2b983a518f7459dea855ccde66453373fef9ff92
[packages/random.git] / Text / PrettyPrint / HughesPJ.hs
1 -----------------------------------------------------------------------------
2 -- |
3 -- Module : Text.PrettyPrint.HughesPJ
4 -- Copyright : (c) The University of Glasgow 2001
5 -- License : BSD-style (see the file libraries/base/LICENSE)
6 --
7 -- Maintainer : libraries@haskell.org
8 -- Stability : provisional
9 -- Portability : portable
10 --
11 -- John Hughes's and Simon Peyton Jones's Pretty Printer Combinators
12 --
13 -- Based on /The Design of a Pretty-printing Library/
14 -- in Advanced Functional Programming,
15 -- Johan Jeuring and Erik Meijer (eds), LNCS 925
16 -- <http://www.cs.chalmers.se/~rjmh/Papers/pretty.ps>
17 --
18 -- Heavily modified by Simon Peyton Jones, Dec 96
19 --
20 -----------------------------------------------------------------------------
21
22 {-
23 Version 3.0 28 May 1997
24 * Cured massive performance bug. If you write
25
26 foldl <> empty (map (text.show) [1..10000])
27
28 you get quadratic behaviour with V2.0. Why? For just the same
29 reason as you get quadratic behaviour with left-associated (++)
30 chains.
31
32 This is really bad news. One thing a pretty-printer abstraction
33 should certainly guarantee is insensivity to associativity. It
34 matters: suddenly GHC's compilation times went up by a factor of
35 100 when I switched to the new pretty printer.
36
37 I fixed it with a bit of a hack (because I wanted to get GHC back
38 on the road). I added two new constructors to the Doc type, Above
39 and Beside:
40
41 <> = Beside
42 $$ = Above
43
44 Then, where I need to get to a "TextBeside" or "NilAbove" form I
45 "force" the Doc to squeeze out these suspended calls to Beside and
46 Above; but in so doing I re-associate. It's quite simple, but I'm
47 not satisfied that I've done the best possible job. I'll send you
48 the code if you are interested.
49
50 * Added new exports:
51 punctuate, hang
52 int, integer, float, double, rational,
53 lparen, rparen, lbrack, rbrack, lbrace, rbrace,
54
55 * fullRender's type signature has changed. Rather than producing a
56 string it now takes an extra couple of arguments that tells it how
57 to glue fragments of output together:
58
59 fullRender :: Mode
60 -> Int -- Line length
61 -> Float -- Ribbons per line
62 -> (TextDetails -> a -> a) -- What to do with text
63 -> a -- What to do at the end
64 -> Doc
65 -> a -- Result
66
67 The "fragments" are encapsulated in the TextDetails data type:
68
69 data TextDetails = Chr Char
70 | Str String
71 | PStr FAST_STRING
72
73 The Chr and Str constructors are obvious enough. The PStr
74 constructor has a packed string (FAST_STRING) inside it. It's
75 generated by using the new "ptext" export.
76
77 An advantage of this new setup is that you can get the renderer to
78 do output directly (by passing in a function of type (TextDetails
79 -> IO () -> IO ()), rather than producing a string that you then
80 print.
81
82
83 Version 2.0 24 April 1997
84 * Made empty into a left unit for <> as well as a right unit;
85 it is also now true that
86 nest k empty = empty
87 which wasn't true before.
88
89 * Fixed an obscure bug in sep that occassionally gave very weird behaviour
90
91 * Added $+$
92
93 * Corrected and tidied up the laws and invariants
94
95 ======================================================================
96 Relative to John's original paper, there are the following new features:
97
98 1. There's an empty document, "empty". It's a left and right unit for
99 both <> and $$, and anywhere in the argument list for
100 sep, hcat, hsep, vcat, fcat etc.
101
102 It is Really Useful in practice.
103
104 2. There is a paragraph-fill combinator, fsep, that's much like sep,
105 only it keeps fitting things on one line until it can't fit any more.
106
107 3. Some random useful extra combinators are provided.
108 <+> puts its arguments beside each other with a space between them,
109 unless either argument is empty in which case it returns the other
110
111
112 hcat is a list version of <>
113 hsep is a list version of <+>
114 vcat is a list version of $$
115
116 sep (separate) is either like hsep or like vcat, depending on what fits
117
118 cat behaves like sep, but it uses <> for horizontal conposition
119 fcat behaves like fsep, but it uses <> for horizontal conposition
120
121 These new ones do the obvious things:
122 char, semi, comma, colon, space,
123 parens, brackets, braces,
124 quotes, doubleQuotes
125
126 4. The "above" combinator, $$, now overlaps its two arguments if the
127 last line of the top argument stops before the first line of the
128 second begins.
