> arrays (see <XRef LinkEnd="sect-mutable">), and
1059 “indexing” to refer to reading a value from an <Emphasis>immutable</Emphasis>
1064 Immutable byte arrays are straightforward to index (all indices in bytes):
1067 indexCharArray# :: ByteArray# -> Int# -> Char#
1068 indexIntArray# :: ByteArray# -> Int# -> Int#
1069 indexAddrArray# :: ByteArray# -> Int# -> Addr#
1070 indexFloatArray# :: ByteArray# -> Int# -> Float#
1071 indexDoubleArray# :: ByteArray# -> Int# -> Double#
1073 indexCharOffAddr# :: Addr# -> Int# -> Char#
1074 indexIntOffAddr# :: Addr# -> Int# -> Int#
1075 indexFloatOffAddr# :: Addr# -> Int# -> Float#
1076 indexDoubleOffAddr# :: Addr# -> Int# -> Double#
1077 indexAddrOffAddr# :: Addr# -> Int# -> Addr#
1078 -- Get an Addr# from an Addr# offset
1081 <IndexTerm><Primary><literal>indexCharArray#</literal></Primary></IndexTerm>
1082 <IndexTerm><Primary><literal>indexIntArray#</literal></Primary></IndexTerm>
1083 <IndexTerm><Primary><literal>indexAddrArray#</literal></Primary></IndexTerm>
1084 <IndexTerm><Primary><literal>indexFloatArray#</literal></Primary></IndexTerm>
1085 <IndexTerm><Primary><literal>indexDoubleArray#</literal></Primary></IndexTerm>
1086 <IndexTerm><Primary><literal>indexCharOffAddr#</literal></Primary></IndexTerm>
1087 <IndexTerm><Primary><literal>indexIntOffAddr#</literal></Primary></IndexTerm>
1088 <IndexTerm><Primary><literal>indexFloatOffAddr#</literal></Primary></IndexTerm>
1089 <IndexTerm><Primary><literal>indexDoubleOffAddr#</literal></Primary></IndexTerm>
1090 <IndexTerm><Primary><literal>indexAddrOffAddr#</literal></Primary></IndexTerm>
1094 The last of these, <Function>indexAddrOffAddr#</Function>, extracts an <Literal>Addr#</Literal> using an offset
1095 from another <Literal>Addr#</Literal>, thereby providing the ability to follow a chain of
1100 Something a bit more interesting goes on when indexing arrays of boxed
1101 objects, because the result is simply the boxed object. So presumably
1102 it should be entered—we never usually return an unevaluated
1103 object! This is a pain: primitive ops aren't supposed to do
1104 complicated things like enter objects. The current solution is to
1105 return a single element unboxed tuple (see <XRef LinkEnd="unboxed-tuples">).
1111 indexArray# :: Array# elt -> Int# -> (# elt #)
1114 <IndexTerm><Primary><literal>indexArray#</literal></Primary></IndexTerm>
1120 <Title>The state type</Title>
1123 <IndexTerm><Primary><literal>state, primitive type</literal></Primary></IndexTerm>
1124 <IndexTerm><Primary><literal>State#</literal></Primary></IndexTerm>
1128 The primitive type <Literal>State#</Literal> represents the state of a state
1129 transformer. It is parameterised on the desired type of state, which
1130 serves to keep states from distinct threads distinct from one another.
1131 But the <Emphasis>only</Emphasis> effect of this parameterisation is in the type
1132 system: all values of type <Literal>State#</Literal> are represented in the same way.
1133 Indeed, they are all represented by nothing at all! The code
1134 generator “knows” to generate no code, and allocate no registers
1135 etc, for primitive states.
1147 The type <Literal>GHC.RealWorld</Literal> is truly opaque: there are no values defined
1148 of this type, and no operations over it. It is “primitive” in that
1149 sense - but it is <Emphasis>not unlifted!</Emphasis> Its only role in life is to be
1150 the type which distinguishes the <Literal>IO</Literal> state transformer.
1164 <Title>State of the world</Title>
1167 A single, primitive, value of type <Literal>State# RealWorld</Literal> is provided.
1173 realWorld# :: State# RealWorld
1176 <IndexTerm><Primary>realWorld# state object</Primary></IndexTerm>
1180 (Note: in the compiler, not a <Literal>PrimOp</Literal>; just a mucho magic
1181 <Literal>Id</Literal>. Exported from <Literal>GHC</Literal>, though).
1186 <Sect2 id="sect-mutable">
1187 <Title>Mutable arrays</Title>
1190 <IndexTerm><Primary>mutable arrays</Primary></IndexTerm>
1191 <IndexTerm><Primary>arrays, mutable</Primary></IndexTerm>
1192 Corresponding to <Literal>Array#</Literal> and <Literal>ByteArray#</Literal>, we have the types of
1193 mutable versions of each. In each case, the representation is a
1194 pointer to a suitable block of (mutable) heap-allocated storage.
1200 type MutableArray# s elt
1201 type MutableByteArray# s
1204 <IndexTerm><Primary><literal>MutableArray#</literal></Primary></IndexTerm>
1205 <IndexTerm><Primary><literal>MutableByteArray#</literal></Primary></IndexTerm>
1209 <Title>Allocation</Title>
1212 <IndexTerm><Primary>mutable arrays, allocation</Primary></IndexTerm>
1213 <IndexTerm><Primary>arrays, allocation</Primary></IndexTerm>
1214 <IndexTerm><Primary>allocation, of mutable arrays</Primary></IndexTerm>
1218 Mutable arrays can be allocated. Only pointer-arrays are initialised;
1219 arrays of non-pointers are filled in by “user code” rather than by
1220 the array-allocation primitive. Reason: only the pointer case has to
1221 worry about GC striking with a partly-initialised array.
1227 newArray# :: Int# -> elt -> State# s -> (# State# s, MutableArray# s elt #)
1229 newCharArray# :: Int# -> State# s -> (# State# s, MutableByteArray# s elt #)
1230 newIntArray# :: Int# -> State# s -> (# State# s, MutableByteArray# s elt #)
1231 newAddrArray# :: Int# -> State# s -> (# State# s, MutableByteArray# s elt #)
1232 newFloatArray# :: Int# -> State# s -> (# State# s, MutableByteArray# s elt #)
1233 newDoubleArray# :: Int# -> State# s -> (# State# s, MutableByteArray# s elt #)
1236 <IndexTerm><Primary><literal>newArray#</literal></Primary></IndexTerm>
1237 <IndexTerm><Primary><literal>newCharArray#</literal></Primary></IndexTerm>
1238 <IndexTerm><Primary><literal>newIntArray#</literal></Primary></IndexTerm>
1239 <IndexTerm><Primary><literal>newAddrArray#</literal></Primary></IndexTerm>
1240 <IndexTerm><Primary><literal>newFloatArray#</literal></Primary></IndexTerm>
1241 <IndexTerm><Primary><literal>newDoubleArray#</literal></Primary></IndexTerm>
1245 The size of a <Literal>ByteArray#</Literal> is given in bytes.
1251 <Title>Reading and writing</Title>
1254 <IndexTerm><Primary>arrays, reading and writing</Primary></IndexTerm>
1260 readArray# :: MutableArray# s elt -> Int# -> State# s -> (# State# s, elt #)
1261 readCharArray# :: MutableByteArray# s -> Int# -> State# s -> (# State# s, Char# #)
1262 readIntArray# :: MutableByteArray# s -> Int# -> State# s -> (# State# s, Int# #)
1263 readAddrArray# :: MutableByteArray# s -> Int# -> State# s -> (# State# s, Addr# #)
1264 readFloatArray# :: MutableByteArray# s -> Int# -> State# s -> (# State# s, Float# #)
1265 readDoubleArray# :: MutableByteArray# s -> Int# -> State# s -> (# State# s, Double# #)
1267 writeArray# :: MutableArray# s elt -> Int# -> elt -> State# s -> State# s
1268 writeCharArray# :: MutableByteArray# s -> Int# -> Char# -> State# s -> State# s
1269 writeIntArray# :: MutableByteArray# s -> Int# -> Int# -> State# s -> State# s
1270 writeAddrArray# :: MutableByteArray# s -> Int# -> Addr# -> State# s -> State# s
1271 writeFloatArray# :: MutableByteArray# s -> Int# -> Float# -> State# s -> State# s
1272 writeDoubleArray# :: MutableByteArray# s -> Int# -> Double# -> State# s -> State# s
1275 <IndexTerm><Primary><literal>readArray#</literal></Primary></IndexTerm>
1276 <IndexTerm><Primary><literal>readCharArray#</literal></Primary></IndexTerm>
1277 <IndexTerm><Primary><literal>readIntArray#</literal></Primary></IndexTerm>
1278 <IndexTerm><Primary><literal>readAddrArray#</literal></Primary></IndexTerm>
1279 <IndexTerm><Primary><literal>readFloatArray#</literal></Primary></IndexTerm>
1280 <IndexTerm><Primary><literal>readDoubleArray#</literal></Primary></IndexTerm>
1281 <IndexTerm><Primary><literal>writeArray#</literal></Primary></IndexTerm>
1282 <IndexTerm><Primary><literal>writeCharArray#</literal></Primary></IndexTerm>
1283 <IndexTerm><Primary><literal>writeIntArray#</literal></Primary></IndexTerm>
1284 <IndexTerm><Primary><literal>writeAddrArray#</literal></Primary></IndexTerm>
1285 <IndexTerm><Primary><literal>writeFloatArray#</literal></Primary></IndexTerm>
1286 <IndexTerm><Primary><literal>writeDoubleArray#</literal></Primary></IndexTerm>
1292 <Title>Equality</Title>
1295 <IndexTerm><Primary>arrays, testing for equality</Primary></IndexTerm>
1299 One can take “equality” of mutable arrays. What is compared is the
1300 <Emphasis>name</Emphasis> or reference to the mutable array, not its contents.
1306 sameMutableArray# :: MutableArray# s elt -> MutableArray# s elt -> Bool
1307 sameMutableByteArray# :: MutableByteArray# s -> MutableByteArray# s -> Bool
1310 <IndexTerm><Primary><literal>sameMutableArray#</literal></Primary></IndexTerm>
1311 <IndexTerm><Primary><literal>sameMutableByteArray#</literal></Primary></IndexTerm>
1317 <Title>Freezing mutable arrays</Title>
1320 <IndexTerm><Primary>arrays, freezing mutable</Primary></IndexTerm>
1321 <IndexTerm><Primary>freezing mutable arrays</Primary></IndexTerm>
1322 <IndexTerm><Primary>mutable arrays, freezing</Primary></IndexTerm>
1326 Only unsafe-freeze has a primitive. (Safe freeze is done directly in Haskell
1327 by copying the array and then using <Function>unsafeFreeze</Function>.)
