layout/format.bend
bend-lint-rules@0.2.1.0 · 2,454 lines · 0 definitions
1# layout/format: a file that differs from its formatted text, where it first differs. One formatter
2# owns indentation, spacing, declaration gaps and wrapping.
3# Fix: writes the formatted text.
4# Options: tabWidth (2); wrapAtWidth (100, or "never": no optional wrapping); endOfLine ("lf",
5# "crlf", or "preserve": the file's first, or LF if it has none).
6# Heuristics: it reads tokens, not Bend's syntax, and changes only blanks, line breaks and the space
7# after a comment's #, so a fix keeps the program. A form its token rules do not know can come out
8# laid out oddly.
9#
10# A String is a JS string at run time: ++ is native, and a match on SCon reads one character and slices off the rest.
11# So the lexer reads the text once, tokens carry what the layout asks of them, and docs carry widths.
12#
13# Each loop is go(input, state): a step that does not recurse makes the next state, so every loop recurses on a smaller
14# part of its input. The lexer and the printer recurse on fuel instead: more steps than they can need.
15import Base
16import bend-lint-api@0.2.1.0/linter/lint.bend as Lint
17import bend-lint-api@0.2.1.0/shared/strings.bend as Strings
18
19# Types
20# =====
21type Lex is Data:
22 Code{}
23 Literal{}
24 Comment{}
25 Newline{}
26
27# The spellings the layout tells apart; Word{} is any other, Inc{} ends in n+.
28type Sym is Data:
29 Word{}
30 Comma{}
31 Semi{}
32 Colon{}
33 Ques{}
34 Bang{}
35 Dot{}
36 Tilde{}
37 At{}
38 Bslash{}
39 Plus{}
40 Minus{}
41 Star{}
42 Slash{}
43 Pct{}
44 Caret{}
45 Amp{}
46 Bar{}
47 Lpar{}
48 Rpar{}
49 Lbrk{}
50 Rbrk{}
51 Lbrc{}
52 Rbrc{}
53 Lt{}
54 Gt{}
55 Assign{}
56 Arrow{}
57 Concat{}
58 AndAnd{}
59 OrOr{}
60 Meet{}
61 BitAnd{}
62 BitOr{}
63 BitXor{}
64 Shl{}
65 Shr{}
66 Cons{}
67 Le{}
68 Ge{}
69 EqEq{}
70 Ne{}
71 Inc{}
72 KwReturn{}
73 KwCase{}
74 KwMatch{}
75 KwFor{}
76 KwExs{}
77 KwIs{}
78 KwDo{}
79 KwDef{}
80 KwType{}
81 KwLaw{}
82 KwImport{}
83
84# Offsets and columns count code points. name: whether the text matches
85# /^[\w.]+$/. tail: for a literal across lines, how many characters follow
86# its last newline. from: the text from the token on.
87type Token is Data:
88 Token{
89 text: String,
90 beg: U32,
91 end: U32,
92 col: U32,
93 lex: Lex,
94 sym: Sym,
95 name: Bool,
96 tail: Maybe<&2, U32>,
97 from: String
98 }
99
100# What the lexer read: the end of the text, one whitespace character, or a
101# token; with the text after it and its last character.
102type Scan is Data:
103 Eof{}
104 Blank{rest: String, last: Char, len: U32}
105 Tok{
106 text: String,
107 len: U32,
108 rest: String,
109 last: Char,
110 lex: Lex,
111 sym: Sym,
112 name: Bool,
113 tail: Maybe<&2, U32>
114 }
115
116# Characters read by run: their text and count, the text after them, and
117# the last one.
118type Run is Data:
119 Run{text: String, len: U32, rest: String, last: Char}
120
121# What the next character of a literal is: inside it, escaped, or past it.
122type Quote is Data:
123 Inside{}
124 Escaped{}
125 Shut{}
126
127type Class is Data:
128 IsBlank{}
129 IsHead{}
130 IsDigit{}
131 IsOther{}
132
133type Tree is Data:
134 Leaf{token: Token}
135 Group{open: Token, close: Token, children: List<&2, Tree>, block: Bool}
136
137# A group the tree has opened: the sym that closes it, children last first,
138# and whether a child is a token that makes it a block.
139type Open is Data:
140 Open{token: Token, close: Sym, children: List<&2, Tree>, block: Bool}
141
142type Built is Data:
143 Built{stack: List<&2, Open>, root: List<&2, Tree>}
144
145# A doc printed flat: its width up to the first break it cannot lose (a
146# hard line, or a newline in text), and whether it has one.
147type Flat is Data:
148 Flat{width: U32, brk: Bool}
149
150# Str: text without its trailing whitespace, that whitespace, the width, and
151# how many characters follow the last newline, if any. Line: the text and
152# width when flat. Grp, Nest and Cat keep their Flat; Cat its docs last
153# first.
154type Doc is Data:
155 Str{body: String, blank: String, width: U32, tail: Maybe<&2, U32>}
156 Line{flat: String, width: U32, hard: Bool, offset: U32}
157 Grp{doc: Doc, flat: Flat}
158 Nest{doc: Doc, amount: U32, flat: Flat}
159 Cat{docs: List<&2, Doc>, flat: Flat}
160
161type Measure is Data:
162 Measure{width: U32, tail: Maybe<&2, U32>, keep: U32}
163
164# How build reads a list of trees: a record, a group's children as a raw
165# block, as comma-separated parts, or a group with no children. A block is
166# an embedded statement body: its line and column boundaries mean
167# something, so it keeps them, while ordinary calls inside it still wrap.
168type Mode is Data:
169 Plain{}
170 Block{open: Token, close: Token}
171 Parts{open: Token, close: Token}
172 Bare{open: Token, close: Token}
173
174# A sequence being built: its docs last first, the docs of an operator
175# chain, the previous tree, the indentation levels deepest first, and
176# whether the rest of a run of newlines is skipped.
177type Seq is Data:
178 Seq{
179 out: List<&2, Doc>,
180 chain: Maybe<&2, List<&2, Doc>>,
181 prev: Maybe<&2, Tree>,
182 levels: List<&2, U32>,
183 skip: Bool
184 }
185
186# A comma-separated part: whether it holds only newlines, and whether its
187# last other tree ends in a comment.
188type Part is Data:
189 Part{doc: Doc, newlines: Bool, comment: Bool}
190
191# A group's parts so far, last first; the current part's flags; whether the
192# next tree ends the current part; whether a child starts with a comment.
193type Split is Data:
194 Split{parts: List<&2, Part>, newlines: Bool, comment: Bool, attach: Bool, comments: Bool}
195
196type State is Data:
197 State{seq: Seq, split: Split}
198
199# What a child of a group does to its parts.
200type Act is Data:
201 Add{}
202 AddAttach{}
203 AddEnd{}
204 End{}
205
206type Job is Data:
207 Job{doc: Doc, indent: U32, flat: Bool}
208
209# The printer: jobs left, the text so far as chunks last first, the
210# whitespace not written yet (dropped at a break or at the end), and the
211# column.
212type Print is Data:
213 Print{jobs: List<&2, Job>, out: List<&2, String>, pending: String, col: U32}
214
215type Fit is Data:
216 Go{jobs: List<&2, Job>, left: U32}
217 Stop{fits: Bool}
218
219# Records so far, last first, and the current record, last first.