129
130 For example: text "hi" $$ nest 5 "there"
131 lays out as
132 hi there
133 rather than
134 hi
135 there
136
137 There are two places this is really useful
138
139 a) When making labelled blocks, like this:
140 Left -> code for left
141 Right -> code for right
142 LongLongLongLabel ->
143 code for longlonglonglabel
144 The block is on the same line as the label if the label is
145 short, but on the next line otherwise.
146
147 b) When laying out lists like this:
148 [ first
149 , second
150 , third
151 ]
152 which some people like. But if the list fits on one line
153 you want [first, second, third]. You can't do this with
154 John's original combinators, but it's quite easy with the
155 new $$.
156
157 The combinator $+$ gives the original "never-overlap" behaviour.
158
159 5. Several different renderers are provided:
160 * a standard one
161 * one that uses cut-marks to avoid deeply-nested documents
162 simply piling up in the right-hand margin
163 * one that ignores indentation (fewer chars output; good for machines)
164 * one that ignores indentation and newlines (ditto, only more so)
165
166 6. Numerous implementation tidy-ups
167 Use of unboxed data types to speed up the implementation
168 -}
169
170 module Text.PrettyPrint.HughesPJ (
171
172 -- * The document type
173 Doc, -- Abstract
174
175 -- * Primitive Documents
176 empty,
177 semi, comma, colon, space, equals,
178 lparen, rparen, lbrack, rbrack, lbrace, rbrace,
179
180 -- * Converting values into documents
181 text, char, ptext,
182 int, integer, float, double, rational,
183
184 -- * Wrapping documents in delimiters
185 parens, brackets, braces, quotes, doubleQuotes,
186
187 -- * Combining documents
188 (<>), (<+>), hcat, hsep,
189 ($$), ($+$), vcat,
190 sep, cat,
191 fsep, fcat,
192 nest,
193 hang, punctuate,
194
195 -- * Predicates on documents
196 isEmpty,
197
198 -- * Rendering documents
199
200 -- ** Default rendering
201 render,
202
203 -- ** Rendering with a particular style
204 Style(..),
205 style,
206 renderStyle,
207
208 -- ** General rendering
209 fullRender,
210 Mode(..), TextDetails(..),
211
212 ) where
213
214
215 import Prelude
216
217 infixl 6 <>
218 infixl 6 <+>
219 infixl 5 $$, $+$
220
221 -- ---------------------------------------------------------------------------
222 -- The interface
223
224 -- The primitive Doc values
225
226 isEmpty :: Doc -> Bool; -- ^ Returns 'True' if the document is empty
227
228 empty :: Doc; -- ^ An empty document
229 semi :: Doc; -- ^ A ';' character
230 comma :: Doc; -- ^ A ',' character
231 colon :: Doc; -- ^ A ':' character
232 space :: Doc; -- ^ A space character
233 equals :: Doc; -- ^ A '=' character
234 lparen :: Doc; -- ^ A '(' character
235 rparen :: Doc; -- ^ A ')' character
236 lbrack :: Doc; -- ^ A '[' character
237 rbrack :: Doc; -- ^ A ']' character
238 lbrace :: Doc; -- ^ A '{' character
239 rbrace :: Doc; -- ^ A '}' character
240
241 text :: String -> Doc
242 ptext :: String -> Doc
243 char :: Char -> Doc
244 int :: Int -> Doc
245 integer :: Integer -> Doc
246 float :: Float -> Doc
247 double :: Double -> Doc
248 rational :: Rational -> Doc
249
250
251 parens :: Doc -> Doc; -- ^ Wrap document in @(...)@
252 brackets :: Doc -> Doc; -- ^ Wrap document in @[...]@
253 braces :: Doc -> Doc; -- ^ Wrap document in @{...}@
254 quotes :: Doc -> Doc; -- ^ Wrap document in @\'...\'@
255 doubleQuotes :: Doc -> Doc; -- ^ Wrap document in @\"...\"@
256
257 -- Combining @Doc@ values
258
259 (<>) :: Doc -> Doc -> Doc; -- ^Beside
260 hcat :: [Doc] -> Doc; -- ^List version of '<>'
261 (<+>) :: Doc -> Doc -> Doc; -- ^Beside, separated by space
262 hsep :: [Doc] -> Doc; -- ^List version of '<+>'
263
264 ($$) :: Doc -> Doc -> Doc; -- ^Above; if there is no
265 -- overlap it \"dovetails\" the two
266 ($+$) :: Doc -> Doc -> Doc; -- ^Above, without dovetailing.