1333 unsafeFreezeArray# :: MutableArray# s elt -> State# s -> (# State# s, Array# s elt #)
1334 unsafeFreezeByteArray# :: MutableByteArray# s -> State# s -> (# State# s, ByteArray# #)
1337 <IndexTerm><Primary><literal>unsafeFreezeArray#</literal></Primary></IndexTerm>
1338 <IndexTerm><Primary><literal>unsafeFreezeByteArray#</literal></Primary></IndexTerm>
1346 <Title>Synchronizing variables (M-vars)</Title>
1349 <IndexTerm><Primary>synchronising variables (M-vars)</Primary></IndexTerm>
1350 <IndexTerm><Primary>M-Vars</Primary></IndexTerm>
1354 Synchronising variables are the primitive type used to implement
1355 Concurrent Haskell's MVars (see the Concurrent Haskell paper for
1356 the operational behaviour of these operations).
1362 type MVar# s elt -- primitive
1364 newMVar# :: State# s -> (# State# s, MVar# s elt #)
1365 takeMVar# :: SynchVar# s elt -> State# s -> (# State# s, elt #)
1366 putMVar# :: SynchVar# s elt -> State# s -> State# s
1369 <IndexTerm><Primary><literal>SynchVar#</literal></Primary></IndexTerm>
1370 <IndexTerm><Primary><literal>newSynchVar#</literal></Primary></IndexTerm>
1371 <IndexTerm><Primary><literal>takeMVar</literal></Primary></IndexTerm>
1372 <IndexTerm><Primary><literal>putMVar</literal></Primary></IndexTerm>
1379 <Sect1 id="glasgow-ST-monad">
1380 <Title>Primitive state-transformer monad
1384 <IndexTerm><Primary>state transformers (Glasgow extensions)</Primary></IndexTerm>
1385 <IndexTerm><Primary>ST monad (Glasgow extension)</Primary></IndexTerm>
1389 This monad underlies our implementation of arrays, mutable and
1390 immutable, and our implementation of I/O, including “C calls”.
1394 The <Literal>ST</Literal> library, which provides access to the
1395 <Function>ST</Function> monad, is described in <xref
1401 <Sect1 id="glasgow-prim-arrays">
1402 <Title>Primitive arrays, mutable and otherwise
1406 <IndexTerm><Primary>primitive arrays (Glasgow extension)</Primary></IndexTerm>
1407 <IndexTerm><Primary>arrays, primitive (Glasgow extension)</Primary></IndexTerm>
1411 GHC knows about quite a few flavours of Large Swathes of Bytes.
1415 First, GHC distinguishes between primitive arrays of (boxed) Haskell
1416 objects (type <Literal>Array# obj</Literal>) and primitive arrays of bytes (type
1417 <Literal>ByteArray#</Literal>).
1421 Second, it distinguishes between…
1425 <Term>Immutable:</Term>
1428 Arrays that do not change (as with “standard” Haskell arrays); you
1429 can only read from them. Obviously, they do not need the care and
1430 attention of the state-transformer monad.
1435 <Term>Mutable:</Term>
1438 Arrays that may be changed or “mutated.” All the operations on them
1439 live within the state-transformer monad and the updates happen
1440 <Emphasis>in-place</Emphasis>.
1445 <Term>“Static” (in C land):</Term>
1448 A C routine may pass an <Literal>Addr#</Literal> pointer back into Haskell land. There
1449 are then primitive operations with which you may merrily grab values
1450 over in C land, by indexing off the “static” pointer.
1455 <Term>“Stable” pointers:</Term>
1458 If, for some reason, you wish to hand a Haskell pointer (i.e.,
1459 <Emphasis>not</Emphasis> an unboxed value) to a C routine, you first make the
1460 pointer “stable,” so that the garbage collector won't forget that it
1461 exists. That is, GHC provides a safe way to pass Haskell pointers to
1466 Please see <XRef LinkEnd="sec-stable-pointers"> for more details.
1471 <Term>“Foreign objects”:</Term>
1474 A “foreign object” is a safe way to pass an external object (a
1475 C-allocated pointer, say) to Haskell and have Haskell do the Right
1476 Thing when it no longer references the object. So, for example, C
1477 could pass a large bitmap over to Haskell and say “please free this
1478 memory when you're done with it.”
1482 Please see <XRef LinkEnd="sec-ForeignObj"> for more details.
1490 The libraries documentatation gives more details on all these
1491 “primitive array” types and the operations on them.
1497 <Sect1 id="pattern-guards">
1498 <Title>Pattern guards</Title>
1501 <IndexTerm><Primary>Pattern guards (Glasgow extension)</Primary></IndexTerm>
1502 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.)
1506 Suppose we have an abstract data type of finite maps, with a
1510 lookup :: FiniteMap -> Int -> Maybe Int
1513 The lookup returns <Function>Nothing</Function> if the supplied key is not in the domain of the mapping, and <Function>(Just v)</Function> otherwise,
1514 where <VarName>v</VarName> is the value that the key maps to. Now consider the following definition:
1518 clunky env var1 var2 | ok1 && ok2 = val1 + val2
1519 | otherwise = var1 + var2
1521 m1 = lookup env var1
1522 m2 = lookup env var2
1523 ok1 = maybeToBool m1
1524 ok2 = maybeToBool m2
1525 val1 = expectJust m1
1526 val2 = expectJust m2
1530 The auxiliary functions are
1534 maybeToBool :: Maybe a -> Bool
1535 maybeToBool (Just x) = True
1536 maybeToBool Nothing = False
1538 expectJust :: Maybe a -> a
1539 expectJust (Just x) = x
1540 expectJust Nothing = error "Unexpected Nothing"
1544 What is <Function>clunky</Function> doing? The guard <Literal>ok1 &&
1545 ok2</Literal> checks that both lookups succeed, using
1546 <Function>maybeToBool</Function> to convert the <Function>Maybe</Function>
1547 types to booleans. The (lazily evaluated) <Function>expectJust</Function>
1548 calls extract the values from the results of the lookups, and binds the
1549 returned values to <VarName>val1</VarName> and <VarName>val2</VarName>
1550 respectively. If either lookup fails, then clunky takes the
1551 <Literal>otherwise</Literal> case and returns the sum of its arguments.
1555 This is certainly legal Haskell, but it is a tremendously verbose and
1556 un-obvious way to achieve the desired effect. Arguably, a more direct way
1557 to write clunky would be to use case expressions:
1561 clunky env var1 var1 = case lookup env var1 of
1563 Just val1 -> case lookup env var2 of
1565 Just val2 -> val1 + val2
1571 This is a bit shorter, but hardly better. Of course, we can rewrite any set
1572 of pattern-matching, guarded equations as case expressions; that is
1573 precisely what the compiler does when compiling equations! The reason that
1574 Haskell provides guarded equations is because they allow us to write down
1575 the cases we want to consider, one at a time, independently of each other.
1576 This structure is hidden in the case version. Two of the right-hand sides
1577 are really the same (<Function>fail</Function>), and the whole expression
1578 tends to become more and more indented.
1582 Here is how I would write clunky:
1586 clunky env var1 var1
1587 | Just val1 <- lookup env var1
1588 , Just val2 <- lookup env var2
1590 ...other equations for clunky...
1594 The semantics should be clear enough. The qualifers are matched in order.
1595 For a <Literal><-</Literal> qualifier, which I call a pattern guard, the
1596 right hand side is evaluated and matched against the pattern on the left.
1597 If the match fails then the whole guard fails and the next equation is
1598 tried. If it succeeds, then the appropriate binding takes place, and the
1599 next qualifier is matched, in the augmented environment. Unlike list
1600 comprehensions, however, the type of the expression to the right of the
1601 <Literal><-</Literal> is the same as the type of the pattern to its
1602 left. The bindings introduced by pattern guards scope over all the
1603 remaining guard qualifiers, and over the right hand side of the equation.
1607 Just as with list comprehensions, boolean expressions can be freely mixed
1608 with among the pattern guards. For example:
1619 Haskell's current guards therefore emerge as a special case, in which the
1620 qualifier list has just one element, a boolean expression.
1624 <sect1 id="sec-ffi">
1625 <title>The foreign interface</title>
1627 <para>The foreign interface consists of the following components:</para>
1631 <para>The Foreign Function Interface language specification
1632 (included in this manual, in <xref linkend="ffi">).</para>
1636 <para>The <literal>Foreign</literal> module (see <xref
1637 linkend="sec-Foreign">) collects together several interfaces
1638 which are useful in specifying foreign language
1639 interfaces, including the following:</para>
1643 <para>The <literal>ForeignObj</literal> module (see <xref
1644 linkend="sec-ForeignObj">), for managing pointers from
1645 Haskell into the outside world.</para>
1649 <para>The <literal>StablePtr</literal> module (see <xref
1650 linkend="sec-stable-pointers">), for managing pointers
1651 into Haskell from the outside world.</para>
1655 <para>The <literal>CTypes</literal> module (see <xref
1656 linkend="sec-CTypes">) gives Haskell equivalents for the
1657 standard C datatypes, for use in making Haskell bindings
1658 to existing C libraries.</para>
1662 <para>The <literal>CTypesISO</literal> module (see <xref
1663 linkend="sec-CTypesISO">) gives Haskell equivalents for C
1664 types defined by the ISO C standard.</para>
1668 <para>The <literal>Storable</literal> library, for
1669 primitive marshalling of data types between Haskell and
1670 the foreign language.</para>
1677 <para>The following sections also give some hints and tips on the use
1678 of the foreign function interface in GHC.</para>
1680 <Sect2 id="glasgow-foreign-headers">
1681 <Title>Using function headers
1685 <IndexTerm><Primary>C calls, function headers</Primary></IndexTerm>
1689 When generating C (using the <Option>-fvia-C</Option> directive), one can assist the
1690 C compiler in detecting type errors by using the <Command>-#include</Command> directive
1691 to provide <Filename>.h</Filename> files containing function headers.
