1 %*********************************************************************************
3 %* John Hughes's and Simon Peyton Jones's Pretty Printer Combinators *
5 %* based on "The Design of a Pretty-printing Library" *
6 %* in Advanced Functional Programming, *
7 %* Johan Jeuring and Erik Meijer (eds), LNCS 925 *
8 %* http://www.cs.chalmers.se/~rjmh/Papers/pretty.ps *
10 %* Heavily modified by Simon Peyton Jones, Dec 96 *
12 %*********************************************************************************
14 Version 3.0 28 May 1997
15 * Cured massive performance bug. If you write
17 foldl <> empty (map (text.show) [1..10000])
19 you get quadratic behaviour with V2.0. Why? For just the same reason as you get
20 quadratic behaviour with left-associated (++) chains.
22 This is really bad news. One thing a pretty-printer abstraction should
23 certainly guarantee is insensivity to associativity. It matters: suddenly
24 GHC's compilation times went up by a factor of 100 when I switched to the
27 I fixed it with a bit of a hack (because I wanted to get GHC back on the
28 road). I added two new constructors to the Doc type, Above and Beside:
33 Then, where I need to get to a "TextBeside" or "NilAbove" form I "force"
34 the Doc to squeeze out these suspended calls to Beside and Above; but in so
35 doing I re-associate. It's quite simple, but I'm not satisfied that I've done
36 the best possible job. I'll send you the code if you are interested.
40 int, integer, float, double, rational,
41 lparen, rparen, lbrack, rbrack, lbrace, rbrace,
43 * fullRender's type signature has changed. Rather than producing a string it
44 now takes an extra couple of arguments that tells it how to glue fragments
49 -> Float -- Ribbons per line
50 -> (TextDetails -> a -> a) -- What to do with text
51 -> a -- What to do at the end
55 The "fragments" are encapsulated in the TextDetails data type:
56 data TextDetails = Chr Char
60 The Chr and Str constructors are obvious enough. The PStr constructor has a packed
61 string (FastString) inside it. It's generated by using the new "ptext" export.
63 An advantage of this new setup is that you can get the renderer to do output
64 directly (by passing in a function of type (TextDetails -> IO () -> IO ()),
65 rather than producing a string that you then print.
68 Version 2.0 24 April 1997
69 * Made empty into a left unit for <> as well as a right unit;
70 it is also now true that
72 which wasn't true before.
74 * Fixed an obscure bug in sep that occassionally gave very wierd behaviour
78 * Corrected and tidied up the laws and invariants
80 ======================================================================
81 Relative to John's original paper, there are the following new features:
83 1. There's an empty document, "empty". It's a left and right unit for
84 both <> and $$, and anywhere in the argument list for
85 sep, hcat, hsep, vcat, fcat etc.
87 It is Really Useful in practice.
89 2. There is a paragraph-fill combinator, fsep, that's much like sep,
90 only it keeps fitting things on one line until it can't fit any more.
92 3. Some random useful extra combinators are provided.
93 <+> puts its arguments beside each other with a space between them,
94 unless either argument is empty in which case it returns the other
97 hcat is a list version of <>
98 hsep is a list version of <+>
99 vcat is a list version of $$
101 sep (separate) is either like hsep or like vcat, depending on what fits
103 cat is behaves like sep, but it uses <> for horizontal conposition
104 fcat is behaves like fsep, but it uses <> for horizontal conposition
106 These new ones do the obvious things:
107 char, semi, comma, colon, space,
108 parens, brackets, braces,
111 4. The "above" combinator, $$, now overlaps its two arguments if the
112 last line of the top argument stops before the first line of the second begins.
113 For example: text "hi" $$ nest 5 "there"
120 There are two places this is really useful
122 a) When making labelled blocks, like this:
123 Left -> code for left
124 Right -> code for right
126 code for longlonglonglabel
127 The block is on the same line as the label if the label is
128 short, but on the next line otherwise.