267 vcat :: [Doc] -> Doc; -- ^List version of '$$'
268
269 cat :: [Doc] -> Doc; -- ^ Either hcat or vcat
270 sep :: [Doc] -> Doc; -- ^ Either hsep or vcat
271 fcat :: [Doc] -> Doc; -- ^ \"Paragraph fill\" version of cat
272 fsep :: [Doc] -> Doc; -- ^ \"Paragraph fill\" version of sep
273
274 nest :: Int -> Doc -> Doc; -- ^ Nested
275
276
277 -- GHC-specific ones.
278
279 hang :: Doc -> Int -> Doc -> Doc; -- ^ @hang d1 n d2 = sep [d1, nest n d2]@
280 punctuate :: Doc -> [Doc] -> [Doc]; -- ^ @punctuate p [d1, ... dn] = [d1 \<> p, d2 \<> p, ... dn-1 \<> p, dn]@
281
282
283 -- Displaying @Doc@ values.
284
285 instance Show Doc where
286 showsPrec prec doc cont = showDoc doc cont
287
288 -- | Renders the document as a string using the default style
289 render :: Doc -> String
290
291 -- | The general rendering interface
292 fullRender :: Mode -- ^Rendering mode
293 -> Int -- ^Line length
294 -> Float -- ^Ribbons per line
295 -> (TextDetails -> a -> a) -- ^What to do with text
296 -> a -- ^What to do at the end
297 -> Doc -- ^The document
298 -> a -- ^Result
299
300 -- | Render the document as a string using a specified style
301 renderStyle :: Style -> Doc -> String
302
303 -- | A rendering style
304 data Style
305 = Style { mode :: Mode -- ^ The rendering mode
306 , lineLength :: Int -- ^ Length of line, in chars
307 , ribbonsPerLine :: Float -- ^ Ratio of ribbon length to line length
308 }
309
310 -- | The default style (@mode=PageMode, lineLength=100, ribbonsPerLine=1.5@)
311 style :: Style
312 style = Style { lineLength = 100, ribbonsPerLine = 1.5, mode = PageMode }
313
314 -- | Rendering mode
315 data Mode = PageMode -- ^Normal
316 | ZigZagMode -- ^With zig-zag cuts
317 | LeftMode -- ^No indentation, infinitely long lines
318 | OneLineMode -- ^All on one line
319
320 -- ---------------------------------------------------------------------------
321 -- The Doc calculus
322
323 -- The Doc combinators satisfy the following laws:
324
325 {-
326 Laws for $$
327 ~~~~~~~~~~~
328 <a1> (x $$ y) $$ z = x $$ (y $$ z)
329 <a2> empty $$ x = x
330 <a3> x $$ empty = x
331
332 ...ditto $+$...
333
334 Laws for <>
335 ~~~~~~~~~~~
336 <b1> (x <> y) <> z = x <> (y <> z)
337 <b2> empty <> x = empty
338 <b3> x <> empty = x
339
340 ...ditto <+>...
341
342 Laws for text
343 ~~~~~~~~~~~~~
344 <t1> text s <> text t = text (s++t)
345 <t2> text "" <> x = x, if x non-empty
346
347 Laws for nest
348 ~~~~~~~~~~~~~
349 <n1> nest 0 x = x
350 <n2> nest k (nest k' x) = nest (k+k') x
351 <n3> nest k (x <> y) = nest k z <> nest k y
352 <n4> nest k (x $$ y) = nest k x $$ nest k y
353 <n5> nest k empty = empty
354 <n6> x <> nest k y = x <> y, if x non-empty
355
356 ** Note the side condition on <n6>! It is this that
357 ** makes it OK for empty to be a left unit for <>.
358
359 Miscellaneous
360 ~~~~~~~~~~~~~
361 <m1> (text s <> x) $$ y = text s <> ((text "" <> x)) $$
362 nest (-length s) y)
363
364 <m2> (x $$ y) <> z = x $$ (y <> z)
365 if y non-empty
366
367
368 Laws for list versions
369 ~~~~~~~~~~~~~~~~~~~~~~
370 <l1> sep (ps++[empty]++qs) = sep (ps ++ qs)
371 ...ditto hsep, hcat, vcat, fill...
372
373 <l2> nest k (sep ps) = sep (map (nest k) ps)
374 ...ditto hsep, hcat, vcat, fill...