220type Recs is Data:
221 Recs{done: List<&2, List<&2, Tree>>, current: List<&2, Tree>}
222
223# A declaration record cut after its first colon, if it has one.
224type Cut is Data:
225 Cut{before: List<&2, Tree>, after: List<&2, Tree>, found: Bool}
226
227type Hash is Data:
228 Hash{before: String, comment: Maybe<&2, String>}
229
230# The rows so far, last first, and what the next record needs from them.
231type Rows is Data:
232 Rows{levels: List<&2, U32>, blank: Bool, unsafe: Bool, comment: Bool, rows: List<&2, String>}
233
234# width: None{} for "never".
235type Settings is Data:
236 Settings{tab: U32, width: Maybe<&2, U32>, newline: String}
237
238# Constants
239# =========
240
241# Fuel per code point of a record: more printer steps than any doc needs.
242def fuel_per_point() -> U32:
243 64
244
245# Functions
246# =========
247
248# Text
249# ----
250
251# Whitespace, as Bend's parser reads it.
252def is_blank(+c: Char) -> Bool:
253 match c:
254 case ' ':
255 True{}
256 case '\t':
257 True{}
258 case '\n':
259 True{}
260 case '\r':
261 True{}
262 case _:
263 False{}
264
265def is_head(+c: Char) -> Bool:
266 Char.is_alpha(c) || Char.is_eq(c, '_')
267
268def is_word(+c: Char) -> Bool:
269 is_head(c) || Char.is_digit(c)
270
271def head_passes(~test: @+c: Char -> Bool, s: String) -> Bool:
272 match s:
273 case SNil{}:
274 False{}
275 case SCon{c, rest}:
276 test(c)
277
278def char(c: Char) -> String:
279 SCon{c, SNil{}}
280
281def drop_blank.go(s: String, blank: Bool) -> String:
282 match s blank:
283 case SNil{} _:
284 SNil{}
285 case SCon{c, rest} False{}:
286 SCon{c, rest}
287 case SCon{c, +rest} True{}:
288 drop_blank.go(rest, head_passes(~is_blank, rest))
289
290def drop_blank(+s: String) -> String:
291 drop_blank.go(s, head_passes(~is_blank, s))
292
293def trim_end(s: String) -> String:
294 String.reverse(drop_blank(String.reverse(s)))
295
296def pad(n: Nat, out: String) -> String:
297 match n:
298 case 0n:
299 out
300 case 1n+k:
301 pad(k, SCon{' ', out})
302
303# Base's String.eq is not tail-recursive, so long text would overflow the stack; is_same is.
304# ok: whether the characters so far were the same.
305def is_same.go(a: String, b: String, ok: Bool) -> Bool:
306 match a b ok:
307 case SCon{x, xs} SCon{y, ys} True{}:
308 is_same.go(xs, ys, Char.is_eq(x, y))
309 case SNil{} SNil{} done:
310 done
311 case _ _ _:
312 False{}
313
314def is_same(a: String, b: String) -> Bool:
315 is_same.go(a, b, True{})
316
317# Lexer
318# -----
319#
320# A branch that uses a string or a character its match took apart gets it
321# rebuilt: a string as character + rest, a rope that the next read copies
322# whole, and a character bit by bit. So a def that needs the whole value in
323# a branch takes it twice, as keep, and matches the other copy.
324def is_not_line_end(s: String) -> Bool:
325 match s:
326 case SNil{}:
327 False{}
328 case SCon{'\n', rest}:
329 False{}
330 case SCon{'\r', rest}:
331 False{}
332 case _:
333 True{}
334
335def is_digit_at(s: String) -> Bool:
336 match s:
337 case SNil{}:
338 False{}
339 case SCon{c, rest}:
340 Char.is_digit(c)
341
342def is_ident_at.of(dot: Bool, c: Char, rest: String) -> Bool:
343 match dot:
344 case True{}:
345 head_passes(~is_head, rest)
346 case False{}:
347 is_word(c)
348
349# A word character, or a dot before the start of a word.
350def is_ident_at(s: String) -> Bool:
351 match s:
352 case SNil{}:
353 False{}
354 case SCon{+c, rest}:
355 is_ident_at.of(Char.is_eq(c, '.'), c, rest)
356
357# `scan`, with the text so far last character first: on the C lane, ++ copies its left side.
358def scan.go(
359 ~test: String -> Bool,
360 s: String,
361 keep: String,
362 ok: Bool,
363 text: String,
364 n: U32,
365 last: Char
366) -> Run:
367 match s ok:
368 case SCon{+c, +rest} True{}:
369 scan.go(~test, rest, rest, test(rest), SCon{c, text}, (n + 1: U32), c)
370 case _ _:
371 Run{String.reverse(text), n, keep, last}
372
373# Reads characters while test holds, after text; test sees the text from
374# each character on, so it can look ahead.
375def scan(
376 ~test: String -> Bool,
377 s: String,
378 keep: String,
379 ok: Bool,
380 text: String,
381 n: U32,
382 last: Char
383) -> Run:
384 scan.go(~test, s, keep, ok, String.reverse(text), n, last)
385
386def tok(r: Run, sym: Sym, name: Bool) -> Scan:
387 match r:
388 case Run{text, len, rest, last}:
389 Tok{text, len, rest, last, Code{}, sym, name, None{}}
390
391def fixed(text: String, len: U32, rest: String, last: Char, sym: Sym) -> Scan:
392 Tok{text, len, rest, last, Code{}, sym, False{}, None{}}
393
394def keyword.is(text: String, word: String, sym: Sym) -> Sym:
395 Bool.pick(Sym, is_same(text, word), sym, Word{})
396
397def keyword(text: String, +keep: String) -> Sym:
398 match text:
399 case SCon{'r', rest}:
400 keyword.is(keep, "return", KwReturn{})
401 case SCon{'c', rest}:
402 keyword.is(keep, "case", KwCase{})
403 case SCon{'m', rest}:
404 keyword.is(keep, "match", KwMatch{})
405 case SCon{'f', rest}:
406 keyword.is(keep, "for", KwFor{})
407 case SCon{'e', rest}:
408 keyword.is(keep, "exs", KwExs{})
409 case SCon{'i', rest}:
410 Bool.pick(Sym, is_same(keep, "is"), KwIs{}, keyword.is(keep, "import", KwImport{}))
411 case SCon{'d', rest}:
412 Bool.pick(Sym, is_same(keep, "do"), KwDo{}, keyword.is(keep, "def", KwDef{}))
413 case SCon{'t', rest}:
414 keyword.is(keep, "type", KwType{})
415 case SCon{'l', rest}:
416 keyword.is(keep, "law", KwLaw{})
417 case _:
418 Word{}
419
420def ident(r: Run) -> Scan:
421 match r:
422 case Run{+text, len, rest, last}:
423 Tok{text, len, rest, last, Code{}, keyword(text, text), True{}, None{}}
424
425def number.tok(r: Run, inc: Bool, name: Bool) -> Scan:
426 tok(r, Bool.pick(Sym, inc, Inc{}, Word{}), name)
427
428def exponent.digits(
429 go: Bool,
430 +rest: String,
431 mark: String,
432 k: U32,
433 keep: String,
434 text: String,
435 n: U32,
436 last: Char,
437 signed: Bool
438) -> Scan:
439 match go:
440 case True{}:
441 number.tok(
442 scan(~is_digit_at, rest, rest, True{}, text ++ mark, (n + k: U32), last),
443 False{},
444 Bool.not(signed)
445 )
446 case False{}:
447 number.tok(Run{text, n, keep, last}, False{}, True{})
448
449def exponent.sign(
450 rest: String,
451 +again: String,
452 mark: String,
453 keep: String,
454 text: String,
455 n: U32,
456 last: Char
457) -> Scan:
458 match rest:
459 case SCon{'+', +r}:
460 exponent.digits(is_digit_at(r), r, mark ++ "+", 2, keep, text, n, last, True{})
461 case SCon{'-', +r}:
462 exponent.digits(is_digit_at(r), r, mark ++ "-", 2, keep, text, n, last, True{})
463 case _:
464 exponent.digits(is_digit_at(again), again, mark, 1, keep, text, n, last, False{})
465
466def exponent.of(s: String, keep: String, text: String, n: U32, last: Char) -> Scan:
467 match s:
468 case SCon{'e', +rest}:
469 exponent.sign(rest, rest, "e", keep, text, n, last)
470 case SCon{'E', +rest}:
471 exponent.sign(rest, rest, "E", keep, text, n, last)
472 case _:
473 number.tok(Run{text, n, keep, last}, False{}, True{})
474
475# After a fraction: [eE][+-]?\d+, if it is there.