1703 void initialiseEFS (HsInt size);
1704 HsInt terminateEFS (void);
1705 HsForeignObj emptyEFS(void);
1706 HsForeignObj updateEFS (HsForeignObj a, HsInt i, HsInt x);
1707 HsInt lookupEFS (HsForeignObj a, HsInt i);
1711 <para>The types <literal>HsInt</literal>,
1712 <literal>HsForeignObj</literal> etc. are described in <xref
1713 linkend="sec-mapping-table">.</Para>
1715 <Para>Note that this approach is only
1716 <Emphasis>essential</Emphasis> for returning
1717 <Literal>float</Literal>s (or if <Literal>sizeof(int) !=
1718 sizeof(int *)</Literal> on your architecture) but is a Good
1719 Thing for anyone who cares about writing solid code. You're
1720 crazy not to do it.</Para>
1726 <Sect1 id="multi-param-type-classes">
1727 <Title>Multi-parameter type classes
1731 This section documents GHC's implementation of multi-parameter type
1732 classes. There's lots of background in the paper <ULink
1733 URL="http://research.microsoft.com/~simonpj/multi.ps.gz" >Type
1734 classes: exploring the design space</ULink > (Simon Peyton Jones, Mark
1735 Jones, Erik Meijer).
1739 I'd like to thank people who reported shorcomings in the GHC 3.02
1740 implementation. Our default decisions were all conservative ones, and
1741 the experience of these heroic pioneers has given useful concrete
1742 examples to support several generalisations. (These appear below as
1743 design choices not implemented in 3.02.)
1747 I've discussed these notes with Mark Jones, and I believe that Hugs
1748 will migrate towards the same design choices as I outline here.
1749 Thanks to him, and to many others who have offered very useful
1754 <Title>Types</Title>
1757 There are the following restrictions on the form of a qualified
1764 forall tv1..tvn (c1, ...,cn) => type
1770 (Here, I write the "foralls" explicitly, although the Haskell source
1771 language omits them; in Haskell 1.4, all the free type variables of an
1772 explicit source-language type signature are universally quantified,
1773 except for the class type variables in a class declaration. However,
1774 in GHC, you can give the foralls if you want. See <XRef LinkEnd="universal-quantification">).
1783 <Emphasis>Each universally quantified type variable
1784 <Literal>tvi</Literal> must be mentioned (i.e. appear free) in <Literal>type</Literal></Emphasis>.
1786 The reason for this is that a value with a type that does not obey
1787 this restriction could not be used without introducing
1788 ambiguity. Here, for example, is an illegal type:
1792 forall a. Eq a => Int
1796 When a value with this type was used, the constraint <Literal>Eq tv</Literal>
1797 would be introduced where <Literal>tv</Literal> is a fresh type variable, and
1798 (in the dictionary-translation implementation) the value would be
1799 applied to a dictionary for <Literal>Eq tv</Literal>. The difficulty is that we
1800 can never know which instance of <Literal>Eq</Literal> to use because we never
1801 get any more information about <Literal>tv</Literal>.
1808 <Emphasis>Every constraint <Literal>ci</Literal> must mention at least one of the
1809 universally quantified type variables <Literal>tvi</Literal></Emphasis>.
1811 For example, this type is OK because <Literal>C a b</Literal> mentions the
1812 universally quantified type variable <Literal>b</Literal>:
1816 forall a. C a b => burble
1820 The next type is illegal because the constraint <Literal>Eq b</Literal> does not
1821 mention <Literal>a</Literal>:
1825 forall a. Eq b => burble
1829 The reason for this restriction is milder than the other one. The
1830 excluded types are never useful or necessary (because the offending
1831 context doesn't need to be witnessed at this point; it can be floated
1832 out). Furthermore, floating them out increases sharing. Lastly,
1833 excluding them is a conservative choice; it leaves a patch of
1834 territory free in case we need it later.
1844 These restrictions apply to all types, whether declared in a type signature
1849 Unlike Haskell 1.4, constraints in types do <Emphasis>not</Emphasis> have to be of
1850 the form <Emphasis>(class type-variables)</Emphasis>. Thus, these type signatures
1857 f :: Eq (m a) => [m a] -> [m a]
1864 This choice recovers principal types, a property that Haskell 1.4 does not have.
1870 <Title>Class declarations</Title>
1878 <Emphasis>Multi-parameter type classes are permitted</Emphasis>. For example:
1882 class Collection c a where
1883 union :: c a -> c a -> c a
1894 <Emphasis>The class hierarchy must be acyclic</Emphasis>. However, the definition
1895 of "acyclic" involves only the superclass relationships. For example,
1901 op :: D b => a -> b -> b
1904 class C a => D a where { ... }
1908 Here, <Literal>C</Literal> is a superclass of <Literal>D</Literal>, but it's OK for a
1909 class operation <Literal>op</Literal> of <Literal>C</Literal> to mention <Literal>D</Literal>. (It
1910 would not be OK for <Literal>D</Literal> to be a superclass of <Literal>C</Literal>.)
1917 <Emphasis>There are no restrictions on the context in a class declaration
1918 (which introduces superclasses), except that the class hierarchy must
1919 be acyclic</Emphasis>. So these class declarations are OK:
1923 class Functor (m k) => FiniteMap m k where
1926 class (Monad m, Monad (t m)) => Transform t m where
1927 lift :: m a -> (t m) a
1936 <Emphasis>In the signature of a class operation, every constraint
1937 must mention at least one type variable that is not a class type
1938 variable</Emphasis>.
1944 class Collection c a where
1945 mapC :: Collection c b => (a->b) -> c a -> c b
1949 is OK because the constraint <Literal>(Collection a b)</Literal> mentions
1950 <Literal>b</Literal>, even though it also mentions the class variable
1951 <Literal>a</Literal>. On the other hand:
1956 op :: Eq a => (a,b) -> (a,b)
1960 is not OK because the constraint <Literal>(Eq a)</Literal> mentions on the class
1961 type variable <Literal>a</Literal>, but not <Literal>b</Literal>. However, any such
1962 example is easily fixed by moving the offending context up to the
1967 class Eq a => C a where
1972 A yet more relaxed rule would allow the context of a class-op signature
1973 to mention only class type variables. However, that conflicts with
1974 Rule 1(b) for types above.
1981 <Emphasis>The type of each class operation must mention <Emphasis>all</Emphasis> of
1982 the class type variables</Emphasis>. For example:
1986 class Coll s a where
1988 insert :: s -> a -> s
1992 is not OK, because the type of <Literal>empty</Literal> doesn't mention
1993 <Literal>a</Literal>. This rule is a consequence of Rule 1(a), above, for
1994 types, and has the same motivation.
1996 Sometimes, offending class declarations exhibit misunderstandings. For
1997 example, <Literal>Coll</Literal> might be rewritten
2001 class Coll s a where
2003 insert :: s a -> a -> s a
2007 which makes the connection between the type of a collection of
2008 <Literal>a</Literal>'s (namely <Literal>(s a)</Literal>) and the element type <Literal>a</Literal>.
2009 Occasionally this really doesn't work, in which case you can split the
2017 class CollE s => Coll s a where
2018 insert :: s -> a -> s
2031 <Sect2 id="instance-decls">
2032 <Title>Instance declarations</Title>
2040 <Emphasis>Instance declarations may not overlap</Emphasis>. The two instance
2045 instance context1 => C type1 where ...
2046 instance context2 => C type2 where ...
2050 "overlap" if <Literal>type1</Literal> and <Literal>type2</Literal> unify
2052 However, if you give the command line option
2053 <Option>-fallow-overlapping-instances</Option><IndexTerm><Primary>-fallow-overlapping-instances
2054 option</Primary></IndexTerm> then two overlapping instance declarations are permitted
2062 EITHER <Literal>type1</Literal> and <Literal>type2</Literal> do not unify
2068 OR <Literal>type2</Literal> is a substitution instance of <Literal>type1</Literal>
2069 (but not identical to <Literal>type1</Literal>)
2082 Notice that these rules
2089 make it clear which instance decl to use
2090 (pick the most specific one that matches)
2097 do not mention the contexts <Literal>context1</Literal>, <Literal>context2</Literal>
2098 Reason: you can pick which instance decl
2099 "matches" based on the type.
2106 Regrettably, GHC doesn't guarantee to detect overlapping instance
2107 declarations if they appear in different modules. GHC can "see" the
2108 instance declarations in the transitive closure of all the modules
2109 imported by the one being compiled, so it can "see" all instance decls
2110 when it is compiling <Literal>Main</Literal>. However, it currently chooses not
2111 to look at ones that can't possibly be of use in the module currently
2112 being compiled, in the interests of efficiency. (Perhaps we should
2113 change that decision, at least for <Literal>Main</Literal>.)
2120 <Emphasis>There are no restrictions on the type in an instance
2121 <Emphasis>head</Emphasis>, except that at least one must not be a type variable</Emphasis>.
2122 The instance "head" is the bit after the "=>" in an instance decl. For
2123 example, these are OK:
2127 instance C Int a where ...
2129 instance D (Int, Int) where ...
2131 instance E [[a]] where ...
2135 Note that instance heads <Emphasis>may</Emphasis> contain repeated type variables.
2136 For example, this is OK:
2140 instance Stateful (ST s) (MutVar s) where ...
2144 The "at least one not a type variable" restriction is to ensure that
2145 context reduction terminates: each reduction step removes one type
2146 constructor. For example, the following would make the type checker
2147 loop if it wasn't excluded:
2151 instance C a => C a where ...
2155 There are two situations in which the rule is a bit of a pain. First,
2156 if one allows overlapping instance declarations then it's quite
2157 convenient to have a "default instance" declaration that applies if
2158 something more specific does not:
2167 Second, sometimes you might want to use the following to get the
2168 effect of a "class synonym":
2172 class (C1 a, C2 a, C3 a) => C a where { }
2174 instance (C1 a, C2 a, C3 a) => C a where { }
2178 This allows you to write shorter signatures:
2190 f :: (C1 a, C2 a, C3 a) => ...
2194 I'm on the lookout for a simple rule that preserves decidability while
2195 allowing these idioms. The experimental flag
2196 <Option>-fallow-undecidable-instances</Option><IndexTerm><Primary>-fallow-undecidable-instances
2197 option</Primary></IndexTerm> lifts this restriction, allowing all the types in an
2198 instance head to be type variables.