375
376 Laws for oneLiner
377 ~~~~~~~~~~~~~~~~~
378 <o1> oneLiner (nest k p) = nest k (oneLiner p)
379 <o2> oneLiner (x <> y) = oneLiner x <> oneLiner y
380
381 You might think that the following verion of <m1> would
382 be neater:
383
384 <3 NO> (text s <> x) $$ y = text s <> ((empty <> x)) $$
385 nest (-length s) y)
386
387 But it doesn't work, for if x=empty, we would have
388
389 text s $$ y = text s <> (empty $$ nest (-length s) y)
390 = text s <> nest (-length s) y
391 -}
392
393 -- ---------------------------------------------------------------------------
394 -- Simple derived definitions
395
396 semi = char ';'
397 colon = char ':'
398 comma = char ','
399 space = char ' '
400 equals = char '='
401 lparen = char '('
402 rparen = char ')'
403 lbrack = char '['
404 rbrack = char ']'
405 lbrace = char '{'
406 rbrace = char '}'
407
408 int n = text (show n)
409 integer n = text (show n)
410 float n = text (show n)
411 double n = text (show n)
412 rational n = text (show n)
413 -- SIGBJORN wrote instead:
414 -- rational n = text (show (fromRationalX n))
415
416 quotes p = char '\'' <> p <> char '\''
417 doubleQuotes p = char '"' <> p <> char '"'
418 parens p = char '(' <> p <> char ')'
419 brackets p = char '[' <> p <> char ']'
420 braces p = char '{' <> p <> char '}'
421
422
423 hcat = foldr (<>) empty
424 hsep = foldr (<+>) empty
425 vcat = foldr ($$) empty
426
427 hang d1 n d2 = sep [d1, nest n d2]
428
429 punctuate p [] = []
430 punctuate p (d:ds) = go d ds
431 where
432 go d [] = [d]
433 go d (e:es) = (d <> p) : go e es
434
435 -- ---------------------------------------------------------------------------
436 -- The Doc data type
437
438 -- A Doc represents a *set* of layouts. A Doc with
439 -- no occurrences of Union or NoDoc represents just one layout.
440
441 -- | The abstract type of documents
442 data Doc
443 = Empty -- empty
444 | NilAbove Doc -- text "" $$ x
445 | TextBeside TextDetails !Int Doc -- text s <> x
446 | Nest !Int Doc -- nest k x
447 | Union Doc Doc -- ul `union` ur
448 | NoDoc -- The empty set of documents
449 | Beside Doc Bool Doc -- True <=> space between
450 | Above Doc Bool Doc -- True <=> never overlap
451
452 type RDoc = Doc -- RDoc is a "reduced Doc", guaranteed not to have a top-level Above or Beside
453
454
455 reduceDoc :: Doc -> RDoc
456 reduceDoc (Beside p g q) = beside p g (reduceDoc q)
457 reduceDoc (Above p g q) = above p g (reduceDoc q)
458 reduceDoc p = p
459
460
461 data TextDetails = Chr Char
462 | Str String
463 | PStr String
464 space_text = Chr ' '
465 nl_text = Chr '\n'
466
467 {-
468 Here are the invariants:
469
470 * The argument of NilAbove is never Empty. Therefore
471 a NilAbove occupies at least two lines.
472
473 * The arugment of @TextBeside@ is never @Nest@.
474
475
476 * The layouts of the two arguments of @Union@ both flatten to the same
477 string.
478
479 * The arguments of @Union@ are either @TextBeside@, or @NilAbove@.
480
481 * The right argument of a union cannot be equivalent to the empty set
482 (@NoDoc@). If the left argument of a union is equivalent to the
483 empty set (@NoDoc@), then the @NoDoc@ appears in the first line.
484
485 * An empty document is always represented by @Empty@. It can't be
486 hidden inside a @Nest@, or a @Union@ of two @Empty@s.
487
488 * The first line of every layout in the left argument of @Union@ is
489 longer than the first line of any layout in the right argument.
490 (1) ensures that the left argument has a first line. In view of
491 (3), this invariant means that the right argument must have at
492 least two lines.