476def exponent(r: Run) -> Scan:
477 match r:
478 case Run{text, n, +s, last}:
479 exponent.of(s, s, text, n, last)
480
481def fraction(go: Bool, +rest: String, keep: String, text: String, n: U32, last: Char) -> Scan:
482 match go:
483 case True{}:
484 exponent(scan(~is_digit_at, rest, rest, True{}, text ++ ".", (n + 1: U32), '.'))
485 case False{}:
486 number.tok(Run{text, n, keep, last}, False{}, True{})
487
488# After "n": n+, or n alone.
489def number.inc(s: String, keep: String, text: String, n: U32) -> Scan:
490 match s:
491 case SCon{'+', rest}:
492 number.tok(Run{text ++ "n+", (n + 2: U32), rest, '+'}, True{}, False{})
493 case _:
494 number.tok(Run{text ++ "n", (n + 1: U32), keep, 'n'}, False{}, True{})
495
496def number.of(s: String, keep: String, text: String, n: U32, last: Char) -> Scan:
497 match s:
498 case SCon{'n', +rest}:
499 number.inc(rest, rest, text, n)
500 case SCon{'.', +rest}:
501 fraction(is_digit_at(rest), rest, keep, text, n, last)
502 case _:
503 number.tok(Run{text, n, keep, last}, False{}, True{})
504
505# After a number's digits: n, n+, or a fraction and its exponent.
506def number(r: Run) -> Scan:
507 match r:
508 case Run{text, n, +s, last}:
509 number.of(s, s, text, n, last)
510
511def tail_after(c: Char, tail: Maybe<&2, U32>) -> Maybe<&2, U32>:
512 match c tail:
513 case '\n' _:
514 Some{0}
515 case _ Some{t}:
516 Some{(t + 1: U32)}
517 case _ None{}:
518 None{}
519
520def quote_after(+c: Char, quote: Char) -> Quote:
521 Bool.pick(
522 Quote,
523 Char.is_eq(c, '\\'),
524 Escaped{},
525 Bool.pick(Quote, Char.is_eq(c, quote), Shut{}, Inside{})
526 )
527
528# `literal`, with the text so far last character first: on the C lane, ++ copies its left side.
529def literal.go(
530 s: String,
531 keep: String,
532 +quote: Char,
533 at: Quote,
534 text: String,
535 n: U32,
536 last: Char,
537 tail: Maybe<&2, U32>
538) -> Scan:
539 match s at:
540 case SCon{+c, +rest} Inside{}:
541 literal.go(
542 rest,
543 rest,
544 quote,
545 quote_after(c, quote),
546 SCon{c, text},
547 (n + 1: U32),
548 c,
549 tail_after(c, tail)
550 )
551 case SCon{+c, +rest} Escaped{}:
552 literal.go(rest, rest, quote, Inside{}, SCon{c, text}, (n + 1: U32), c, tail_after(c, tail))
553 case _ _:
554 Tok{String.reverse(text), n, keep, last, Literal{}, Word{}, False{}, tail}
555
556def literal(
557 s: String,
558 keep: String,
559 quote: Char,
560 at: Quote,
561 text: String,
562 n: U32,
563 last: Char,
564 tail: Maybe<&2, U32>
565) -> Scan:
566 literal.go(s, keep, quote, at, String.reverse(text), n, last, tail)
567
568def class_of(+c: Char) -> Class:
569 Bool.pick(
570 Class,
571 is_blank(c),
572 IsBlank{},
573 Bool.pick(Class, is_head(c), IsHead{}, Bool.pick(Class, Char.is_digit(c), IsDigit{}, IsOther{}))
574 )
575
576def single(c: Char) -> Sym:
577 match c:
578 case ',':
579 Comma{}
580 case ';':
581 Semi{}
582 case ':':
583 Colon{}
584 case '?':
585 Ques{}
586 case '~':
587 Tilde{}
588 case '@':
589 At{}
590 case '\\':
591 Bslash{}
592 case '*':
593 Star{}
594 case '/':
595 Slash{}
596 case '%':
597 Pct{}
598 case '^':
599 Caret{}
600 case '(':
601 Lpar{}
602 case ')':
603 Rpar{}
604 case '[':
605 Lbrk{}
606 case ']':
607 Rbrk{}
608 case '{':
609 Lbrc{}
610 case '}':
611 Rbrc{}
612 case _:
613 Word{}
614
615def lex.other(class: Class, +c: Char, +rest: String) -> Scan:
616 match class:
617 case IsBlank{}:
618 Blank{rest, c, 1}
619 case IsHead{}:
620 ident(scan(~is_ident_at, rest, rest, is_ident_at(rest), char(c), 1, c))
621 case IsDigit{}:
622 number(scan(~is_digit_at, rest, rest, is_digit_at(rest), char(c), 1, c))
623 case IsOther{}:
624 fixed(char(c), 1, rest, c, single(c))
625
626def lex.char(s: String) -> Scan:
627 match s:
628 case SNil{}:
629 Eof{}
630 case SCon{+c, rest}:
631 lex.other(class_of(c), c, rest)
632
633def lex.dash(word: Bool, keep: String, r: String) -> Scan:
634 match word:
635 case True{}:
636 fixed("<", 1, keep, '<', Lt{})
637 case False{}:
638 fixed("<-", 2, r, '-', Word{})
639
640def lex.lt(rest: String, keep: String, +prev: Char) -> Scan:
641 match rest:
642 case SCon{'&', SCon{'>', r}}:
643 fixed("<&>", 3, r, '>', Meet{})
644 case SCon{'-', r}:
645 lex.dash(is_word(prev) || Char.is_eq(prev, '.'), keep, r)
646 case SCon{'=', r}:
647 fixed("<=", 2, r, '=', Le{})
648 case SCon{'<', r}:
649 fixed("<<", 2, r, '<', Shl{})