2205 <Emphasis>Unlike Haskell 1.4, instance heads may use type
2206 synonyms</Emphasis>. As always, using a type synonym is just shorthand for
2207 writing the RHS of the type synonym definition. For example:
2211 type Point = (Int,Int)
2212 instance C Point where ...
2213 instance C [Point] where ...
2217 is legal. However, if you added
2221 instance C (Int,Int) where ...
2225 as well, then the compiler will complain about the overlapping
2226 (actually, identical) instance declarations. As always, type synonyms
2227 must be fully applied. You cannot, for example, write:
2232 instance Monad P where ...
2236 This design decision is independent of all the others, and easily
2237 reversed, but it makes sense to me.
2244 <Emphasis>The types in an instance-declaration <Emphasis>context</Emphasis> must all
2245 be type variables</Emphasis>. Thus
2249 instance C a b => Eq (a,b) where ...
2257 instance C Int b => Foo b where ...
2261 is not OK. Again, the intent here is to make sure that context
2262 reduction terminates.
2264 Voluminous correspondence on the Haskell mailing list has convinced me
2265 that it's worth experimenting with a more liberal rule. If you use
2266 the flag <Option>-fallow-undecidable-instances</Option> can use arbitrary
2267 types in an instance context. Termination is ensured by having a
2268 fixed-depth recursion stack. If you exceed the stack depth you get a
2269 sort of backtrace, and the opportunity to increase the stack depth
2270 with <Option>-fcontext-stack</Option><Emphasis>N</Emphasis>.
2283 <Sect1 id="implicit-parameters">
2284 <Title>Implicit parameters
2287 <Para> Implicit paramters are implemented as described in
2288 "Implicit parameters: dynamic scoping with static types",
2289 J Lewis, MB Shields, E Meijer, J Launchbury,
2290 27th ACM Symposium on Principles of Programming Languages (POPL'00),
2295 There should be more documentation, but there isn't (yet). Yell if you need it.
2299 <Para> You can't have an implicit parameter in the context of a class or instance
2300 declaration. For example, both these declarations are illegal:
2302 class (?x::Int) => C a where ...
2303 instance (?x::a) => Foo [a] where ...
2305 Reason: exactly which implicit parameter you pick up depends on exactly where
2306 you invoke a function. But the ``invocation'' of instance declarations is done
2307 behind the scenes by the compiler, so it's hard to figure out exactly where it is done.
2308 Easiest thing is to outlaw the offending types.</para>
2316 <Sect1 id="functional-dependencies">
2317 <Title>Functional dependencies
2320 <Para> Functional dependencies are implemented as described by Mark Jones
2321 in "Type Classes with Functional Dependencies", Mark P. Jones,
2322 In Proceedings of the 9th European Symposium on Programming,
2323 ESOP 2000, Berlin, Germany, March 2000, Springer-Verlag LNCS 1782.
2327 There should be more documentation, but there isn't (yet). Yell if you need it.
2332 <Sect1 id="universal-quantification">
2333 <Title>Explicit universal quantification
2337 GHC's type system supports explicit universal quantification in
2338 constructor fields and function arguments. This is useful for things
2339 like defining <Literal>runST</Literal> from the state-thread world.
2340 GHC's syntax for this now agrees with Hugs's, namely:
2346 forall a b. (Ord a, Eq b) => a -> b -> a
2352 The context is, of course, optional. You can't use <Literal>forall</Literal> as
2353 a type variable any more!
2357 Haskell type signatures are implicitly quantified. The <Literal>forall</Literal>
2358 allows us to say exactly what this means. For example:
2376 g :: forall b. (b -> b)
2382 The two are treated identically.
2386 <Title>Universally-quantified data type fields
2390 In a <Literal>data</Literal> or <Literal>newtype</Literal> declaration one can quantify
2391 the types of the constructor arguments. Here are several examples:
2397 data T a = T1 (forall b. b -> b -> b) a
2399 data MonadT m = MkMonad { return :: forall a. a -> m a,
2400 bind :: forall a b. m a -> (a -> m b) -> m b
2403 newtype Swizzle = MkSwizzle (Ord a => [a] -> [a])
2409 The constructors now have so-called <Emphasis>rank 2</Emphasis> polymorphic
2410 types, in which there is a for-all in the argument types.:
2416 T1 :: forall a. (forall b. b -> b -> b) -> a -> T a
2417 MkMonad :: forall m. (forall a. a -> m a)
2418 -> (forall a b. m a -> (a -> m b) -> m b)
2420 MkSwizzle :: (Ord a => [a] -> [a]) -> Swizzle
2426 Notice that you don't need to use a <Literal>forall</Literal> if there's an
2427 explicit context. For example in the first argument of the
2428 constructor <Function>MkSwizzle</Function>, an implicit "<Literal>forall a.</Literal>" is
2429 prefixed to the argument type. The implicit <Literal>forall</Literal>
2430 quantifies all type variables that are not already in scope, and are
2431 mentioned in the type quantified over.
2435 As for type signatures, implicit quantification happens for non-overloaded
2436 types too. So if you write this:
2439 data T a = MkT (Either a b) (b -> b)
2442 it's just as if you had written this:
2445 data T a = MkT (forall b. Either a b) (forall b. b -> b)
2448 That is, since the type variable <Literal>b</Literal> isn't in scope, it's
2449 implicitly universally quantified. (Arguably, it would be better
2450 to <Emphasis>require</Emphasis> explicit quantification on constructor arguments
2451 where that is what is wanted. Feedback welcomed.)
2457 <Title>Construction </Title>
2460 You construct values of types <Literal>T1, MonadT, Swizzle</Literal> by applying
2461 the constructor to suitable values, just as usual. For example,
2467 (T1 (\xy->x) 3) :: T Int
2469 (MkSwizzle sort) :: Swizzle
2470 (MkSwizzle reverse) :: Swizzle
2477 MkMonad r b) :: MonadT Maybe
2483 The type of the argument can, as usual, be more general than the type
2484 required, as <Literal>(MkSwizzle reverse)</Literal> shows. (<Function>reverse</Function>
2485 does not need the <Literal>Ord</Literal> constraint.)
2491 <Title>Pattern matching</Title>
2494 When you use pattern matching, the bound variables may now have
2495 polymorphic types. For example:
2501 f :: T a -> a -> (a, Char)
2502 f (T1 f k) x = (f k x, f 'c' 'd')
2504 g :: (Ord a, Ord b) => Swizzle -> [a] -> (a -> b) -> [b]
2505 g (MkSwizzle s) xs f = s (map f (s xs))
2507 h :: MonadT m -> [m a] -> m [a]
2508 h m [] = return m []
2509 h m (x:xs) = bind m x $ \y ->
2510 bind m (h m xs) $ \ys ->
2517 In the function <Function>h</Function> we use the record selectors <Literal>return</Literal>
2518 and <Literal>bind</Literal> to extract the polymorphic bind and return functions
2519 from the <Literal>MonadT</Literal> data structure, rather than using pattern
2524 You cannot pattern-match against an argument that is polymorphic.
2528 newtype TIM s a = TIM (ST s (Maybe a))
2530 runTIM :: (forall s. TIM s a) -> Maybe a
2531 runTIM (TIM m) = runST m
2537 Here the pattern-match fails, because you can't pattern-match against
2538 an argument of type <Literal>(forall s. TIM s a)</Literal>. Instead you
2539 must bind the variable and pattern match in the right hand side:
2542 runTIM :: (forall s. TIM s a) -> Maybe a
2543 runTIM tm = case tm of { TIM m -> runST m }
2546 The <Literal>tm</Literal> on the right hand side is (invisibly) instantiated, like
2547 any polymorphic value at its occurrence site, and now you can pattern-match
2554 <Title>The partial-application restriction</Title>
2557 There is really only one way in which data structures with polymorphic
2558 components might surprise you: you must not partially apply them.
2559 For example, this is illegal:
2565 map MkSwizzle [sort, reverse]
2571 The restriction is this: <Emphasis>every subexpression of the program must
2572 have a type that has no for-alls, except that in a function
2573 application (f e1…en) the partial applications are not subject to
2574 this rule</Emphasis>. The restriction makes type inference feasible.
2578 In the illegal example, the sub-expression <Literal>MkSwizzle</Literal> has the
2579 polymorphic type <Literal>(Ord b => [b] -> [b]) -> Swizzle</Literal> and is not
2580 a sub-expression of an enclosing application. On the other hand, this
2587 map (T1 (\a b -> a)) [1,2,3]
2593 even though it involves a partial application of <Function>T1</Function>, because
2594 the sub-expression <Literal>T1 (\a b -> a)</Literal> has type <Literal>Int -> T
2601 <Title>Type signatures
2605 Once you have data constructors with universally-quantified fields, or
2606 constants such as <Constant>runST</Constant> that have rank-2 types, it isn't long
2607 before you discover that you need more! Consider:
2613 mkTs f x y = [T1 f x, T1 f y]
2619 <Function>mkTs</Function> is a fuction that constructs some values of type
2620 <Literal>T</Literal>, using some pieces passed to it. The trouble is that since
2621 <Literal>f</Literal> is a function argument, Haskell assumes that it is
2622 monomorphic, so we'll get a type error when applying <Function>T1</Function> to
2623 it. This is a rather silly example, but the problem really bites in
2624 practice. Lots of people trip over the fact that you can't make
2625 "wrappers functions" for <Constant>runST</Constant> for exactly the same reason.
2626 In short, it is impossible to build abstractions around functions with
2631 The solution is fairly clear. We provide the ability to give a rank-2
2632 type signature for <Emphasis>ordinary</Emphasis> functions (not only data
2633 constructors), thus:
2639 mkTs :: (forall b. b -> b -> b) -> a -> [T a]
2640 mkTs f x y = [T1 f x, T1 f y]
2646 This type signature tells the compiler to attribute <Literal>f</Literal> with
2647 the polymorphic type <Literal>(forall b. b -> b -> b)</Literal> when type
2648 checking the body of <Function>mkTs</Function>, so now the application of
2649 <Function>T1</Function> is fine.
2653 There are two restrictions:
2662 You can only define a rank 2 type, specified by the following
2667 rank2type ::= [forall tyvars .] [context =>] funty
2668 funty ::= ([forall tyvars .] [context =>] ty) -> funty
2670 ty ::= ...current Haskell monotype syntax...
2674 Informally, the universal quantification must all be right at the beginning,
2675 or at the top level of a function argument.