493 -}
494
495 -- Arg of a NilAbove is always an RDoc
496 nilAbove_ p = NilAbove p
497
498 -- Arg of a TextBeside is always an RDoc
499 textBeside_ s sl p = TextBeside s sl p
500
501 -- Arg of Nest is always an RDoc
502 nest_ k p = Nest k p
503
504 -- Args of union are always RDocs
505 union_ p q = Union p q
506
507
508 -- Notice the difference between
509 -- * NoDoc (no documents)
510 -- * Empty (one empty document; no height and no width)
511 -- * text "" (a document containing the empty string;
512 -- one line high, but has no width)
513
514
515 -- ---------------------------------------------------------------------------
516 -- @empty@, @text@, @nest@, @union@
517
518 empty = Empty
519
520 isEmpty Empty = True
521 isEmpty _ = False
522
523 char c = textBeside_ (Chr c) 1 Empty
524 text s = case length s of {sl -> textBeside_ (Str s) sl Empty}
525 ptext s = case length s of {sl -> textBeside_ (PStr s) sl Empty}
526
527 nest k p = mkNest k (reduceDoc p) -- Externally callable version
528
529 -- mkNest checks for Nest's invariant that it doesn't have an Empty inside it
530 mkNest k _ | k `seq` False = undefined
531 mkNest k (Nest k1 p) = mkNest (k + k1) p
532 mkNest k NoDoc = NoDoc
533 mkNest k Empty = Empty
534 mkNest 0 p = p -- Worth a try!
535 mkNest k p = nest_ k p
536
537 -- mkUnion checks for an empty document
538 mkUnion Empty q = Empty
539 mkUnion p q = p `union_` q
540
541 -- ---------------------------------------------------------------------------
542 -- Vertical composition @$$@
543
544 p $$ q = Above p False q
545 p $+$ q = Above p True q
546
547 above :: Doc -> Bool -> RDoc -> RDoc
548 above (Above p g1 q1) g2 q2 = above p g1 (above q1 g2 q2)
549 above p@(Beside _ _ _) g q = aboveNest (reduceDoc p) g 0 (reduceDoc q)
550 above p g q = aboveNest p g 0 (reduceDoc q)
551
552 aboveNest :: RDoc -> Bool -> Int -> RDoc -> RDoc
553 -- Specfication: aboveNest p g k q = p $g$ (nest k q)
554
555 aboveNest _ _ k _ | k `seq` False = undefined
556 aboveNest NoDoc g k q = NoDoc
557 aboveNest (p1 `Union` p2) g k q = aboveNest p1 g k q `union_`
558 aboveNest p2 g k q
559
560 aboveNest Empty g k q = mkNest k q
561 aboveNest (Nest k1 p) g k q = nest_ k1 (aboveNest p g (k - k1) q)
562 -- p can't be Empty, so no need for mkNest
563
564 aboveNest (NilAbove p) g k q = nilAbove_ (aboveNest p g k q)
565 aboveNest (TextBeside s sl p) g k q = k1 `seq` textBeside_ s sl rest
566 where
567 k1 = k - sl
568 rest = case p of
569 Empty -> nilAboveNest g k1 q
570 other -> aboveNest p g k1 q
571
572
573 nilAboveNest :: Bool -> Int -> RDoc -> RDoc
574 -- Specification: text s <> nilaboveNest g k q
575 -- = text s <> (text "" $g$ nest k q)
576
577 nilAboveNest _ k _ | k `seq` False = undefined
578 nilAboveNest g k Empty = Empty -- Here's why the "text s <>" is in the spec!
579 nilAboveNest g k (Nest k1 q) = nilAboveNest g (k + k1) q
580
581 nilAboveNest g k q | (not g) && (k > 0) -- No newline if no overlap
582 = textBeside_ (Str (spaces k)) k q
583 | otherwise -- Put them really above
584 = nilAbove_ (mkNest k q)
585
586 -- ---------------------------------------------------------------------------
587 -- Horizontal composition @<>@
588
589 p <> q = Beside p False q
590 p <+> q = Beside p True q
591
592 beside :: Doc -> Bool -> RDoc -> RDoc
593 -- Specification: beside g p q = p <g> q
594
595 beside NoDoc g q = NoDoc
596 beside (p1 `Union` p2) g q = (beside p1 g q) `union_` (beside p2 g q)
597 beside Empty g q = q
598 beside (Nest k p) g q = nest_ k (beside p g q) -- p non-empty
599 beside p@(Beside p1 g1 q1) g2 q2
600 {- (A `op1` B) `op2` C == A `op1` (B `op2` C) iff op1 == op2
601 [ && (op1 == <> || op1 == <+>) ] -}