650 case SCon{'>', r}:
651 fixed("<>", 2, r, '>', Cons{})
652 case _:
653 fixed("<", 1, keep, '<', Lt{})
654
655def lex.shr(blank: Bool, keep: String, r: String) -> Scan:
656 match blank:
657 case True{}:
658 fixed(">>", 2, r, '>', Shr{})
659 case False{}:
660 fixed(">", 1, keep, '>', Gt{})
661
662def lex.gt(rest: String, keep: String, prev: Char) -> Scan:
663 match rest:
664 case SCon{'>', r}:
665 lex.shr(is_blank(prev), keep, r)
666 case SCon{'=', r}:
667 fixed(">=", 2, r, '=', Ge{})
668 case _:
669 fixed(">", 1, keep, '>', Gt{})
670
671def lex.dot(rest: String, keep: String) -> Scan:
672 match rest:
673 case SCon{'&', SCon{'.', r}}:
674 fixed(".&.", 3, r, '.', BitAnd{})
675 case SCon{'|', SCon{'.', r}}:
676 fixed(".|.", 3, r, '.', BitOr{})
677 case SCon{'^', SCon{'.', r}}:
678 fixed(".^.", 3, r, '.', BitXor{})
679 case _:
680 Tok{".", 1, keep, '.', Code{}, Dot{}, True{}, None{}}
681
682def lex.minus(rest: String, keep: String) -> Scan:
683 match rest:
684 case SCon{'>', r}:
685 fixed("->", 2, r, '>', Arrow{})
686 case _:
687 fixed("-", 1, keep, '-', Minus{})
688
689def lex.eq(rest: String, keep: String) -> Scan:
690 match rest:
691 case SCon{'=', r}:
692 fixed("==", 2, r, '=', EqEq{})
693 case SCon{'>', r}:
694 fixed("=>", 2, r, '>', Word{})
695 case _:
696 fixed("=", 1, keep, '=', Assign{})
697
698def lex.bang(rest: String, keep: String) -> Scan:
699 match rest:
700 case SCon{'=', r}:
701 fixed("!=", 2, r, '=', Ne{})
702 case _:
703 fixed("!", 1, keep, '!', Bang{})
704
705def lex.amp(rest: String, keep: String) -> Scan:
706 match rest:
707 case SCon{'&', r}:
708 fixed("&&", 2, r, '&', AndAnd{})
709 case SCon{'0', r}:
710 fixed("&0", 2, r, '0', Word{})
711 case SCon{'1', r}:
712 fixed("&1", 2, r, '1', Word{})
713 case SCon{'2', r}:
714 fixed("&2", 2, r, '2', Word{})
715 case _:
716 fixed("&", 1, keep, '&', Amp{})
717
718def lex.bar(rest: String, keep: String) -> Scan:
719 match rest:
720 case SCon{'|', r}:
721 fixed("||", 2, r, '|', OrOr{})
722 case _:
723 fixed("|", 1, keep, '|', Bar{})
724
725def lex.plus(rest: String, keep: String) -> Scan:
726 match rest:
727 case SCon{'+', r}:
728 fixed("++", 2, r, '+', Concat{})
729 case _:
730 fixed("+", 1, keep, '+', Plus{})
731
732def comment(r: Run) -> Scan:
733 match r:
734 case Run{text, len, rest, last}:
735 Tok{text, len, rest, last, Comment{}, Word{}, False{}, None{}}
736
737# A run of whitespace on one line.
738def blanks(s: String, keep: String, n: U32, last: Char) -> Scan:
739 match s:
740 case SCon{' ', +rest}:
741 blanks(rest, rest, (n + 1: U32), ' ')
742 case SCon{'\t', +rest}:
743 blanks(rest, rest, (n + 1: U32), '\t')
744 case SCon{'\r', +rest}:
745 blanks(rest, rest, (n + 1: U32), '\r')
746 case _:
747 Blank{keep, last, n}
748
749# The token at the start of s; prev is the character before it. Literals
750# and comment text are kept verbatim: a quote or # inside a string is not
751# syntax. Successor notation (1n+) and quantities (&2) are single tokens.
752def lex.token(s: String, keep: String, prev: Char) -> Scan:
753 match s:
754 case SNil{}:
755 Eof{}
756 case SCon{' ', +rest}:
757 blanks(rest, rest, 1, ' ')
758 case SCon{'\n', rest}:
759 Tok{"\n", 1, rest, '\n', Newline{}, Word{}, False{}, None{}}
760 case SCon{'#', +rest}:
761 comment(scan(~is_not_line_end, rest, rest, is_not_line_end(rest), "#", 1, '#'))
762 case SCon{'"', +rest}:
763 literal(rest, rest, '"', Inside{}, "\"", 1, '"', None{})
764 case SCon{'\'', +rest}:
765 literal(rest, rest, '\'', Inside{}, "'", 1, '\'', None{})
766 case SCon{'<', +rest}:
767 lex.lt(rest, rest, prev)
768 case SCon{'>', +rest}:
769 lex.gt(rest, rest, prev)
770 case SCon{'.', +rest}:
771 lex.dot(rest, rest)
772 case SCon{'-', +rest}:
773 lex.minus(rest, rest)
774 case SCon{'=', +rest}:
775 lex.eq(rest, rest)
776 case SCon{'!', +rest}:
777 lex.bang(rest, rest)
778 case SCon{'&', +rest}:
779 lex.amp(rest, rest)
780 case SCon{'|', +rest}:
781 lex.bar(rest, rest)
782 case SCon{'+', +rest}:
783 lex.plus(rest, rest)
784 case _:
785 lex.char(keep)
786
787def next_col(lex: Lex, tail: Maybe<&2, U32>, col: U32, len: U32) -> U32:
788 match lex tail:
789 case Newline{} _:
790 0
791 case _ Some{t}:
792 t
793 case _ None{}:
794 (col + len: U32)
795
796# Each step reads at least one character, so the text's length is fuel.