2682 There is a restriction on the definition of a function whose
2683 type signature is a rank-2 type: the polymorphic arguments must be
2684 matched on the left hand side of the "<Literal>=</Literal>" sign. You can't
2685 define <Function>mkTs</Function> like this:
2689 mkTs :: (forall b. b -> b -> b) -> a -> [T a]
2690 mkTs = \ f x y -> [T1 f x, T1 f y]
2695 The same partial-application rule applies to ordinary functions with
2696 rank-2 types as applied to data constructors.
2709 <Title>Type synonyms and hoisting
2713 GHC also allows you to write a <Literal>forall</Literal> in a type synonym, thus:
2715 type Discard a = forall b. a -> b -> a
2720 However, it is often convenient to use these sort of synonyms at the right hand
2721 end of an arrow, thus:
2723 type Discard a = forall b. a -> b -> a
2725 g :: Int -> Discard Int
2728 Simply expanding the type synonym would give
2730 g :: Int -> (forall b. Int -> b -> Int)
2732 but GHC "hoists" the <Literal>forall</Literal> to give the isomorphic type
2734 g :: forall b. Int -> Int -> b -> Int
2736 In general, the rule is this: <Emphasis>to determine the type specified by any explicit
2737 user-written type (e.g. in a type signature), GHC expands type synonyms and then repeatedly
2738 performs the transformation:</Emphasis>
2740 <Emphasis>type1</Emphasis> -> forall a. <Emphasis>type2</Emphasis>
2742 forall a. <Emphasis>type1</Emphasis> -> <Emphasis>type2</Emphasis>
2744 (In fact, GHC tries to retain as much synonym information as possible for use in
2745 error messages, but that is a usability issue.) This rule applies, of course, whether
2746 or not the <Literal>forall</Literal> comes from a synonym. For example, here is another
2747 valid way to write <Literal>g</Literal>'s type signature:
2749 g :: Int -> Int -> forall b. b -> Int
2756 <Sect1 id="existential-quantification">
2757 <Title>Existentially quantified data constructors
2761 The idea of using existential quantification in data type declarations
2762 was suggested by Laufer (I believe, thought doubtless someone will
2763 correct me), and implemented in Hope+. It's been in Lennart
2764 Augustsson's <Command>hbc</Command> Haskell compiler for several years, and
2765 proved very useful. Here's the idea. Consider the declaration:
2771 data Foo = forall a. MkFoo a (a -> Bool)
2778 The data type <Literal>Foo</Literal> has two constructors with types:
2784 MkFoo :: forall a. a -> (a -> Bool) -> Foo
2791 Notice that the type variable <Literal>a</Literal> in the type of <Function>MkFoo</Function>
2792 does not appear in the data type itself, which is plain <Literal>Foo</Literal>.
2793 For example, the following expression is fine:
2799 [MkFoo 3 even, MkFoo 'c' isUpper] :: [Foo]
2805 Here, <Literal>(MkFoo 3 even)</Literal> packages an integer with a function
2806 <Function>even</Function> that maps an integer to <Literal>Bool</Literal>; and <Function>MkFoo 'c'
2807 isUpper</Function> packages a character with a compatible function. These
2808 two things are each of type <Literal>Foo</Literal> and can be put in a list.
2812 What can we do with a value of type <Literal>Foo</Literal>?. In particular,
2813 what happens when we pattern-match on <Function>MkFoo</Function>?
2819 f (MkFoo val fn) = ???
2825 Since all we know about <Literal>val</Literal> and <Function>fn</Function> is that they
2826 are compatible, the only (useful) thing we can do with them is to
2827 apply <Function>fn</Function> to <Literal>val</Literal> to get a boolean. For example:
2834 f (MkFoo val fn) = fn val
2840 What this allows us to do is to package heterogenous values
2841 together with a bunch of functions that manipulate them, and then treat
2842 that collection of packages in a uniform manner. You can express
2843 quite a bit of object-oriented-like programming this way.
2846 <Sect2 id="existential">
2847 <Title>Why existential?
2851 What has this to do with <Emphasis>existential</Emphasis> quantification?
2852 Simply that <Function>MkFoo</Function> has the (nearly) isomorphic type
2858 MkFoo :: (exists a . (a, a -> Bool)) -> Foo
2864 But Haskell programmers can safely think of the ordinary
2865 <Emphasis>universally</Emphasis> quantified type given above, thereby avoiding
2866 adding a new existential quantification construct.
2872 <Title>Type classes</Title>
2875 An easy extension (implemented in <Command>hbc</Command>) is to allow
2876 arbitrary contexts before the constructor. For example:
2882 data Baz = forall a. Eq a => Baz1 a a
2883 | forall b. Show b => Baz2 b (b -> b)
2889 The two constructors have the types you'd expect:
2895 Baz1 :: forall a. Eq a => a -> a -> Baz
2896 Baz2 :: forall b. Show b => b -> (b -> b) -> Baz
2902 But when pattern matching on <Function>Baz1</Function> the matched values can be compared
2903 for equality, and when pattern matching on <Function>Baz2</Function> the first matched
2904 value can be converted to a string (as well as applying the function to it).
2905 So this program is legal:
2912 f (Baz1 p q) | p == q = "Yes"
2914 f (Baz1 v fn) = show (fn v)
2920 Operationally, in a dictionary-passing implementation, the
2921 constructors <Function>Baz1</Function> and <Function>Baz2</Function> must store the
2922 dictionaries for <Literal>Eq</Literal> and <Literal>Show</Literal> respectively, and
2923 extract it on pattern matching.
2927 Notice the way that the syntax fits smoothly with that used for
2928 universal quantification earlier.
2934 <Title>Restrictions</Title>
2937 There are several restrictions on the ways in which existentially-quantified
2938 constructors can be use.
2947 When pattern matching, each pattern match introduces a new,
2948 distinct, type for each existential type variable. These types cannot
2949 be unified with any other type, nor can they escape from the scope of
2950 the pattern match. For example, these fragments are incorrect:
2958 Here, the type bound by <Function>MkFoo</Function> "escapes", because <Literal>a</Literal>
2959 is the result of <Function>f1</Function>. One way to see why this is wrong is to
2960 ask what type <Function>f1</Function> has:
2964 f1 :: Foo -> a -- Weird!
2968 What is this "<Literal>a</Literal>" in the result type? Clearly we don't mean
2973 f1 :: forall a. Foo -> a -- Wrong!
2977 The original program is just plain wrong. Here's another sort of error
2981 f2 (Baz1 a b) (Baz1 p q) = a==q
2985 It's ok to say <Literal>a==b</Literal> or <Literal>p==q</Literal>, but
2986 <Literal>a==q</Literal> is wrong because it equates the two distinct types arising
2987 from the two <Function>Baz1</Function> constructors.
2995 You can't pattern-match on an existentially quantified
2996 constructor in a <Literal>let</Literal> or <Literal>where</Literal> group of
2997 bindings. So this is illegal:
3001 f3 x = a==b where { Baz1 a b = x }
3005 You can only pattern-match
3006 on an existentially-quantified constructor in a <Literal>case</Literal> expression or
3007 in the patterns of a function definition.
3009 The reason for this restriction is really an implementation one.
3010 Type-checking binding groups is already a nightmare without
3011 existentials complicating the picture. Also an existential pattern
3012 binding at the top level of a module doesn't make sense, because it's
3013 not clear how to prevent the existentially-quantified type "escaping".
3014 So for now, there's a simple-to-state restriction. We'll see how
3022 You can't use existential quantification for <Literal>newtype</Literal>
3023 declarations. So this is illegal:
3027 newtype T = forall a. Ord a => MkT a
3031 Reason: a value of type <Literal>T</Literal> must be represented as a pair
3032 of a dictionary for <Literal>Ord t</Literal> and a value of type <Literal>t</Literal>.
3033 That contradicts the idea that <Literal>newtype</Literal> should have no
3034 concrete representation. You can get just the same efficiency and effect
3035 by using <Literal>data</Literal> instead of <Literal>newtype</Literal>. If there is no
3036 overloading involved, then there is more of a case for allowing
3037 an existentially-quantified <Literal>newtype</Literal>, because the <Literal>data</Literal>
3038 because the <Literal>data</Literal> version does carry an implementation cost,
3039 but single-field existentially quantified constructors aren't much
3040 use. So the simple restriction (no existential stuff on <Literal>newtype</Literal>)
3041 stands, unless there are convincing reasons to change it.
3049 You can't use <Literal>deriving</Literal> to define instances of a
3050 data type with existentially quantified data constructors.
3052 Reason: in most cases it would not make sense. For example:#
3055 data T = forall a. MkT [a] deriving( Eq )
3058 To derive <Literal>Eq</Literal> in the standard way we would need to have equality
3059 between the single component of two <Function>MkT</Function> constructors:
3063 (MkT a) == (MkT b) = ???
3066 But <VarName>a</VarName> and <VarName>b</VarName> have distinct types, and so can't be compared.
3067 It's just about possible to imagine examples in which the derived instance
3068 would make sense, but it seems altogether simpler simply to prohibit such
3069 declarations. Define your own instances!
3081 <Sect1 id="sec-assertions">
3083 <IndexTerm><Primary>Assertions</Primary></IndexTerm>
3087 If you want to make use of assertions in your standard Haskell code, you
3088 could define a function like the following:
3094 assert :: Bool -> a -> a
3095 assert False x = error "assertion failed!"
3102 which works, but gives you back a less than useful error message --
3103 an assertion failed, but which and where?
3107 One way out is to define an extended <Function>assert</Function> function which also
3108 takes a descriptive string to include in the error message and
3109 perhaps combine this with the use of a pre-processor which inserts
3110 the source location where <Function>assert</Function> was used.
3114 Ghc offers a helping hand here, doing all of this for you. For every
3115 use of <Function>assert</Function> in the user's source:
3121 kelvinToC :: Double -> Double
3122 kelvinToC k = assert (k >= 0.0) (k+273.15)
3128 Ghc will rewrite this to also include the source location where the
3135 assert pred val ==> assertError "Main.hs|15" pred val
3141 The rewrite is only performed by the compiler when it spots
3142 applications of <Function>Exception.assert</Function>, so you can still define and
3143 use your own versions of <Function>assert</Function>, should you so wish. If not,
3144 import <Literal>Exception</Literal> to make use <Function>assert</Function> in your code.