130 b) When laying out lists like this:
135 which some people like. But if the list fits on one line
136 you want [first, second, third]. You can't do this with
137 John's original combinators, but it's quite easy with the
140 The combinator $+$ gives the original "never-overlap" behaviour.
142 5. Several different renderers are provided:
144 * one that uses cut-marks to avoid deeply-nested documents
145 simply piling up in the right-hand margin
146 * one that ignores indentation (fewer chars output; good for machines)
147 * one that ignores indentation and newlines (ditto, only more so)
149 6. Numerous implementation tidy-ups
150 Use of unboxed data types to speed up the implementation
155 {-# OPTIONS -fno-warn-unused-imports #-}
156 -- XXX GHC 6.9 seems to be confused by unpackCString# being used only in
161 Mode(..), TextDetails(..),
163 empty, isEmpty, nest,
165 text, char, ftext, ptext,
166 int, integer, float, double, rational,
167 parens, brackets, braces, quotes, doubleQuotes,
168 semi, comma, colon, space, equals,
169 lparen, rparen, lbrack, rbrack, lbrace, rbrace, cparen,
171 (<>), (<+>), hcat, hsep,
178 -- renderStyle, -- Haskell 1.3 only
179 render, fullRender, printDoc, showDocWith
187 import Numeric (fromRat)
189 --import Foreign.Ptr (castPtr)
191 #if defined(__GLASGOW_HASKELL__)
193 import GHC.Base ( unpackCString# )
194 import GHC.Exts ( Int# )
195 import GHC.Ptr ( Ptr(..) )
198 -- Don't import Util( assertPanic ) because it makes a loop in the module structure
208 -- Disable ASSERT checks; they are expensive!
209 #define LOCAL_ASSERT(x)
214 %*********************************************************
216 \subsection{The interface}
218 %*********************************************************
220 The primitive @Doc@ values
224 isEmpty :: Doc -> Bool
225 text :: String -> Doc
228 semi, comma, colon, space, equals :: Doc
229 lparen, rparen, lbrack, rbrack, lbrace, rbrace :: Doc
231 parens, brackets, braces :: Doc -> Doc
232 quotes, doubleQuotes :: Doc -> Doc
235 integer :: Integer -> Doc
236 float :: Float -> Doc
237 double :: Double -> Doc
238 rational :: Rational -> Doc
241 Combining @Doc@ values
244 (<>) :: Doc -> Doc -> Doc -- Beside
245 hcat :: [Doc] -> Doc -- List version of <>
246 (<+>) :: Doc -> Doc -> Doc -- Beside, separated by space
247 hsep :: [Doc] -> Doc -- List version of <+>
249 ($$) :: Doc -> Doc -> Doc -- Above; if there is no
250 -- overlap it "dovetails" the two
251 vcat :: [Doc] -> Doc -- List version of $$
253 cat :: [Doc] -> Doc -- Either hcat or vcat
254 sep :: [Doc] -> Doc -- Either hsep or vcat
255 fcat :: [Doc] -> Doc -- ``Paragraph fill'' version of cat
256 fsep :: [Doc] -> Doc -- ``Paragraph fill'' version of sep
258 nest :: Int -> Doc -> Doc -- Nested
264 hang :: Doc -> Int -> Doc -> Doc
265 punctuate :: Doc -> [Doc] -> [Doc] -- punctuate p [d1, ... dn] = [d1 <> p, d2 <> p, ... dn-1 <> p, dn]
268 Displaying @Doc@ values.
271 instance Show Doc where
272 showsPrec _ doc cont = showDoc doc cont
274 render :: Doc -> String -- Uses default style
276 -> Int -- Line length
277 -> Float -- Ribbons per line
278 -> (TextDetails -> a -> a) -- What to do with text
279 -> a -- What to do at the end
283 {- When we start using 1.3
284 renderStyle :: Style -> Doc -> String
285 data Style = Style { lineLength :: Int, -- In chars
286 ribbonsPerLine :: Float, -- Ratio of ribbon length to line length
289 style :: Style -- The default style
290 style = Style { lineLength = 100, ribbonsPerLine = 2.5, mode = PageMode }
293 data Mode = PageMode -- Normal
294 | ZigZagMode -- With zig-zag cuts
295 | LeftMode -- No indentation, infinitely long lines
296 | OneLineMode -- All on one line
301 %*********************************************************
303 \subsection{The @Doc@ calculus}
305 %*********************************************************
307 The @Doc@ combinators satisfy the following laws:
311 <a1> (x $$ y) $$ z = x $$ (y $$ z)
319 <b1> (x <> y) <> z = x <> (y <> z)
320 <b2> empty <> x = empty
327 <t1> text s <> text t = text (s++t)
328 <t2> text "" <> x = x, if x non-empty
333 <n2> nest k (nest k' x) = nest (k+k') x
334 <n3> nest k (x <> y) = nest k z <> nest k y
335 <n4> nest k (x $$ y) = nest k x $$ nest k y
336 <n5> nest k empty = empty
337 <n6> x <> nest k y = x <> y, if x non-empty
339 - Note the side condition on <n6>! It is this that
340 makes it OK for empty to be a left unit for <>.