797def lex.go(
798 fuel: Nat,
799 sc: Scan,
800 from: String,
801 +at: U32,
802 +col: U32,
803 acc: List<&2, Token>
804) -> List<&2, Token>:
805 match fuel sc:
806 case 1n+k Tok{text, +len, +rest, last, +lex, sym, name, +tail}:
807 lex.go(
808 k,
809 lex.token(rest, rest, last),
810 rest,
811 (at + len: U32),
812 next_col(lex, tail, col, len),
813 Con{Token{text, at, (at + len: U32), col, lex, sym, name, tail, from}, acc}
814 )
815 case 1n+k Blank{+rest, last, +len}:
816 lex.go(k, lex.token(rest, rest, last), rest, (at + len: U32), (col + len: U32), acc)
817 case _ _:
818 List.reverse(&2, Token, acc)
819
820def tokens(+s: String) -> List<&2, Token>:
821 lex.go(1n+String.length(s), lex.token(s, s, '\0'), s, 0, 0, [])
822
823# Tokens
824# ------
825def Token.text(t: Token) -> String:
826 match t:
827 case Token{text, beg, end, col, lex, sym, name, tail, from}:
828 text
829
830def Token.beg(t: Token) -> U32:
831 match t:
832 case Token{text, beg, end, col, lex, sym, name, tail, from}:
833 beg
834
835def Token.end(t: Token) -> U32:
836 match t:
837 case Token{text, beg, end, col, lex, sym, name, tail, from}:
838 end
839
840def Token.col(t: Token) -> U32:
841 match t:
842 case Token{text, beg, end, col, lex, sym, name, tail, from}:
843 col
844
845def Token.sym(t: Token) -> Sym:
846 match t:
847 case Token{text, beg, end, col, lex, sym, name, tail, from}:
848 sym
849
850def Token.is_name(t: Token) -> Bool:
851 match t:
852 case Token{text, beg, end, col, lex, sym, name, tail, from}:
853 name
854
855def Token.from(t: Token) -> String:
856 match t:
857 case Token{text, beg, end, col, lex, sym, name, tail, from}:
858 from
859
860def is_comment(t: Token) -> Bool:
861 match t:
862 case Token{text, beg, end, col, Comment{}, sym, name, tail, from}:
863 True{}
864 case _:
865 False{}
866
867def is_operator(s: Sym) -> Bool:
868 match s:
869 case Plus{}:
870 True{}
871 case Minus{}:
872 True{}
873 case Star{}:
874 True{}
875 case Slash{}:
876 True{}
877 case Pct{}:
878 True{}
879 case Caret{}:
880 True{}
881 case Concat{}:
882 True{}
883 case AndAnd{}:
884 True{}
885 case OrOr{}:
886 True{}
887 case Meet{}:
888 True{}
889 case BitAnd{}:
890 True{}
891 case BitOr{}:
892 True{}
893 case BitXor{}:
894 True{}
895 case Shl{}:
896 True{}
897 case Shr{}:
898 True{}
899 case Cons{}:
900 True{}
901 case Lt{}:
902 True{}
903 case Gt{}:
904 True{}
905 case Le{}:
906 True{}
907 case Ge{}:
908 True{}
909 case EqEq{}:
910 True{}
911 case Ne{}:
912 True{}
913 case Amp{}:
914 True{}
915 case Bar{}:
916 True{}
917 case _:
918 False{}
919
920# A sign or quantity glued to what follows it.
921def is_prefix(s: Sym) -> Bool:
922 match s:
923 case Plus{}:
924 True{}
925 case Minus{}:
926 True{}
927 case Pct{}:
928 True{}
929 case Amp{}:
930 True{}
931 case _:
932 False{}
933
934def is_tight(s: Sym) -> Bool:
935 match s:
936 case Comma{}:
937 True{}
938 case Semi{}:
939 True{}
940 case Colon{}:
941 True{}
942 case Ques{}:
943 True{}
944 case Bang{}:
945 True{}
946 case Dot{}:
947 True{}
948 case _:
949 False{}
950
951def is_punct(s: Sym) -> Bool:
952 match s:
953 case Comma{}:
954 True{}
955 case Semi{}:
956 True{}
957 case Colon{}:
958 True{}
959 case _:
960 False{}
961
962def is_sigil(s: Sym) -> Bool:
963 match s:
964 case Tilde{}:
965 True{}
966 case At{}:
967 True{}
968 case Ques{}:
969 True{}
970 case Bang{}:
971 True{}
972 case Bslash{}:
973 True{}
974 case _:
975 False{}
976
977def is_opener(s: Sym) -> Bool:
978 match s:
979 case Lpar{}:
980 True{}
981 case Lbrk{}:
982 True{}
983 case Lbrc{}:
984 True{}
985 case Lt{}:
986 True{}
987 case _:
988 False{}
989
990def is_closer(s: Sym) -> Bool:
991 match s:
992 case Rpar{}:
993 True{}
994 case Rbrk{}:
995 True{}
996 case Rbrc{}:
997 True{}
998 case _:
999 False{}
1000
1001def is_keyword(s: Sym) -> Bool:
1002 match s:
1003 case KwReturn{}:
1004 True{}
1005 case KwCase{}:
1006 True{}
1007 case KwMatch{}:
1008 True{}
1009 case KwFor{}:
1010 True{}
1011 case KwExs{}:
1012 True{}
1013 case KwIs{}:
1014 True{}
1015 case _:
1016 False{}
1017
1018def is_block_word(s: Sym) -> Bool:
1019 match s:
1020 case KwMatch{}:
1021 True{}
1022 case KwCase{}:
1023 True{}
1024 case KwDo{}:
1025 True{}
1026 case Assign{}:
1027 True{}
1028 case Semi{}:
1029 True{}
1030 case Pct{}:
1031 True{}
1032 case Bslash{}:
1033 True{}
1034 case _:
1035 False{}
1036
1037# A record that ends in one of these goes on at the next line.
1038def is_continued(s: Sym) -> Bool:
1039 match s:
1040 case Arrow{}:
1041 True{}
1042 case Assign{}:
1043 True{}
1044 case Concat{}:
1045 True{}
1046 case AndAnd{}:
1047 True{}
1048 case OrOr{}:
1049 True{}
1050 case _:
1051 False{}
1052
1053def is_declaration(s: Sym) -> Bool:
1054 match s:
1055 case KwDef{}:
1056 True{}
1057 case KwType{}:
1058 True{}
1059 case KwLaw{}:
1060 True{}
1061 case At{}:
1062 True{}
1063 case _:
1064 False{}
1065
1066def is_import(s: Sym) -> Bool:
1067 match s:
1068 case KwImport{}:
1069 True{}
1070 case _:
1071 False{}
1072
1073def is_sym(s: Sym, t: Sym) -> Bool:
1074 match s t:
1075 case Comma{} Comma{}:
1076 True{}
1077 case Colon{} Colon{}:
1078 True{}
1079 case Lbrc{} Lbrc{}:
1080 True{}
1081 case Gt{} Gt{}:
1082 True{}
1083 case Rpar{} Rpar{}:
1084 True{}
1085 case Rbrk{} Rbrk{}:
1086 True{}
1087 case Rbrc{} Rbrc{}:
1088 True{}
1089 case Bang{} Bang{}:
1090 True{}
1091 case At{} At{}:
1092 True{}
1093 case _ _:
1094 False{}
1095
1096# Trees
1097# -----
1098def first(n: Tree) -> Token:
1099 match n:
1100 case Leaf{t}:
1101 t
1102 case Group{open, close, children, block}:
1103 open
1104
1105def last(n: Tree) -> Token:
1106 match n:
1107 case Leaf{t}:
1108 t
1109 case Group{open, close, children, block}:
1110 close
1111
1112def is_newline(n: Tree) -> Bool:
1113 match n:
1114 case Leaf{Token{text, beg, end, col, Newline{}, sym, name, tail, from}}:
1115 True{}
1116 case _:
1117 False{}
1118
1119def is_leaf(n: Tree) -> Bool:
1120 match n:
1121 case Leaf{t}:
1122 True{}
1123 case _:
1124 False{}
1125
1126def is_leaf_of(n: Tree, s: Sym) -> Bool:
1127 match n:
1128 case Leaf{t}:
1129 is_sym(Token.sym(t), s)
1130 case _:
1131 False{}
1132
1133def head(xs: List<&2, Tree>) -> Maybe<&2, Tree>:
1134 match xs:
1135 case Nil{}:
1136 None{}
1137 case Con{x, rest}:
1138 Some{x}
1139
1140def last_of(xs: List<&2, Tree>, at: Maybe<&2, Tree>) -> Maybe<&2, Tree>:
1141 match xs:
1142 case Nil{}:
1143 at
1144 case Con{x, rest}:
1145 last_of(rest, Some{x})
1146
1147# The first tree after a run of newlines.