602 | g1 == g2 = beside p1 g1 (beside q1 g2 q2)
603 | otherwise = beside (reduceDoc p) g2 q2
604 beside p@(Above _ _ _) g q = beside (reduceDoc p) g q
605 beside (NilAbove p) g q = nilAbove_ (beside p g q)
606 beside (TextBeside s sl p) g q = textBeside_ s sl rest
607 where
608 rest = case p of
609 Empty -> nilBeside g q
610 other -> beside p g q
611
612
613 nilBeside :: Bool -> RDoc -> RDoc
614 -- Specification: text "" <> nilBeside g p
615 -- = text "" <g> p
616
617 nilBeside g Empty = Empty -- Hence the text "" in the spec
618 nilBeside g (Nest _ p) = nilBeside g p
619 nilBeside g p | g = textBeside_ space_text 1 p
620 | otherwise = p
621
622 -- ---------------------------------------------------------------------------
623 -- Separate, @sep@, Hughes version
624
625 -- Specification: sep ps = oneLiner (hsep ps)
626 -- `union`
627 -- vcat ps
628
629 sep = sepX True -- Separate with spaces
630 cat = sepX False -- Don't
631
632 sepX x [] = empty
633 sepX x (p:ps) = sep1 x (reduceDoc p) 0 ps
634
635
636 -- Specification: sep1 g k ys = sep (x : map (nest k) ys)
637 -- = oneLiner (x <g> nest k (hsep ys))
638 -- `union` x $$ nest k (vcat ys)
639
640 sep1 :: Bool -> RDoc -> Int -> [Doc] -> RDoc
641 sep1 g _ k ys | k `seq` False = undefined
642 sep1 g NoDoc k ys = NoDoc
643 sep1 g (p `Union` q) k ys = sep1 g p k ys
644 `union_`
645 (aboveNest q False k (reduceDoc (vcat ys)))
646
647 sep1 g Empty k ys = mkNest k (sepX g ys)
648 sep1 g (Nest n p) k ys = nest_ n (sep1 g p (k - n) ys)
649
650 sep1 g (NilAbove p) k ys = nilAbove_ (aboveNest p False k (reduceDoc (vcat ys)))
651 sep1 g (TextBeside s sl p) k ys = textBeside_ s sl (sepNB g p (k - sl) ys)
652
653 -- Specification: sepNB p k ys = sep1 (text "" <> p) k ys
654 -- Called when we have already found some text in the first item
655 -- We have to eat up nests
656
657 sepNB g (Nest _ p) k ys = sepNB g p k ys
658
659 sepNB g Empty k ys = oneLiner (nilBeside g (reduceDoc rest))
660 `mkUnion`
661 nilAboveNest False k (reduceDoc (vcat ys))
662 where
663 rest | g = hsep ys
664 | otherwise = hcat ys
665
666 sepNB g p k ys = sep1 g p k ys
667
668 -- ---------------------------------------------------------------------------
669 -- @fill@
670
671 fsep = fill True
672 fcat = fill False
673
674 -- Specification:
675 -- fill [] = empty
676 -- fill [p] = p
677 -- fill (p1:p2:ps) = oneLiner p1 <#> nest (length p1)
678 -- (fill (oneLiner p2 : ps))
679 -- `union`
680 -- p1 $$ fill ps
681
682 fill g [] = empty
683 fill g (p:ps) = fill1 g (reduceDoc p) 0 ps
684
685
686 fill1 :: Bool -> RDoc -> Int -> [Doc] -> Doc
687 fill1 g _ k ys | k `seq` False = undefined
688 fill1 g NoDoc k ys = NoDoc
689 fill1 g (p `Union` q) k ys = fill1 g p k ys
690 `union_`
691 (aboveNest q False k (fill g ys))
692
693 fill1 g Empty k ys = mkNest k (fill g ys)
694 fill1 g (Nest n p) k ys = nest_ n (fill1 g p (k - n) ys)
695
696 fill1 g (NilAbove p) k ys = nilAbove_ (aboveNest p False k (fill g ys))
697 fill1 g (TextBeside s sl p) k ys = textBeside_ s sl (fillNB g p (k - sl) ys)
698
699 fillNB g _ k ys | k `seq` False = undefined
700 fillNB g (Nest _ p) k ys = fillNB g p k ys
701 fillNB g Empty k [] = Empty
702 fillNB g Empty k (y:ys) = nilBeside g (fill1 g (oneLiner (reduceDoc y)) k1 ys)
703 `mkUnion`
704 nilAboveNest False k (fill g (y:ys))
705 where
706 k1 | g = k - 1
707 | otherwise = k
708
709 fillNB g p k ys = fill1 g p k ys
710
711
712 -- ---------------------------------------------------------------------------
713 -- Selecting the best layout
714
715 best :: Mode
716 -> Int -- Line length
717 -> Int -- Ribbon length
718 -> RDoc
719 -> RDoc -- No unions in here!