344 <m1> (text s <> x) $$ y = text s <> ((text "" <> x)) $$
347 <m2> (x $$ y) <> z = x $$ (y <> z)
351 Laws for list versions
352 ~~~~~~~~~~~~~~~~~~~~~~
353 <l1> sep (ps++[empty]++qs) = sep (ps ++ qs)
354 ...ditto hsep, hcat, vcat, fill...
356 <l2> nest k (sep ps) = sep (map (nest k) ps)
357 ...ditto hsep, hcat, vcat, fill...
361 <o1> oneLiner (nest k p) = nest k (oneLiner p)
362 <o2> oneLiner (x <> y) = oneLiner x <> oneLiner y
366 You might think that the following verion of <m1> would
369 <3 NO> (text s <> x) $$ y = text s <> ((empty <> x)) $$
372 But it doesn't work, for if x=empty, we would have
374 text s $$ y = text s <> (empty $$ nest (-length s) y)
375 = text s <> nest (-length s) y
380 %*********************************************************
382 \subsection{Simple derived definitions}
384 %*********************************************************
399 int n = text (show n)
400 integer n = text (show n)
401 float n = text (show n)
402 double n = text (show n)
403 rational n = text (show (fromRat n :: Double))
404 --rational n = text (show (fromRationalX n)) -- _showRational 30 n)
406 quotes p = char '`' <> p <> char '\''
407 doubleQuotes p = char '"' <> p <> char '"'
408 parens p = char '(' <> p <> char ')'
409 brackets p = char '[' <> p <> char ']'
410 braces p = char '{' <> p <> char '}'
412 cparen :: Bool -> Doc -> Doc
416 hcat = foldr (<>) empty
417 hsep = foldr (<+>) empty
418 vcat = foldr ($$) empty
420 hang d1 n d2 = sep [d1, nest n d2]
423 punctuate p (d:ds) = go d ds
426 go d (e:es) = (d <> p) : go e es
430 %*********************************************************
432 \subsection{The @Doc@ data type}
434 %*********************************************************
436 A @Doc@ represents a {\em set} of layouts. A @Doc@ with
437 no occurrences of @Union@ or @NoDoc@ represents just one layout.
441 | NilAbove Doc -- text "" $$ x
442 | TextBeside !TextDetails FastInt Doc -- text s <> x
443 | Nest FastInt Doc -- nest k x
444 | Union Doc Doc -- ul `union` ur
445 | NoDoc -- The empty set of documents
446 | Beside Doc Bool Doc -- True <=> space between
447 | Above Doc Bool Doc -- True <=> never overlap
449 type RDoc = Doc -- RDoc is a "reduced Doc", guaranteed not to have a top-level Above or Beside
452 reduceDoc :: Doc -> RDoc
453 reduceDoc (Beside p g q) = beside p g (reduceDoc q)
454 reduceDoc (Above p g q) = above p g (reduceDoc q)
458 data TextDetails = Chr {-#UNPACK#-}!Char
460 | PStr FastString -- a hashed string
461 | LStr {-#UNPACK#-}!LitString FastInt -- a '\0'-terminated
464 space_text :: TextDetails
466 nl_text :: TextDetails
470 Here are the invariants:
473 The argument of @NilAbove@ is never @Empty@. Therefore
474 a @NilAbove@ occupies at least two lines.
477 The arugment of @TextBeside@ is never @Nest@.
480 The layouts of the two arguments of @Union@ both flatten to the same string.
483 The arguments of @Union@ are either @TextBeside@, or @NilAbove@.
486 The right argument of a union cannot be equivalent to the empty set (@NoDoc@).
487 If the left argument of a union is equivalent to the empty set (@NoDoc@),
488 then the @NoDoc@ appears in the first line.