1148def upcoming(xs: List<&2, Tree>) -> Maybe<&2, Tree>:
1149 match xs:
1150 case Nil{}:
1151 None{}
1152 case Con{Leaf{Token{text, beg, end, col, Newline{}, sym, name, tail, from}}, rest}:
1153 upcoming(rest)
1154 case Con{x, rest}:
1155 Some{x}
1156
1157# What closes the group s opens, or Word{}.
1158def closer(s: Sym) -> Sym:
1159 match s:
1160 case Lpar{}:
1161 Rpar{}
1162 case Lbrk{}:
1163 Rbrk{}
1164 case Lbrc{}:
1165 Rbrc{}
1166 case Lt{}:
1167 Gt{}
1168 case _:
1169 Word{}
1170
1171def is_angled(prev: Maybe<&2, Token>, t: Token) -> Bool:
1172 match prev:
1173 case Some{+p}:
1174 U32.is_eq(Token.end(p), Token.beg(t)) && Token.is_name(p)
1175 case None{}:
1176 False{}
1177
1178# Whether t opens a group: a glued < opens type arguments.
1179def opens.of(s: Sym, t: Token, prev: Maybe<&2, Token>) -> Bool:
1180 match s:
1181 case Lpar{}:
1182 True{}
1183 case Lbrk{}:
1184 True{}
1185 case Lbrc{}:
1186 True{}
1187 case Lt{}:
1188 is_angled(prev, t)
1189 case _:
1190 False{}
1191
1192def opens(+t: Token, prev: Maybe<&2, Token>) -> Bool:
1193 opens.of(Token.sym(t), t, prev)
1194
1195def is_block_leaf(n: Tree) -> Bool:
1196 match n:
1197 case Leaf{t}:
1198 is_block_word(Token.sym(t))
1199 case _:
1200 False{}
1201
1202def add(+n: Tree, stack: List<&2, Open>, root: List<&2, Tree>) -> Built:
1203 match stack:
1204 case Nil{}:
1205 Built{Nil{}, Con{n, root}}
1206 case Con{Open{t, close, children, block}, rest}:
1207 Built{Con{Open{t, close, Con{n, children}, block || is_block_leaf(n)}, rest}, root}
1208
1209def waits(stack: List<&2, Open>, s: Sym) -> Bool:
1210 match stack:
1211 case Nil{}:
1212 False{}
1213 case Con{Open{t, close, children, block}, rest}:
1214 is_sym(close, s)
1215
1216def pop(t: Token, stack: List<&2, Open>, root: List<&2, Tree>) -> Built:
1217 match stack:
1218 case Nil{}:
1219 Built{Nil{}, root}
1220 case Con{Open{open, close, children, block}, rest}:
1221 add(Group{open, t, List.reverse(&2, Tree, children), block}, rest, root)
1222
1223def tree.close(closes: Bool, t: Token, stack: List<&2, Open>, root: List<&2, Tree>) -> Built:
1224 match closes:
1225 case True{}:
1226 pop(t, stack, root)
1227 case False{}:
1228 add(Leaf{t}, stack, root)
1229
1230def tree.step(opens: Bool, +t: Token, +stack: List<&2, Open>, root: List<&2, Tree>) -> Built:
1231 match opens:
1232 case True{}:
1233 Built{Con{Open{t, closer(Token.sym(t)), [], False{}}, stack}, root}
1234 case False{}:
1235 tree.close(waits(stack, Token.sym(t)), t, stack, root)
1236
1237def tree.go(ts: List<&2, Token>, prev: Maybe<&2, Token>, b: Built) -> Maybe<&2, List<&2, Tree>>:
1238 match ts b:
1239 case Nil{} Built{Nil{}, root}:
1240 Some{List.reverse(&2, Tree, root)}
1241 case Nil{} Built{Con{o, os}, root}:
1242 None{}
1243 case Con{+t, rest} Built{stack, root}:
1244 tree.go(rest, Some{t}, tree.step(opens(t, prev), t, stack, root))
1245
1246# Docs
1247# ----
1248def measure(s: String, +n: U32, tail: Maybe<&2, U32>, keep: U32) -> Measure:
1249 match s:
1250 case SNil{}:
1251 Measure{n, tail, keep}
1252 case SCon{+c, rest}:
1253 measure(
1254 rest,
1255 (n + 1: U32),
1256 tail_after(c, tail),
1257 Bool.pick(U32, is_blank(c), keep, (n + 1: U32))
1258 )
1259
1260def str.split(whole: Bool, +s: String, width: U32, tail: Maybe<&2, U32>, +keep: U32) -> Doc:
1261 match whole:
1262 case True{}:
1263 Str{s, "", width, tail}
1264 case False{}:
1265 Str{Strings.take(s, U32.to_nat(keep)), String.drop(s, U32.to_nat(keep)), width, tail}
1266
1267def str.of(s: String, m: Measure) -> Doc:
1268 match m:
1269 case Measure{+width, tail, +keep}:
1270 str.split(U32.is_eq(width, keep), s, width, tail, keep)
1271
1272def str(+s: String) -> Doc:
1273 str.of(s, measure(s, 0, None{}, 0))
1274
1275# A doc for text of this width, with no newline and no trailing whitespace.
1276def text(s: String, width: U32) -> Doc:
1277 Str{s, "", width, None{}}
1278
1279def space(s: String, width: U32) -> Doc:
1280 Str{"", s, width, None{}}
1281
1282def nothing() -> Doc:
1283 Str{"", "", 0, None{}}
1284
1285def soft(flat: String, width: U32) -> Doc:
1286 Line{flat, width, False{}, 0}
1287
1288def hard(offset: U32) -> Doc:
1289 Line{"", 0, True{}, offset}
1290
1291def flat_of(d: Doc) -> Flat:
1292 match d:
1293 case Str{body, blank, width, Some{t}}:
1294 Flat{0, True{}}
1295 case Str{body, blank, width, None{}}:
1296 Flat{width, False{}}
1297 case Line{text, width, True{}, offset}:
1298 Flat{0, True{}}
1299 case Line{text, width, False{}, offset}:
1300 Flat{width, False{}}
1301 case Grp{doc, f}:
1302 f
1303 case Nest{doc, amount, f}:
1304 f
1305 case Cat{docs, f}:
1306 f
1307
1308# The Flat of a doc a, then of what follows it.