3148 To have the compiler ignore uses of assert, use the compiler option
3149 <Option>-fignore-asserts</Option>. <IndexTerm><Primary>-fignore-asserts option</Primary></IndexTerm> That is,
3150 expressions of the form <Literal>assert pred e</Literal> will be rewritten to <Literal>e</Literal>.
3154 Assertion failures can be caught, see the documentation for the
3155 <literal>Exception</literal> library (<xref linkend="sec-Exception">)
3161 <Sect1 id="scoped-type-variables">
3162 <Title>Scoped Type Variables
3166 A <Emphasis>pattern type signature</Emphasis> can introduce a <Emphasis>scoped type
3167 variable</Emphasis>. For example
3173 f (xs::[a]) = ys ++ ys
3182 The pattern <Literal>(xs::[a])</Literal> includes a type signature for <VarName>xs</VarName>.
3183 This brings the type variable <Literal>a</Literal> into scope; it scopes over
3184 all the patterns and right hand sides for this equation for <Function>f</Function>.
3185 In particular, it is in scope at the type signature for <VarName>y</VarName>.
3189 At ordinary type signatures, such as that for <VarName>ys</VarName>, any type variables
3190 mentioned in the type signature <Emphasis>that are not in scope</Emphasis> are
3191 implicitly universally quantified. (If there are no type variables in
3192 scope, all type variables mentioned in the signature are universally
3193 quantified, which is just as in Haskell 98.) In this case, since <VarName>a</VarName>
3194 is in scope, it is not universally quantified, so the type of <VarName>ys</VarName> is
3195 the same as that of <VarName>xs</VarName>. In Haskell 98 it is not possible to declare
3196 a type for <VarName>ys</VarName>; a major benefit of scoped type variables is that
3197 it becomes possible to do so.
3201 Scoped type variables are implemented in both GHC and Hugs. Where the
3202 implementations differ from the specification below, those differences
3207 So much for the basic idea. Here are the details.
3211 <Title>Scope and implicit quantification</Title>
3219 All the type variables mentioned in the patterns for a single
3220 function definition equation, that are not already in scope,
3221 are brought into scope by the patterns. We describe this set as
3222 the <Emphasis>type variables bound by the equation</Emphasis>.
3229 The type variables thus brought into scope may be mentioned
3230 in ordinary type signatures or pattern type signatures anywhere within
3238 In ordinary type signatures, any type variable mentioned in the
3239 signature that is in scope is <Emphasis>not</Emphasis> universally quantified.
3246 Ordinary type signatures do not bring any new type variables
3247 into scope (except in the type signature itself!). So this is illegal:
3256 It's illegal because <VarName>a</VarName> is not in scope in the body of <Function>f</Function>,
3257 so the ordinary signature <Literal>x::a</Literal> is equivalent to <Literal>x::forall a.a</Literal>;
3258 and that is an incorrect typing.
3265 There is no implicit universal quantification on pattern type
3266 signatures, nor may one write an explicit <Literal>forall</Literal> type in a pattern
3267 type signature. The pattern type signature is a monotype.
3275 The type variables in the head of a <Literal>class</Literal> or <Literal>instance</Literal> declaration
3276 scope over the methods defined in the <Literal>where</Literal> part. For example:
3290 (Not implemented in Hugs yet, Dec 98).
3301 <Title>Polymorphism</Title>
3309 Pattern type signatures are completely orthogonal to ordinary, separate
3310 type signatures. The two can be used independently or together. There is
3311 no scoping associated with the names of the type variables in a separate type signature.
3316 f (xs::[b]) = reverse xs
3325 The function must be polymorphic in the type variables
3326 bound by all its equations. Operationally, the type variables bound
3327 by one equation must not:
3334 Be unified with a type (such as <Literal>Int</Literal>, or <Literal>[a]</Literal>).
3340 Be unified with a type variable free in the environment.
3346 Be unified with each other. (They may unify with the type variables
3347 bound by another equation for the same function, of course.)
3354 For example, the following all fail to type check:
3358 f (x::a) (y::b) = [x,y] -- a unifies with b
3360 g (x::a) = x + 1::Int -- a unifies with Int
3362 h x = let k (y::a) = [x,y] -- a is free in the
3363 in k x -- environment
3365 k (x::a) True = ... -- a unifies with Int
3366 k (x::Int) False = ...
3369 w (x::a) = x -- a unifies with [b]
3378 The pattern-bound type variable may, however, be constrained
3379 by the context of the principal type, thus:
3383 f (x::a) (y::a) = x+y*2
3387 gets the inferred type: <Literal>forall a. Num a => a -> a -> a</Literal>.
3398 <Title>Result type signatures</Title>
3406 The result type of a function can be given a signature,
3411 f (x::a) :: [a] = [x,x,x]
3415 The final <Literal>:: [a]</Literal> after all the patterns gives a signature to the
3416 result type. Sometimes this is the only way of naming the type variable
3421 f :: Int -> [a] -> [a]
3422 f n :: ([a] -> [a]) = let g (x::a, y::a) = (y,x)
3423 in \xs -> map g (reverse xs `zip` xs)
3435 Result type signatures are not yet implemented in Hugs.
3441 <Title>Pattern signatures on other constructs</Title>
3449 A pattern type signature can be on an arbitrary sub-pattern, not
3454 f ((x,y)::(a,b)) = (y,x) :: (b,a)
3463 Pattern type signatures, including the result part, can be used
3464 in lambda abstractions:
3468 (\ (x::a, y) :: a -> x)
3472 Type variables bound by these patterns must be polymorphic in
3473 the sense defined above.
3478 f1 (x::c) = f1 x -- ok
3479 f2 = \(x::c) -> f2 x -- not ok
3483 Here, <Function>f1</Function> is OK, but <Function>f2</Function> is not, because <VarName>c</VarName> gets unified
3484 with a type variable free in the environment, in this
3485 case, the type of <Function>f2</Function>, which is in the environment when
3486 the lambda abstraction is checked.
3493 Pattern type signatures, including the result part, can be used
3494 in <Literal>case</Literal> expressions:
3498 case e of { (x::a, y) :: a -> x }
3502 The pattern-bound type variables must, as usual,
3503 be polymorphic in the following sense: each case alternative,
3504 considered as a lambda abstraction, must be polymorphic.
3509 case (True,False) of { (x::a, y) -> x }
3513 Even though the context is that of a pair of booleans,
3514 the alternative itself is polymorphic. Of course, it is
3519 case (True,False) of { (x::Bool, y) -> x }
3528 To avoid ambiguity, the type after the “<Literal>::</Literal>” in a result
3529 pattern signature on a lambda or <Literal>case</Literal> must be atomic (i.e. a single
3530 token or a parenthesised type of some sort). To see why,
3531 consider how one would parse this:
3544 Pattern type signatures that bind new type variables
3545 may not be used in pattern bindings at all.
3550 f x = let (y, z::a) = x in ...
3554 But these are OK, because they do not bind fresh type variables:
3558 f1 x = let (y, z::Int) = x in ...
3559 f2 (x::(Int,a)) = let (y, z::a) = x in ...
3563 However a single variable is considered a degenerate function binding,
3564 rather than a degerate pattern binding, so this is permitted, even
3565 though it binds a type variable:
3569 f :: (b->b) = \(x::b) -> x
3578 Such degnerate function bindings do not fall under the monomorphism
3585 g :: a -> a -> Bool = \x y. x==y
3591 Here <Function>g</Function> has type <Literal>forall a. Eq a => a -> a -> Bool</Literal>, just as if
3592 <Function>g</Function> had a separate type signature. Lacking a type signature, <Function>g</Function>
3593 would get a monomorphic type.
3599 <Title>Existentials</Title>
3607 Pattern type signatures can bind existential type variables.
3612 data T = forall a. MkT [a]
3615 f (MkT [t::a]) = MkT t3
3632 <Sect1 id="pragmas">
3637 GHC supports several pragmas, or instructions to the compiler placed
3638 in the source code. Pragmas don't affect the meaning of the program,
3639 but they might affect the efficiency of the generated code.
3642 <Sect2 id="inline-pragma">
3643 <Title>INLINE pragma
3645 <IndexTerm><Primary>INLINE pragma</Primary></IndexTerm>
3646 <IndexTerm><Primary>pragma, INLINE</Primary></IndexTerm></Title>
3649 GHC (with <Option>-O</Option>, as always) tries to inline (or “unfold”)
3650 functions/values that are “small enough,” thus avoiding the call
3651 overhead and possibly exposing other more-wonderful optimisations.
3655 You will probably see these unfoldings (in Core syntax) in your
3660 Normally, if GHC decides a function is “too expensive” to inline, it
3661 will not do so, nor will it export that unfolding for other modules to
3666 The sledgehammer you can bring to bear is the
3667 <Literal>INLINE</Literal><IndexTerm><Primary>INLINE pragma</Primary></IndexTerm> pragma, used thusly:
3670 key_function :: Int -> String -> (Bool, Double)
3672 #ifdef __GLASGOW_HASKELL__
3673 {-# INLINE key_function #-}
3677 (You don't need to do the C pre-processor carry-on unless you're going
3678 to stick the code through HBC—it doesn't like <Literal>INLINE</Literal> pragmas.)
3682 The major effect of an <Literal>INLINE</Literal> pragma is to declare a function's
3683 “cost” to be very low. The normal unfolding machinery will then be
3684 very keen to inline it.
3688 An <Literal>INLINE</Literal> pragma for a function can be put anywhere its type
3689 signature could be put.
3693 <Literal>INLINE</Literal> pragmas are a particularly good idea for the
3694 <Literal>then</Literal>/<Literal>return</Literal> (or <Literal>bind</Literal>/<Literal>unit</Literal>) functions in a monad.
3695 For example, in GHC's own <Literal>UniqueSupply</Literal> monad code, we have:
3698 #ifdef __GLASGOW_HASKELL__
3699 {-# INLINE thenUs #-}
3700 {-# INLINE returnUs #-}
3708 <Sect2 id="noinline-pragma">
3709 <Title>NOINLINE pragma
3713 <IndexTerm><Primary>NOINLINE pragma</Primary></IndexTerm>
3714 <IndexTerm><Primary>pragma, NOINLINE</Primary></IndexTerm>
3718 The <Literal>NOINLINE</Literal> pragma does exactly what you'd expect: it stops the
3719 named function from being inlined by the compiler. You shouldn't ever
3720 need to do this, unless you're very cautious about code size.