491 An empty document is always represented by @Empty@.
492 It can't be hidden inside a @Nest@, or a @Union@ of two @Empty@s.
495 The first line of every layout in the left argument of @Union@
496 is longer than the first line of any layout in the right argument.
497 (1) ensures that the left argument has a first line. In view of (3),
498 this invariant means that the right argument must have at least two
503 -- Arg of a NilAbove is always an RDoc
504 nilAbove_ :: Doc -> Doc
505 nilAbove_ p = LOCAL_ASSERT( _ok p ) NilAbove p
510 -- Arg of a TextBeside is always an RDoc
511 textBeside_ :: TextDetails -> FastInt -> Doc -> Doc
512 textBeside_ s sl p = TextBeside s sl (LOCAL_ASSERT( _ok p ) p)
514 _ok (Nest _ _) = False
517 -- Arg of Nest is always an RDoc
518 nest_ :: FastInt -> Doc -> Doc
519 nest_ k p = Nest k (LOCAL_ASSERT( _ok p ) p)
524 -- Args of union are always RDocs
525 union_ :: Doc -> Doc -> Doc
526 union_ p q = Union (LOCAL_ASSERT( _ok p ) p) (LOCAL_ASSERT( _ok q ) q)
528 _ok (TextBeside _ _ _) = True
529 _ok (NilAbove _) = True
530 _ok (Union _ _) = True
535 Notice the difference between
536 * NoDoc (no documents)
537 * Empty (one empty document; no height and no width)
538 * text "" (a document containing the empty string;
539 one line high, but has no width)
543 %*********************************************************
545 \subsection{@empty@, @text@, @nest@, @union@}
547 %*********************************************************
555 char c = textBeside_ (Chr c) (_ILIT(1)) Empty
556 text s = case iUnbox (length s) of {sl -> textBeside_ (Str s) sl Empty}
557 ftext :: FastString -> Doc
558 ftext s = case iUnbox (lengthFS s) of {sl -> textBeside_ (PStr s) sl Empty}
559 ptext :: LitString -> Doc
560 ptext s_= case iUnbox (strLength s) of {sl -> textBeside_ (LStr s sl) sl Empty}
561 where s = {-castPtr-} s_
563 #if defined(__GLASGOW_HASKELL__)
564 -- RULE that turns (text "abc") into (ptext (A# "abc"#)) to avoid the
565 -- intermediate packing/unpacking of the string.
567 "text/str" forall a. text (unpackCString# a) = ptext (Ptr a)
571 nest k p = mkNest (iUnbox k) (reduceDoc p) -- Externally callable version
573 -- mkNest checks for Nest's invariant that it doesn't have an Empty inside it
574 mkNest :: Int# -> Doc -> Doc
575 mkNest k (Nest k1 p) = mkNest (k +# k1) p
576 mkNest _ NoDoc = NoDoc
577 mkNest _ Empty = Empty
578 mkNest k p | k ==# _ILIT(0) = p -- Worth a try!
579 mkNest k p = nest_ k p
581 -- mkUnion checks for an empty document
582 mkUnion :: Doc -> Doc -> Doc
583 mkUnion Empty _ = Empty
584 mkUnion p q = p `union_` q
587 %*********************************************************
589 \subsection{Vertical composition @$$@}
591 %*********************************************************
595 p $$ q = Above p False q
596 ($+$) :: Doc -> Doc -> Doc
597 p $+$ q = Above p True q
599 above :: Doc -> Bool -> RDoc -> RDoc
600 above (Above p g1 q1) g2 q2 = above p g1 (above q1 g2 q2)
601 above p@(Beside _ _ _) g q = aboveNest (reduceDoc p) g (_ILIT(0)) (reduceDoc q)
602 above p g q = aboveNest p g (_ILIT(0)) (reduceDoc q)
604 aboveNest :: RDoc -> Bool -> FastInt -> RDoc -> RDoc
605 -- Specfication: aboveNest p g k q = p $g$ (nest k q)
607 aboveNest NoDoc _ _ _ = NoDoc
608 aboveNest (p1 `Union` p2) g k q = aboveNest p1 g k q `union_`
611 aboveNest Empty _ k q = mkNest k q
612 aboveNest (Nest k1 p) g k q = nest_ k1 (aboveNest p g (k -# k1) q)
613 -- p can't be Empty, so no need for mkNest
615 aboveNest (NilAbove p) g k q = nilAbove_ (aboveNest p g k q)
616 aboveNest (TextBeside s sl p) g k q = textBeside_ s sl rest
620 Empty -> nilAboveNest g k1 q
621 _ -> aboveNest p g k1 q
622 aboveNest _ _ _ _ = panic "aboveNest: Unhandled case"
626 nilAboveNest :: Bool -> FastInt -> RDoc -> RDoc
627 -- Specification: text s <> nilaboveNest g k q
628 -- = text s <> (text "" $g$ nest k q)
630 nilAboveNest _ _ Empty = Empty -- Here's why the "text s <>" is in the spec!