1309def flat_then(a: Flat, rest: Flat) -> Flat:
1310 match a rest:
1311 case Flat{w, True{}} r:
1312 Flat{w, True{}}
1313 case Flat{w, False{}} Flat{v, b}:
1314 Flat{(w + v: U32), b}
1315
1316def flat_rev(docs: List<&2, Doc>, acc: Flat) -> Flat:
1317 match docs:
1318 case Nil{}:
1319 acc
1320 case Con{d, rest}:
1321 flat_rev(rest, flat_then(flat_of(d), acc))
1322
1323def grp(+d: Doc) -> Doc:
1324 Grp{d, flat_of(d)}
1325
1326def nest(+d: Doc, amount: U32) -> Doc:
1327 Nest{d, amount, flat_of(d)}
1328
1329# A Cat of docs last first.
1330def cat_rev(+docs: List<&2, Doc>) -> Doc:
1331 Cat{docs, flat_rev(docs, Flat{0, False{}})}
1332
1333# A Cat of a few docs, in order.
1334def cat(docs: List<&2, Doc>) -> Doc:
1335 cat_rev(List.reverse(&2, Doc, docs))
1336
1337def code(t: Token) -> Doc:
1338 match t:
1339 case Token{text, beg, end, col, lex, sym, name, tail, from}:
1340 Str{text, "", (end - beg: U32), None{}}
1341
1342def is_glued_prefix(+a: Token, b: Maybe<&2, Tree>) -> Bool:
1343 match b:
1344 case Some{n}:
1345 is_prefix(Token.sym(a)) && U32.is_eq(Token.end(a), Token.beg(first(n)))
1346 case None{}:
1347 False{}
1348
1349# Whether a group after x is its argument list or type arguments.
1350def is_callee(+x: Token) -> Bool:
1351 (
1352 Token.is_name(x) || is_sym(Token.sym(x), Bang{}) || is_closer(Token.sym(x))
1353 )
1354 && Bool.not(is_keyword(Token.sym(x)))
1355
1356def is_inc(s: Sym) -> Bool:
1357 match s:
1358 case Inc{}:
1359 True{}
1360 case _:
1361 False{}
1362
1363# What goes between x and the tree b that y starts. A constructor's brace
1364# is glued to its name; a brace after a space can instead be the next
1365# (annotated) argument of a call without commas, so the space stays. Other
1366# atoms need a space.
1367def separator.of(+x: Token, +y: Token, +group: Bool, b: Tree) -> Doc:
1368 Bool.pick(
1369 Doc,
1370 is_comment(y),
1371 space(" ", 2),
1372 Bool.pick(
1373 Doc,
1374 is_tight(Token.sym(y)),
1375 nothing(),
1376 Bool.pick(
1377 Doc,
1378 is_punct(Token.sym(x)),
1379 space(" ", 1),
1380 Bool.pick(
1381 Doc,
1382 is_sigil(Token.sym(x)),
1383 nothing(),
1384 Bool.pick(
1385 Doc,
1386 is_glued_prefix(x, Some{b}),
1387 nothing(),
1388 Bool.pick(
1389 Doc,
1390 is_inc(Token.sym(x)),
1391 nothing(),
1392 Bool.pick(
1393 Doc,
1394 group && is_sym(Token.sym(y), Lbrc{}) && U32.is_ne(Token.end(x), Token.beg(y)),
1395 space(" ", 1),
1396 Bool.pick(
1397 Doc,
1398 group && is_opener(Token.sym(y)) && is_callee(x),
1399 nothing(),
1400 space(" ", 1)
1401 )
1402 )
1403 )
1404 )
1405 )
1406 )
1407 )
1408 )
1409
1410def separator(a: Tree, +b: Tree) -> Doc:
1411 separator.of(last(a), first(b), Bool.not(is_leaf(b)), b)
1412
1413# "#x" gets a space; empty comments, whitespace, test expectations (#|),
1414# directives (#!) and headings (##) stay as they are.
1415def comment_text.after(rest: String, +keep: String) -> String:
1416 match rest:
1417 case SNil{}:
1418 "#"
1419 case SCon{+c, more}:
1420 Bool.pick(
1421 String,
1422 is_blank(c) || Char.is_eq(c, '#') || Char.is_eq(c, '!') || Char.is_eq(c, '|'),
1423 "#" ++ keep,
1424 "# " ++ keep
1425 )
1426
1427def comment_text(text: String, keep: String) -> String:
1428 match text:
1429 case SCon{'#', +rest}:
1430 comment_text.after(rest, rest)
1431 case _:
1432 keep
1433
1434def leaf_doc(+t: Token) -> Doc:
1435 match t:
1436 case Token{text, beg, end, col, Newline{}, sym, name, tail, from}:
1437 nothing()
1438 case Token{+text, beg, end, col, Comment{}, sym, name, tail, from}:
1439 str(comment_text(text, text))
1440 case Token{text, beg, end, col, Literal{}, sym, name, tail, from}:
1441 str(text)
1442 case Token{text, beg, end, col, Code{}, sym, name, tail, from}:
1443 code(t)
1444
1445# Levels
1446# ------
1447def should_pop(levels: List<&2, U32>, col: U32) -> Bool:
1448 match levels:
1449 case Con{top, Con{x, xs}}:
1450 U32.is_lt(col, top)
1451 case _:
1452 False{}
1453
1454def pop_levels(levels: List<&2, U32>, +col: U32, go: Bool) -> List<&2, U32>:
1455 match levels go:
1456 case Con{top, +more} True{}:
1457 pop_levels(more, col, should_pop(more, col))
1458 case Con{top, more} False{}:
1459 Con{top, more}
1460 case Nil{} _:
1461 Nil{}
1462
1463def deepen(levels: List<&2, U32>, +col: U32) -> List<&2, U32>:
1464 match levels:
1465 case Nil{}:
1466 [col]
1467 case Con{+top, +more}:
1468 Bool.pick(List<&2, U32>, U32.is_gt(col, top), Con{col, Con{top, more}}, Con{top, more})
1469
1470# Closes the levels deeper than col, then opens col if it is deeper.
1471def indented(+levels: List<&2, U32>, +col: U32) -> List<&2, U32>:
1472 deepen(pop_levels(levels, col, should_pop(levels, col)), col)
1473
1474def depth.go(levels: List<&2, U32>, n: U32) -> U32:
1475 match levels:
1476 case Nil{}:
1477 n
1478 case Con{top, more}:
1479 depth.go(more, (n + 1: U32))
1480
1481def depth(levels: List<&2, U32>) -> U32:
1482 (depth.go(levels, 0) - 1: U32)
1483
1484# Sequences
1485# ---------
1486def fresh(base: U32) -> Seq:
1487 Seq{[], None{}, None{}, [base], False{}}
1488
1489def Seq.put(s: Seq, d: Doc) -> Seq:
1490 match s:
1491 case Seq{out, None{}, prev, levels, skip}:
1492 Seq{Con{d, out}, None{}, prev, levels, skip}
1493 case Seq{out, Some{chain}, prev, levels, skip}:
1494 Seq{out, Some{Con{d, chain}}, prev, levels, skip}
1495
1496def Seq.chain(s: Seq) -> Seq:
1497 match s:
1498 case Seq{out, None{}, prev, levels, skip}:
1499 Seq{out, Some{[soft(" ", 1)]}, prev, levels, skip}
1500 case Seq{out, Some{chain}, prev, levels, skip}:
1501 Seq{out, Some{Con{soft(" ", 1), chain}}, prev, levels, skip}
1502
1503def Seq.doc(s: Seq, tab: U32) -> Doc:
1504 match s:
1505 case Seq{out, None{}, prev, levels, skip}:
1506 grp(cat_rev(out))
1507 case Seq{out, Some{chain}, prev, levels, skip}:
1508 grp(cat_rev(Con{nest(cat_rev(chain), tab), out}))
1509
1510def is_prev_comment(prev: Maybe<&2, Tree>) -> Bool:
1511 match prev:
1512 case Some{p}:
1513 is_comment(last(p))
1514 case None{}:
1515 False{}
1516
1517# A run of newlines breaks the line at the next tree's level.