720
721 best OneLineMode w r p
722 = get p
723 where
724 get Empty = Empty
725 get NoDoc = NoDoc
726 get (NilAbove p) = nilAbove_ (get p)
727 get (TextBeside s sl p) = textBeside_ s sl (get p)
728 get (Nest k p) = get p -- Elide nest
729 get (p `Union` q) = first (get p) (get q)
730
731 best mode w r p
732 = get w p
733 where
734 get :: Int -- (Remaining) width of line
735 -> Doc -> Doc
736 get w _ | w==0 && False = undefined
737 get w Empty = Empty
738 get w NoDoc = NoDoc
739 get w (NilAbove p) = nilAbove_ (get w p)
740 get w (TextBeside s sl p) = textBeside_ s sl (get1 w sl p)
741 get w (Nest k p) = nest_ k (get (w - k) p)
742 get w (p `Union` q) = nicest w r (get w p) (get w q)
743
744 get1 :: Int -- (Remaining) width of line
745 -> Int -- Amount of first line already eaten up
746 -> Doc -- This is an argument to TextBeside => eat Nests
747 -> Doc -- No unions in here!
748
749 get1 w _ _ | w==0 && False = undefined
750 get1 w sl Empty = Empty
751 get1 w sl NoDoc = NoDoc
752 get1 w sl (NilAbove p) = nilAbove_ (get (w - sl) p)
753 get1 w sl (TextBeside t tl p) = textBeside_ t tl (get1 w (sl + tl) p)
754 get1 w sl (Nest k p) = get1 w sl p
755 get1 w sl (p `Union` q) = nicest1 w r sl (get1 w sl p)
756 (get1 w sl q)
757
758 nicest w r p q = nicest1 w r 0 p q
759 nicest1 w r sl p q | fits ((w `minn` r) - sl) p = p
760 | otherwise = q
761
762 fits :: Int -- Space available
763 -> Doc
764 -> Bool -- True if *first line* of Doc fits in space available
765
766 fits n p | n < 0 = False
767 fits n NoDoc = False
768 fits n Empty = True
769 fits n (NilAbove _) = True
770 fits n (TextBeside _ sl p) = fits (n - sl) p
771
772 minn x y | x < y = x
773 | otherwise = y
774
775 -- @first@ and @nonEmptySet@ are similar to @nicest@ and @fits@, only simpler.
776 -- @first@ returns its first argument if it is non-empty, otherwise its second.
777
778 first p q | nonEmptySet p = p
779 | otherwise = q
780
781 nonEmptySet NoDoc = False
782 nonEmptySet (p `Union` q) = True
783 nonEmptySet Empty = True
784 nonEmptySet (NilAbove p) = True -- NoDoc always in first line
785 nonEmptySet (TextBeside _ _ p) = nonEmptySet p
786 nonEmptySet (Nest _ p) = nonEmptySet p
787
788 -- @oneLiner@ returns the one-line members of the given set of @Doc@s.
789
790 oneLiner :: Doc -> Doc
791 oneLiner NoDoc = NoDoc
792 oneLiner Empty = Empty
793 oneLiner (NilAbove p) = NoDoc
794 oneLiner (TextBeside s sl p) = textBeside_ s sl (oneLiner p)
795 oneLiner (Nest k p) = nest_ k (oneLiner p)
796 oneLiner (p `Union` q) = oneLiner p
797
798
799 -- ---------------------------------------------------------------------------
800 -- Displaying the best layout
801
802 renderStyle style doc
803 = fullRender (mode style)
804 (lineLength style)
805 (ribbonsPerLine style)
806 string_txt
807 ""
808 doc
809
810 render doc = showDoc doc ""
811 showDoc doc rest = fullRender PageMode 100 1.5 string_txt rest doc
812
813 string_txt (Chr c) s = c:s
814 string_txt (Str s1) s2 = s1 ++ s2
815 string_txt (PStr s1) s2 = s1 ++ s2
816
817
818 fullRender OneLineMode _ _ txt end doc = easy_display space_text txt end (reduceDoc doc)
819 fullRender LeftMode _ _ txt end doc = easy_display nl_text txt end (reduceDoc doc)
820
821 fullRender mode line_length ribbons_per_line txt end doc
822 = display mode line_length ribbon_length txt end best_doc
823 where
824 best_doc = best mode hacked_line_length ribbon_length (reduceDoc doc)
825