631 nilAboveNest g k (Nest k1 q) = nilAboveNest g (k +# k1) q
633 nilAboveNest g k q | (not g) && (k ># _ILIT(0)) -- No newline if no overlap
634 = textBeside_ (Str (spaces k)) k q
635 | otherwise -- Put them really above
636 = nilAbove_ (mkNest k q)
640 %*********************************************************
642 \subsection{Horizontal composition @<>@}
644 %*********************************************************
647 p <> q = Beside p False q
648 p <+> q = Beside p True q
650 beside :: Doc -> Bool -> RDoc -> RDoc
651 -- Specification: beside g p q = p <g> q
653 beside NoDoc _ _ = NoDoc
654 beside (p1 `Union` p2) g q = (beside p1 g q) `union_` (beside p2 g q)
656 beside (Nest k p) g q = nest_ k $! beside p g q -- p non-empty
657 beside p@(Beside p1 g1 q1) g2 q2
658 {- (A `op1` B) `op2` C == A `op1` (B `op2` C) iff op1 == op2
659 [ && (op1 == <> || op1 == <+>) ] -}
660 | g1 == g2 = beside p1 g1 $! beside q1 g2 q2
661 | otherwise = beside (reduceDoc p) g2 q2
662 beside p@(Above _ _ _) g q = let d = reduceDoc p in d `seq` beside d g q
663 beside (NilAbove p) g q = nilAbove_ $! beside p g q
664 beside (TextBeside s sl p) g q = textBeside_ s sl $! rest
667 Empty -> nilBeside g q
672 nilBeside :: Bool -> RDoc -> RDoc
673 -- Specification: text "" <> nilBeside g p
676 nilBeside _ Empty = Empty -- Hence the text "" in the spec
677 nilBeside g (Nest _ p) = nilBeside g p
678 nilBeside g p | g = textBeside_ space_text (_ILIT(1)) p
682 %*********************************************************
684 \subsection{Separate, @sep@, Hughes version}
686 %*********************************************************
689 -- Specification: sep ps = oneLiner (hsep ps)
693 sep = sepX True -- Separate with spaces
694 cat = sepX False -- Don't
696 sepX :: Bool -> [Doc] -> Doc
698 sepX x (p:ps) = sep1 x (reduceDoc p) (_ILIT(0)) ps
701 -- Specification: sep1 g k ys = sep (x : map (nest k) ys)
702 -- = oneLiner (x <g> nest k (hsep ys))
703 -- `union` x $$ nest k (vcat ys)
705 sep1 :: Bool -> RDoc -> FastInt -> [Doc] -> RDoc
706 sep1 _ NoDoc _ _ = NoDoc
707 sep1 g (p `Union` q) k ys = sep1 g p k ys
709 (aboveNest q False k (reduceDoc (vcat ys)))
711 sep1 g Empty k ys = mkNest k (sepX g ys)
712 sep1 g (Nest n p) k ys = nest_ n (sep1 g p (k -# n) ys)
714 sep1 _ (NilAbove p) k ys = nilAbove_ (aboveNest p False k (reduceDoc (vcat ys)))
715 sep1 g (TextBeside s sl p) k ys = textBeside_ s sl (sepNB g p (k -# sl) ys)
716 sep1 _ _ _ _ = panic "sep1: Unhandled case"
718 -- Specification: sepNB p k ys = sep1 (text "" <> p) k ys
719 -- Called when we have already found some text in the first item
720 -- We have to eat up nests
722 sepNB :: Bool -> Doc -> FastInt -> [Doc] -> Doc
723 sepNB g (Nest _ p) k ys = sepNB g p k ys
725 sepNB g Empty k ys = oneLiner (nilBeside g (reduceDoc rest))
727 nilAboveNest False k (reduceDoc (vcat ys))
730 | otherwise = hcat ys
732 sepNB g p k ys = sep1 g p k ys