3725 <Sect2 id="specialize-pragma">
3726 <Title>SPECIALIZE pragma
3730 <IndexTerm><Primary>SPECIALIZE pragma</Primary></IndexTerm>
3731 <IndexTerm><Primary>pragma, SPECIALIZE</Primary></IndexTerm>
3732 <IndexTerm><Primary>overloading, death to</Primary></IndexTerm>
3736 (UK spelling also accepted.) For key overloaded functions, you can
3737 create extra versions (NB: more code space) specialised to particular
3738 types. Thus, if you have an overloaded function:
3744 hammeredLookup :: Ord key => [(key, value)] -> key -> value
3750 If it is heavily used on lists with <Literal>Widget</Literal> keys, you could
3751 specialise it as follows:
3754 {-# SPECIALIZE hammeredLookup :: [(Widget, value)] -> Widget -> value #-}
3760 To get very fancy, you can also specify a named function to use for
3761 the specialised value, by adding <Literal>= blah</Literal>, as in:
3764 {-# SPECIALIZE hammeredLookup :: ...as before... = blah #-}
3767 It's <Emphasis>Your Responsibility</Emphasis> to make sure that <Function>blah</Function> really
3768 behaves as a specialised version of <Function>hammeredLookup</Function>!!!
3772 NOTE: the <Literal>=blah</Literal> feature isn't implemented in GHC 4.xx.
3776 An example in which the <Literal>= blah</Literal> form will Win Big:
3779 toDouble :: Real a => a -> Double
3780 toDouble = fromRational . toRational
3782 {-# SPECIALIZE toDouble :: Int -> Double = i2d #-}
3783 i2d (I# i) = D# (int2Double# i) -- uses Glasgow prim-op directly
3786 The <Function>i2d</Function> function is virtually one machine instruction; the
3787 default conversion—via an intermediate <Literal>Rational</Literal>—is obscenely
3788 expensive by comparison.
3792 By using the US spelling, your <Literal>SPECIALIZE</Literal> pragma will work with
3793 HBC, too. Note that HBC doesn't support the <Literal>= blah</Literal> form.
3797 A <Literal>SPECIALIZE</Literal> pragma for a function can be put anywhere its type
3798 signature could be put.
3803 <Sect2 id="specialize-instance-pragma">
3804 <Title>SPECIALIZE instance pragma
3808 <IndexTerm><Primary>SPECIALIZE pragma</Primary></IndexTerm>
3809 <IndexTerm><Primary>overloading, death to</Primary></IndexTerm>
3810 Same idea, except for instance declarations. For example:
3813 instance (Eq a) => Eq (Foo a) where { ... usual stuff ... }
3815 {-# SPECIALIZE instance Eq (Foo [(Int, Bar)] #-}
3818 Compatible with HBC, by the way.
3823 <Sect2 id="line-pragma">
3828 <IndexTerm><Primary>LINE pragma</Primary></IndexTerm>
3829 <IndexTerm><Primary>pragma, LINE</Primary></IndexTerm>
3833 This pragma is similar to C's <Literal>#line</Literal> pragma, and is mainly for use in
3834 automatically generated Haskell code. It lets you specify the line
3835 number and filename of the original code; for example
3841 {-# LINE 42 "Foo.vhs" #-}
3847 if you'd generated the current file from something called <Filename>Foo.vhs</Filename>
3848 and this line corresponds to line 42 in the original. GHC will adjust
3849 its error messages to refer to the line/file named in the <Literal>LINE</Literal>
3856 <Title>RULES pragma</Title>
3859 The RULES pragma lets you specify rewrite rules. It is described in
3860 <XRef LinkEnd="rewrite-rules">.
3867 <Sect1 id="rewrite-rules">
3868 <Title>Rewrite rules
3870 <IndexTerm><Primary>RULES pagma</Primary></IndexTerm>
3871 <IndexTerm><Primary>pragma, RULES</Primary></IndexTerm>
3872 <IndexTerm><Primary>rewrite rules</Primary></IndexTerm></Title>
3875 The programmer can specify rewrite rules as part of the source program
3876 (in a pragma). GHC applies these rewrite rules wherever it can.
3884 "map/map" forall f g xs. map f (map g xs) = map (f.g) xs
3891 <Title>Syntax</Title>
3894 From a syntactic point of view:
3900 Each rule has a name, enclosed in double quotes. The name itself has
3901 no significance at all. It is only used when reporting how many times the rule fired.
3907 There may be zero or more rules in a <Literal>RULES</Literal> pragma.
3913 Layout applies in a <Literal>RULES</Literal> pragma. Currently no new indentation level
3914 is set, so you must lay out your rules starting in the same column as the
3915 enclosing definitions.
3921 Each variable mentioned in a rule must either be in scope (e.g. <Function>map</Function>),
3922 or bound by the <Literal>forall</Literal> (e.g. <Function>f</Function>, <Function>g</Function>, <Function>xs</Function>). The variables bound by
3923 the <Literal>forall</Literal> are called the <Emphasis>pattern</Emphasis> variables. They are separated
3924 by spaces, just like in a type <Literal>forall</Literal>.
3930 A pattern variable may optionally have a type signature.
3931 If the type of the pattern variable is polymorphic, it <Emphasis>must</Emphasis> have a type signature.
3932 For example, here is the <Literal>foldr/build</Literal> rule:
3935 "fold/build" forall k z (g::forall b. (a->b->b) -> b -> b) .
3936 foldr k z (build g) = g k z
3939 Since <Function>g</Function> has a polymorphic type, it must have a type signature.
3946 The left hand side of a rule must consist of a top-level variable applied
3947 to arbitrary expressions. For example, this is <Emphasis>not</Emphasis> OK:
3950 "wrong1" forall e1 e2. case True of { True -> e1; False -> e2 } = e1
3951 "wrong2" forall f. f True = True
3954 In <Literal>"wrong1"</Literal>, the LHS is not an application; in <Literal>"wrong2"</Literal>, the LHS has a pattern variable
3961 A rule does not need to be in the same module as (any of) the
3962 variables it mentions, though of course they need to be in scope.
3968 Rules are automatically exported from a module, just as instance declarations are.
3979 <Title>Semantics</Title>
3982 From a semantic point of view:
3988 Rules are only applied if you use the <Option>-O</Option> flag.
3994 Rules are regarded as left-to-right rewrite rules.
3995 When GHC finds an expression that is a substitution instance of the LHS
3996 of a rule, it replaces the expression by the (appropriately-substituted) RHS.
3997 By "a substitution instance" we mean that the LHS can be made equal to the
3998 expression by substituting for the pattern variables.
4005 The LHS and RHS of a rule are typechecked, and must have the
4013 GHC makes absolutely no attempt to verify that the LHS and RHS
4014 of a rule have the same meaning. That is undecideable in general, and
4015 infeasible in most interesting cases. The responsibility is entirely the programmer's!
4022 GHC makes no attempt to make sure that the rules are confluent or
4023 terminating. For example:
4026 "loop" forall x,y. f x y = f y x
4029 This rule will cause the compiler to go into an infinite loop.
4036 If more than one rule matches a call, GHC will choose one arbitrarily to apply.
4042 GHC currently uses a very simple, syntactic, matching algorithm
4043 for matching a rule LHS with an expression. It seeks a substitution
4044 which makes the LHS and expression syntactically equal modulo alpha
4045 conversion. The pattern (rule), but not the expression, is eta-expanded if
4046 necessary. (Eta-expanding the epression can lead to laziness bugs.)
4047 But not beta conversion (that's called higher-order matching).
4051 Matching is carried out on GHC's intermediate language, which includes
4052 type abstractions and applications. So a rule only matches if the
4053 types match too. See <XRef LinkEnd="rule-spec"> below.
4059 GHC keeps trying to apply the rules as it optimises the program.
4060 For example, consider:
4069 The expression <Literal>s (t xs)</Literal> does not match the rule <Literal>"map/map"</Literal>, but GHC
4070 will substitute for <VarName>s</VarName> and <VarName>t</VarName>, giving an expression which does match.
4071 If <VarName>s</VarName> or <VarName>t</VarName> was (a) used more than once, and (b) large or a redex, then it would
4072 not be substituted, and the rule would not fire.
4079 In the earlier phases of compilation, GHC inlines <Emphasis>nothing
4080 that appears on the LHS of a rule</Emphasis>, because once you have substituted
4081 for something you can't match against it (given the simple minded
4082 matching). So if you write the rule
4085 "map/map" forall f,g. map f . map g = map (f.g)
4088 this <Emphasis>won't</Emphasis> match the expression <Literal>map f (map g xs)</Literal>.
4089 It will only match something written with explicit use of ".".
4090 Well, not quite. It <Emphasis>will</Emphasis> match the expression
4096 where <Function>wibble</Function> is defined:
4099 wibble f g = map f . map g
4102 because <Function>wibble</Function> will be inlined (it's small).
4104 Later on in compilation, GHC starts inlining even things on the
4105 LHS of rules, but still leaves the rules enabled. This inlining
4106 policy is controlled by the per-simplification-pass flag <Option>-finline-phase</Option><Emphasis>n</Emphasis>.
4113 All rules are implicitly exported from the module, and are therefore
4114 in force in any module that imports the module that defined the rule, directly
4115 or indirectly. (That is, if A imports B, which imports C, then C's rules are
4116 in force when compiling A.) The situation is very similar to that for instance
4128 <Title>List fusion</Title>
4131 The RULES mechanism is used to implement fusion (deforestation) of common list functions.
4132 If a "good consumer" consumes an intermediate list constructed by a "good producer", the
4133 intermediate list should be eliminated entirely.