826 hacked_line_length, ribbon_length :: Int
827 ribbon_length = round (fromIntegral line_length / ribbons_per_line)
828 hacked_line_length = case mode of { ZigZagMode -> maxBound; other -> line_length }
829
830 display mode page_width ribbon_width txt end doc
831 = case page_width - ribbon_width of { gap_width ->
832 case gap_width `quot` 2 of { shift ->
833 let
834 lay k _ | k `seq` False = undefined
835 lay k (Nest k1 p) = lay (k + k1) p
836 lay k Empty = end
837
838 lay k (NilAbove p) = nl_text `txt` lay k p
839
840 lay k (TextBeside s sl p)
841 = case mode of
842 ZigZagMode | k >= gap_width
843 -> nl_text `txt` (
844 Str (multi_ch shift '/') `txt` (
845 nl_text `txt` (
846 lay1 (k - shift) s sl p)))
847
848 | k < 0
849 -> nl_text `txt` (
850 Str (multi_ch shift '\\') `txt` (
851 nl_text `txt` (
852 lay1 (k + shift) s sl p )))
853
854 other -> lay1 k s sl p
855
856 lay1 k _ sl _ | k+sl `seq` False = undefined
857 lay1 k s sl p = Str (indent k) `txt` (s `txt` lay2 (k + sl) p)
858
859 lay2 k _ | k `seq` False = undefined
860 lay2 k (NilAbove p) = nl_text `txt` lay k p
861 lay2 k (TextBeside s sl p) = s `txt` (lay2 (k + sl) p)
862 lay2 k (Nest _ p) = lay2 k p
863 lay2 k Empty = end
864 in
865 lay 0 doc
866 }}
867
868 cant_fail = error "easy_display: NoDoc"
869 easy_display nl_text txt end doc
870 = lay doc cant_fail
871 where
872 lay NoDoc no_doc = no_doc
873 lay (Union p q) no_doc = {- lay p -} (lay q cant_fail) -- Second arg can't be NoDoc
874 lay (Nest k p) no_doc = lay p no_doc
875 lay Empty no_doc = end
876 lay (NilAbove p) no_doc = nl_text `txt` lay p cant_fail -- NoDoc always on first line
877 lay (TextBeside s sl p) no_doc = s `txt` lay p no_doc
878
879 -- OLD version: we shouldn't rely on tabs being 8 columns apart in the output.
880 -- indent n | n >= 8 = '\t' : indent (n - 8)
881 -- | otherwise = spaces n
882 indent n = spaces n
883
884 multi_ch 0 ch = ""
885 multi_ch n ch = ch : multi_ch (n - 1) ch
886
887 -- (spaces n) generates a list of n spaces
888 --
889 -- It should never be called with 'n' < 0, but that can happen for reasons I don't understand
890 -- Here's a test case:
891 -- ncat x y = nest 4 $ cat [ x, y ]
892 -- d1 = foldl1 ncat $ take 50 $ repeat $ char 'a'
893 -- d2 = parens $ sep [ d1, text "+" , d1 ]
894 -- main = print d2
895 -- I don't feel motivated enough to find the Real Bug, so meanwhile we just test for n<=0
896 spaces n | n <= 0 = ""
897 | otherwise = ' ' : spaces (n - 1)
898
899 {- Comments from Johannes Waldmann about what the problem might be:
900
901 In the example above, d2 and d1 are deeply nested, but `text "+"' is not,
902 so the layout function tries to "out-dent" it.
903
904 when I look at the Doc values that are generated, there are lots of
905 Nest constructors with negative arguments. see this sample output of
906 d1 (obtained with hugs, :s -u)
907
908 tBeside (TextDetails_Chr 'a') 1 Doc_Empty) (Doc_NilAbove (Doc_Nest
909 (-241) (Doc_TextBeside (TextDetails_Chr 'a') 1 Doc_Empty)))))
910 (Doc_NilAbove (Doc_Nest (-236) (Doc_TextBeside (TextDetails_Chr 'a') 1
911 (Doc_NilAbove (Doc_Nest (-5) (Doc_TextBeside (TextDetails_Chr 'a') 1
912 Doc_Empty)))))))) (Doc_NilAbove (Doc_Nest (-231) (Doc_TextBeside
913 (TextDetails_Chr 'a') 1 (Doc_NilAbove (Doc_Nest (-5) (Doc_TextBeside
914 (TextDetails_Chr 'a') 1 (Doc_NilAbove (Doc_Nest (-5) (Doc_TextBeside
915 (TextDetails_Chr 'a') 1 Doc_Empty))))))))))) (Doc_NilAbove (Doc_Nest
916 -}