735 %*********************************************************
739 %*********************************************************
748 -- fill (p1:p2:ps) = oneLiner p1 <#> nest (length p1)
749 -- (fill (oneLiner p2 : ps))
753 fill :: Bool -> [Doc] -> Doc
755 fill g (p:ps) = fill1 g (reduceDoc p) (_ILIT(0)) ps
758 fill1 :: Bool -> RDoc -> FastInt -> [Doc] -> Doc
759 fill1 _ NoDoc _ _ = NoDoc
760 fill1 g (p `Union` q) k ys = fill1 g p k ys
762 (aboveNest q False k (fill g ys))
764 fill1 g Empty k ys = mkNest k (fill g ys)
765 fill1 g (Nest n p) k ys = nest_ n (fill1 g p (k -# n) ys)
767 fill1 g (NilAbove p) k ys = nilAbove_ (aboveNest p False k (fill g ys))
768 fill1 g (TextBeside s sl p) k ys = textBeside_ s sl (fillNB g p (k -# sl) ys)
769 fill1 _ _ _ _ = panic "fill1: Unhandled case"
771 fillNB :: Bool -> Doc -> Int# -> [Doc] -> Doc
772 fillNB g (Nest _ p) k ys = fillNB g p k ys
773 fillNB _ Empty _ [] = Empty
774 fillNB g Empty k (y:ys) = nilBeside g (fill1 g (oneLiner (reduceDoc y)) k1 ys)
776 nilAboveNest False k (fill g (y:ys))
778 k1 | g = k -# _ILIT(1)
781 fillNB g p k ys = fill1 g p k ys
785 %*********************************************************
787 \subsection{Selecting the best layout}
789 %*********************************************************
792 best :: Int -- Line length
793 -> Int -- Ribbon length
795 -> RDoc -- No unions in here!
801 get :: FastInt -- (Remaining) width of line
805 get w (NilAbove p) = nilAbove_ (get w p)
806 get w (TextBeside s sl p) = textBeside_ s sl (get1 w sl p)
807 get w (Nest k p) = nest_ k (get (w -# k) p)
808 get w (p `Union` q) = nicest w r (get w p) (get w q)
809 get _ _ = panic "best/get: Unhandled case"
811 get1 :: FastInt -- (Remaining) width of line
812 -> FastInt -- Amount of first line already eaten up
813 -> Doc -- This is an argument to TextBeside => eat Nests
814 -> Doc -- No unions in here!
816 get1 _ _ Empty = Empty
817 get1 _ _ NoDoc = NoDoc
818 get1 w sl (NilAbove p) = nilAbove_ (get (w -# sl) p)
819 get1 w sl (TextBeside t tl p) = textBeside_ t tl (get1 w (sl +# tl) p)
820 get1 w sl (Nest _ p) = get1 w sl p
821 get1 w sl (p `Union` q) = nicest1 w r sl (get1 w sl p)
823 get1 _ _ _ = panic "best/get1: Unhandled case"
825 nicest :: FastInt -> FastInt -> Doc -> Doc -> Doc
826 nicest w r p q = nicest1 w r (_ILIT(0)) p q
827 nicest1 :: FastInt -> FastInt -> Int# -> Doc -> Doc -> Doc
828 nicest1 w r sl p q | fits ((w `minFastInt` r) -# sl) p = p
831 fits :: FastInt -- Space available
833 -> Bool -- True if *first line* of Doc fits in space available
835 fits n _ | n <# _ILIT(0) = False
838 fits _ (NilAbove _) = True
839 fits n (TextBeside _ sl p) = fits (n -# sl) p
840 fits _ _ = panic "fits: Unhandled case"
843 @first@ and @nonEmptySet@ are similar to @nicest@ and @fits@, only simpler.
844 @first@ returns its first argument if it is non-empty, otherwise its second.