4137 The following are good producers:
4149 Enumerations of <Literal>Int</Literal> and <Literal>Char</Literal> (e.g. <Literal>['a'..'z']</Literal>).
4155 Explicit lists (e.g. <Literal>[True, False]</Literal>)
4161 The cons constructor (e.g <Literal>3:4:[]</Literal>)
4167 <Function>++</Function>
4173 <Function>map</Function>
4179 <Function>filter</Function>
4185 <Function>iterate</Function>, <Function>repeat</Function>
4191 <Function>zip</Function>, <Function>zipWith</Function>
4200 The following are good consumers:
4212 <Function>array</Function> (on its second argument)
4218 <Function>length</Function>
4224 <Function>++</Function> (on its first argument)
4230 <Function>map</Function>
4236 <Function>filter</Function>
4242 <Function>concat</Function>
4248 <Function>unzip</Function>, <Function>unzip2</Function>, <Function>unzip3</Function>, <Function>unzip4</Function>
4254 <Function>zip</Function>, <Function>zipWith</Function> (but on one argument only; if both are good producers, <Function>zip</Function>
4255 will fuse with one but not the other)
4261 <Function>partition</Function>
4267 <Function>head</Function>
4273 <Function>and</Function>, <Function>or</Function>, <Function>any</Function>, <Function>all</Function>
4279 <Function>sequence_</Function>
4285 <Function>msum</Function>
4291 <Function>sortBy</Function>
4300 So, for example, the following should generate no intermediate lists:
4303 array (1,10) [(i,i*i) | i <- map (+ 1) [0..9]]
4309 This list could readily be extended; if there are Prelude functions that you use
4310 a lot which are not included, please tell us.
4314 If you want to write your own good consumers or producers, look at the
4315 Prelude definitions of the above functions to see how to do so.
4320 <Sect2 id="rule-spec">
4321 <Title>Specialisation
4325 Rewrite rules can be used to get the same effect as a feature
4326 present in earlier version of GHC:
4329 {-# SPECIALIZE fromIntegral :: Int8 -> Int16 = int8ToInt16 #-}
4332 This told GHC to use <Function>int8ToInt16</Function> instead of <Function>fromIntegral</Function> whenever
4333 the latter was called with type <Literal>Int8 -> Int16</Literal>. That is, rather than
4334 specialising the original definition of <Function>fromIntegral</Function> the programmer is
4335 promising that it is safe to use <Function>int8ToInt16</Function> instead.
4339 This feature is no longer in GHC. But rewrite rules let you do the
4344 "fromIntegral/Int8/Int16" fromIntegral = int8ToInt16
4348 This slightly odd-looking rule instructs GHC to replace <Function>fromIntegral</Function>
4349 by <Function>int8ToInt16</Function> <Emphasis>whenever the types match</Emphasis>. Speaking more operationally,
4350 GHC adds the type and dictionary applications to get the typed rule
4353 forall (d1::Integral Int8) (d2::Num Int16) .
4354 fromIntegral Int8 Int16 d1 d2 = int8ToInt16
4358 this rule does not need to be in the same file as fromIntegral,
4359 unlike the <Literal>SPECIALISE</Literal> pragmas which currently do (so that they
4360 have an original definition available to specialise).
4366 <Title>Controlling what's going on</Title>
4374 Use <Option>-ddump-rules</Option> to see what transformation rules GHC is using.
4380 Use <Option>-ddump-simpl-stats</Option> to see what rules are being fired.
4381 If you add <Option>-dppr-debug</Option> you get a more detailed listing.
4387 The defintion of (say) <Function>build</Function> in <FileName>PrelBase.lhs</FileName> looks llike this:
4390 build :: forall a. (forall b. (a -> b -> b) -> b -> b) -> [a]
4391 {-# INLINE build #-}
4395 Notice the <Literal>INLINE</Literal>! That prevents <Literal>(:)</Literal> from being inlined when compiling
4396 <Literal>PrelBase</Literal>, so that an importing module will “see” the <Literal>(:)</Literal>, and can
4397 match it on the LHS of a rule. <Literal>INLINE</Literal> prevents any inlining happening
4398 in the RHS of the <Literal>INLINE</Literal> thing. I regret the delicacy of this.
4405 In <Filename>ghc/lib/std/PrelBase.lhs</Filename> look at the rules for <Function>map</Function> to
4406 see how to write rules that will do fusion and yet give an efficient
4407 program even if fusion doesn't happen. More rules in <Filename>PrelList.lhs</Filename>.
4419 <Sect1 id="generic-classes">
4420 <Title>Generic classes</Title>
4423 The ideas behind this extension are described in detail in "Derivable type classes",
4424 Ralf Hinze and Simon Peyton Jones, Haskell Workshop, Montreal Sept 2000, pp94-105.
4425 An example will give the idea:
4433 fromBin :: [Int] -> (a, [Int])
4435 toBin {| Unit |} Unit = []
4436 toBin {| a :+: b |} (Inl x) = 0 : toBin x
4437 toBin {| a :+: b |} (Inr y) = 1 : toBin y
4438 toBin {| a :*: b |} (x :*: y) = toBin x ++ toBin y
4440 fromBin {| Unit |} bs = (Unit, bs)
4441 fromBin {| a :+: b |} (0:bs) = (Inl x, bs') where (x,bs') = fromBin bs
4442 fromBin {| a :+: b |} (1:bs) = (Inr y, bs') where (y,bs') = fromBin bs
4443 fromBin {| a :*: b |} bs = (x :*: y, bs'') where (x,bs' ) = fromBin bs
4444 (y,bs'') = fromBin bs'
4447 This class declaration explains how <Literal>toBin</Literal> and <Literal>fromBin</Literal>
4448 work for arbitrary data types. They do so by giving cases for unit, product, and sum,
4449 which are defined thus in the library module <Literal>Generics</Literal>:
4453 data a :+: b = Inl a | Inr b
4454 data a :*: b = a :*: b
4457 Now you can make a data type into an instance of Bin like this:
4459 instance (Bin a, Bin b) => Bin (a,b)
4460 instance Bin a => Bin [a]
4462 That is, just leave off the "where" clasuse. Of course, you can put in the
4463 where clause and over-ride whichever methods you please.
4467 <Title> Using generics </Title>
4468 <Para>To use generics you need to</para>
4471 <Para>Use the <Option>-fgenerics</Option> flag.</Para>
4474 <Para>Import the module <Literal>Generics</Literal> from the
4475 <Literal>lang</Literal> package. This import brings into
4476 scope the data types <Literal>Unit</Literal>,
4477 <Literal>:*:</Literal>, and <Literal>:+:</Literal>. (You
4478 don't need this import if you don't mention these types
4479 explicitly; for example, if you are simply giving instance
4480 declarations.)</Para>
4485 <Sect2> <Title> Changes wrt the paper </Title>
4487 Note that the type constructors <Literal>:+:</Literal> and <Literal>:*:</Literal>
4488 can be written infix (indeed, you can now use
4489 any operator starting in a colon as an infix type constructor). Also note that
4490 the type constructors are not exactly as in the paper (Unit instead of 1, etc).
4491 Finally, note that the syntax of the type patterns in the class declaration
4492 uses "<Literal>{|</Literal>" and "<Literal>{|</Literal>" brackets; curly braces
4493 alone would ambiguous when they appear on right hand sides (an extension we
4494 anticipate wanting).
4498 <Sect2> <Title>Terminology and restrictions</Title>
4500 Terminology. A "generic default method" in a class declaration
4501 is one that is defined using type patterns as above.
4502 A "polymorphic default method" is a default method defined as in Haskell 98.
4503 A "generic class declaration" is a class declaration with at least one
4504 generic default method.
4512 Alas, we do not yet implement the stuff about constructor names and
4519 A generic class can have only one parameter; you can't have a generic
4520 multi-parameter class.
4526 A default method must be defined entirely using type patterns, or entirely
4527 without. So this is illegal:
4530 op :: a -> (a, Bool)
4531 op {| Unit |} Unit = (Unit, True)
4534 However it is perfectly OK for some methods of a generic class to have
4535 generic default methods and others to have polymorphic default methods.
4541 The type variable(s) in the type pattern for a generic method declaration
4542 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:
4546 op {| p :*: q |} (x :*: y) = op (x :: p)
4554 The type patterns in a generic default method must take one of the forms:
4560 where "a" and "b" are type variables. Furthermore, all the type patterns for
4561 a single type constructor (<Literal>:*:</Literal>, say) must be identical; they
4562 must use the same type variables. So this is illegal:
4566 op {| a :+: b |} (Inl x) = True
4567 op {| p :+: q |} (Inr y) = False
4569 The type patterns must be identical, even in equations for different methods of the class.
4570 So this too is illegal:
4574 op {| a :*: b |} (Inl x) = True
4577 op {| p :*: q |} (Inr y) = False
4579 (The reason for this restriction is that we gather all the equations for a particular type consructor
4580 into a single generic instance declaration.)
4586 A generic method declaration must give a case for each of the three type constructors.
4592 The type for a generic method can be built only from:
4594 <ListItem> <Para> Function arrows </Para> </ListItem>
4595 <ListItem> <Para> Type variables </Para> </ListItem>
4596 <ListItem> <Para> Tuples </Para> </ListItem>
4597 <ListItem> <Para> Arbitrary types not involving type variables </Para> </ListItem>
4599 Here are some example type signatures for generic methods:
4602 op2 :: Bool -> (a,Bool)
4603 op3 :: [Int] -> a -> a
4606 Here, op1, op2, op3 are OK, but op4 is rejected, because it has a type variable
4610 This restriction is an implementation restriction: we just havn't got around to
4611 implementing the necessary bidirectional maps over arbitrary type constructors.
4612 It would be relatively easy to add specific type constructors, such as Maybe and list,
4613 to the ones that are allowed.</para>
4618 In an instance declaration for a generic class, the idea is that the compiler
4619 will fill in the methods for you, based on the generic templates. However it can only
4624 The instance type is simple (a type constructor applied to type variables, as in Haskell 98).
4629 No constructor of the instance type has unboxed fields.
4633 (Of course, these things can only arise if you are already using GHC extensions.)
4634 However, you can still give an instance declarations for types which break these rules,
4635 provided you give explicit code to override any generic default methods.
4643 The option <Option>-ddump-deriv</Option> dumps incomprehensible stuff giving details of
4644 what the compiler does with generic declarations.
4649 <Sect2> <Title> Another example </Title>
4651 Just to finish with, here's another example I rather like:
4655 nCons {| Unit |} _ = 1
4656 nCons {| a :*: b |} _ = 1
4657 nCons {| a :+: b |} _ = nCons (bot::a) + nCons (bot::b)
4660 tag {| Unit |} _ = 1
4661 tag {| a :*: b |} _ = 1
4662 tag {| a :+: b |} (Inl x) = tag x
4663 tag {| a :+: b |} (Inr y) = nCons (bot::a) + tag y
4670 ;;; Local Variables: ***
4672 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter" "sect1") ***
projects / ghc-hetmet.git / blob