847 first :: Doc -> Doc -> Doc
848 first p q | nonEmptySet p = p
851 nonEmptySet :: Doc -> Bool
852 nonEmptySet NoDoc = False
853 nonEmptySet (_ `Union` _) = True
854 nonEmptySet Empty = True
855 nonEmptySet (NilAbove _) = True -- NoDoc always in first line
856 nonEmptySet (TextBeside _ _ p) = nonEmptySet p
857 nonEmptySet (Nest _ p) = nonEmptySet p
858 nonEmptySet _ = panic "nonEmptySet: Unhandled case"
861 @oneLiner@ returns the one-line members of the given set of @Doc@s.
864 oneLiner :: Doc -> Doc
865 oneLiner NoDoc = NoDoc
866 oneLiner Empty = Empty
867 oneLiner (NilAbove _) = NoDoc
868 oneLiner (TextBeside s sl p) = textBeside_ s sl (oneLiner p)
869 oneLiner (Nest k p) = nest_ k (oneLiner p)
870 oneLiner (p `Union` _) = oneLiner p
871 oneLiner _ = panic "oneLiner: Unhandled case"
876 %*********************************************************
878 \subsection{Displaying the best layout}
880 %*********************************************************
885 renderStyle Style{mode, lineLength, ribbonsPerLine} doc
886 = fullRender mode lineLength ribbonsPerLine doc ""
889 render doc = showDocWith PageMode doc
891 showDoc :: Doc -> String -> String
892 showDoc doc rest = showDocWithAppend PageMode doc rest
894 showDocWithAppend :: Mode -> Doc -> String -> String
895 showDocWithAppend mode doc rest = fullRender mode 100 1.5 string_txt rest doc
897 showDocWith :: Mode -> Doc -> String
898 showDocWith mode doc = showDocWithAppend mode doc ""
900 string_txt :: TextDetails -> String -> String
901 string_txt (Chr c) s = c:s
902 string_txt (Str s1) s2 = s1 ++ s2
903 string_txt (PStr s1) s2 = unpackFS s1 ++ s2
904 string_txt (LStr s1 _) s2 = unpackLitString s1 ++ s2
909 fullRender OneLineMode _ _ txt end doc
910 = lay (reduceDoc doc)
912 lay NoDoc = cant_fail
913 lay (Union _ q) = lay q -- Second arg can't be NoDoc
914 lay (Nest _ p) = lay p
916 lay (NilAbove p) = space_text `txt` lay p -- NoDoc always on
918 lay (TextBeside s _ p) = s `txt` lay p
919 lay _ = panic "fullRender/OneLineMode/lay: Unhandled case"
921 fullRender LeftMode _ _ txt end doc
922 = lay (reduceDoc doc)
924 lay NoDoc = cant_fail
925 lay (Union p q) = lay (first p q)
926 lay (Nest _ p) = lay p
928 lay (NilAbove p) = nl_text `txt` lay p -- NoDoc always on first line
929 lay (TextBeside s _ p) = s `txt` lay p
930 lay _ = panic "fullRender/LeftMode/lay: Unhandled case"
932 fullRender mode line_length ribbons_per_line txt end doc
933 = display mode line_length ribbon_length txt end best_doc
935 best_doc = best hacked_line_length ribbon_length (reduceDoc doc)
937 hacked_line_length, ribbon_length :: Int
938 ribbon_length = round (fromIntegral line_length / ribbons_per_line)
939 hacked_line_length = case mode of
940 ZigZagMode -> maxBound
943 display :: Mode -> Int -> Int -> (TextDetails -> t -> t) -> t -> Doc -> t
944 display mode page_width ribbon_width txt end doc
945 = case (iUnbox page_width) -# (iUnbox ribbon_width) of { gap_width ->
946 case gap_width `quotFastInt` _ILIT(2) of { shift ->
948 lay k (Nest k1 p) = lay (k +# k1) p
951 lay k (NilAbove p) = nl_text `txt` lay k p
953 lay k (TextBeside s sl p)
955 ZigZagMode | k >=# gap_width
957 Str (multi_ch shift '/') `txt` (
959 lay1 (k -# shift) s sl p)))
963 Str (multi_ch shift '\\') `txt` (
965 lay1 (k +# shift) s sl p )))
968 lay _ _ = panic "display/lay: Unhandled case"
970 lay1 k s sl p = indent k (s `txt` lay2 (k +# sl) p)
972 lay2 k (NilAbove p) = nl_text `txt` lay k p
973 lay2 k (TextBeside s sl p) = s `txt` (lay2 (k +# sl) p)
974 lay2 k (Nest _ p) = lay2 k p
976 lay2 _ _ = panic "display/lay2: Unhandled case"
978 -- optimise long indentations using LitString chunks of 8 spaces
979 indent n r | n >=# _ILIT(8) = LStr (sLit " ") (_ILIT(8)) `txt`
980 indent (n -# _ILIT(8)) r
981 | otherwise = Str (spaces n) `txt` r
987 cant_fail = error "easy_display: NoDoc"
989 multi_ch :: Int# -> Char -> String
990 multi_ch n ch | n <=# _ILIT(0) = ""
991 | otherwise = ch : multi_ch (n -# _ILIT(1)) ch
993 spaces :: Int# -> String
994 spaces n | n <=# _ILIT(0) = ""
995 | otherwise = ' ' : spaces (n -# _ILIT(1))
1001 pprCols = 100 -- could make configurable
1003 -- NB. printDoc prints FastStrings in UTF-8: hPutFS below does no decoding.
1004 -- This is what we usually want, because the IO library has no encoding
1005 -- functionality, and we're assuming UTF-8 source code so we might as well
1006 -- assume UTF-8 output too.
1007 printDoc :: Mode -> Handle -> Doc -> IO ()
1008 printDoc LeftMode hdl doc
1009 = do { printLeftRender hdl doc; hFlush hdl }
1010 printDoc mode hdl doc
1011 = do { fullRender mode pprCols 1.5 put done doc ;
1014 put (Chr c) next = hPutChar hdl c >> next
1015 put (Str s) next = hPutStr hdl s >> next
1016 put (PStr s) next = hPutFS hdl s >> next
1017 put (LStr s l) next = hPutLitString hdl s l >> next
1019 done = hPutChar hdl '\n'
1021 -- some versions of hPutBuf will barf if the length is zero
1022 hPutLitString :: Handle -> Ptr a -> Int# -> IO ()
1023 hPutLitString handle a l = if l ==# _ILIT(0)
1025 else hPutBuf handle a (iBox l)
1027 -- Printing output in LeftMode is performance critical: it's used when
1028 -- dumping C and assembly output, so we allow ourselves a few dirty
1031 -- (1) we specialise fullRender for LeftMode with IO output.
1033 -- (2) we add a layer of buffering on top of Handles. Handles
1034 -- don't perform well with lots of hPutChars, which is mostly
1035 -- what we're doing here, because Handles have to be thread-safe
1036 -- and async exception-safe. We only have a single thread and don't
1037 -- care about exceptions, so we add a layer of fast buffering
1038 -- over the Handle interface.
1040 -- (3) a few hacks in layLeft below to convince GHC to generate the right
1043 printLeftRender :: Handle -> Doc -> IO ()
1044 printLeftRender hdl doc = do
1045 b <- newBufHandle hdl
1046 layLeft b (reduceDoc doc)
1049 -- HACK ALERT! the "return () >>" below convinces GHC to eta-expand
1050 -- this function with the IO state lambda. Otherwise we end up with
1051 -- closures in all the case branches.
1052 layLeft :: BufHandle -> Doc -> IO ()
1053 layLeft b _ | b `seq` False = undefined -- make it strict in b
1054 layLeft _ NoDoc = cant_fail
1055 layLeft b (Union p q) = return () >> layLeft b (first p q)
1056 layLeft b (Nest _ p) = return () >> layLeft b p
1057 layLeft b Empty = bPutChar b '\n'
1058 layLeft b (NilAbove p) = bPutChar b '\n' >> layLeft b p
1059 layLeft b (TextBeside s _ p) = put b s >> layLeft b p
1061 put b _ | b `seq` False = undefined
1062 put b (Chr c) = bPutChar b c
1063 put b (Str s) = bPutStr b s
1064 put b (PStr s) = bPutFS b s
1065 put b (LStr s l) = bPutLitString b s l
1066 layLeft _ _ = panic "layLeft: Unhandled case"