1518def Seq.level(
1519 +levels: List<&2, U32>,
1520 tab: U32,
1521 out: List<&2, Doc>,
1522 chain: Maybe<&2, List<&2, Doc>>
1523) -> Seq:
1524 Seq.put(Seq{out, chain, None{}, levels, True{}}, hard((depth(levels) * tab: U32)))
1525
1526def Seq.end(comment: Bool, s: Seq) -> Seq:
1527 match comment:
1528 case True{}:
1529 Seq.put(s, hard(0))
1530 case False{}:
1531 s
1532
1533def Seq.break(up: Maybe<&2, Tree>, tab: U32, s: Seq) -> Seq:
1534 match up s:
1535 case Some{u} Seq{out, chain, prev, levels, skip}:
1536 Seq.level(indented(levels, Token.col(first(u))), tab, out, chain)
1537 case None{} Seq{out, chain, prev, levels, skip}:
1538 Seq.end(is_prev_comment(prev), Seq{out, chain, None{}, levels, True{}})
1539
1540def Seq.newline.if(trigger: Bool, rest: List<&2, Tree>, tab: U32, s: Seq) -> Seq:
1541 match trigger:
1542 case True{}:
1543 Seq.break(upcoming(rest), tab, s)
1544 case False{}:
1545 s
1546
1547def Seq.newline(rest: List<&2, Tree>, tab: U32, raw: Bool, s: Seq) -> Seq:
1548 match s:
1549 case Seq{out, chain, prev, levels, True{}}:
1550 Seq{out, chain, prev, levels, True{}}
1551 case Seq{out, chain, +prev, levels, False{}}:
1552 Seq.newline.if(
1553 raw || is_prev_comment(prev),
1554 rest,
1555 tab,
1556 Seq{out, chain, prev, levels, False{}}
1557 )
1558
1559def Seq.joint(chain: Bool, p: Tree, n: Tree, s: Seq) -> Seq:
1560 match chain:
1561 case True{}:
1562 Seq.chain(s)
1563 case False{}:
1564 Seq.put(s, separator(p, n))
1565
1566def Seq.tree(+n: Tree, d: Doc, next: Maybe<&2, Tree>, raw: Bool, s: Seq) -> Seq:
1567 match s:
1568 case Seq{out, chain, None{}, levels, skip}:
1569 Seq.put(Seq{out, chain, Some{n}, levels, False{}}, d)
1570 case Seq{out, chain, Some{p}, levels, skip}:
1571 Seq.put(
1572 Seq.joint(
1573 Bool.not(
1574 raw
1575 )
1576 && is_leaf(n)
1577 && is_operator(Token.sym(first(n)))
1578 && Bool.not(is_glued_prefix(first(n), next)),
1579 p,
1580 n,
1581 Seq{out, chain, Some{n}, levels, False{}}
1582 ),
1583 d
1584 )
1585
1586def Seq.step(
1587 newline: Bool,
1588 n: Tree,
1589 d: Doc,
1590 rest: List<&2, Tree>,
1591 tab: U32,
1592 raw: Bool,
1593 s: Seq
1594) -> Seq:
1595 match newline:
1596 case True{}:
1597 Seq.newline(rest, tab, raw, s)
1598 case False{}:
1599 Seq.tree(n, d, head(rest), raw, s)
1600
1601def Seq.visit(+n: Tree, d: Doc, rest: List<&2, Tree>, tab: U32, raw: Bool, s: Seq) -> Seq:
1602 Seq.step(is_newline(n), n, d, rest, tab, raw, s)
1603
1604# Parts
1605# -----
1606def Split.fresh() -> Split:
1607 Split{[], True{}, False{}, False{}, False{}}
1608
1609def Split.track(+n: Tree, attach: Bool, sp: Split) -> Split:
1610 match sp:
1611 case Split{parts, newlines, comment, a, comments}:
1612 Split{
1613 parts,
1614 newlines && is_newline(n),
1615 Bool.pick(Bool, is_newline(n), comment, is_comment(last(n))),
1616 attach,
1617 comments
1618 }
1619
1620def Split.end(tab: U32, base: U32, st: State) -> State:
1621 match st:
1622 case State{seq, Split{parts, newlines, comment, attach, comments}}:
1623 State{
1624 fresh(base),
1625 Split{
1626 Con{Part{Seq.doc(seq, tab), newlines, comment}, parts},
1627 True{},
1628 False{},
1629 False{},
1630 comments
1631 }
1632 }
1633
1634def Split.comments(n: Tree, sp: Split) -> Split:
1635 match sp:
1636 case Split{parts, newlines, comment, attach, comments}:
1637 Split{parts, newlines, comment, attach, comments || is_comment(first(n))}
1638
1639def has_head_comment(xs: List<&2, Tree>) -> Bool:
1640 match xs:
1641 case Con{x, rest}:
1642 is_comment(first(x))
1643 case Nil{}:
1644 False{}
1645
1646def Split.act(n: Tree, rest: List<&2, Tree>, sp: Split) -> Act:
1647 match sp:
1648 case Split{parts, newlines, comment, True{}, comments}:
1649 AddEnd{}
1650 case Split{parts, newlines, comment, False{}, comments}:
1651 Bool.pick(
1652 Act,
1653 is_leaf_of(n, Comma{}),
1654 Bool.pick(Act, has_head_comment(rest), AddAttach{}, End{}),
1655 Add{}
1656 )
1657
1658def Split.apply(
1659 act: Act,
1660 +n: Tree,
1661 d: Doc,
1662 rest: List<&2, Tree>,
1663 +tab: U32,
1664 base: U32,
1665 st: State
1666) -> State:
1667 match act st:
1668 case Add{} State{seq, sp}:
1669 State{Seq.visit(n, d, rest, tab, False{}, seq), Split.track(n, False{}, sp)}
1670 case AddAttach{} State{seq, sp}:
1671 State{Seq.visit(n, d, rest, tab, False{}, seq), Split.track(n, True{}, sp)}
1672 case AddEnd{} State{seq, sp}:
1673 Split.end(
1674 tab,
1675 base,
1676 State{Seq.visit(n, d, rest, tab, False{}, seq), Split.track(n, False{}, sp)}
1677 )
1678 case End{} State{seq, sp}:
1679 Split.end(tab, base, State{seq, sp})
1680
1681def Split.visit(+n: Tree, d: Doc, +rest: List<&2, Tree>, tab: U32, base: U32, st: State) -> State:
1682 match st:
1683 case State{seq, +sp}:
1684 Split.apply(Split.act(n, rest, sp), n, d, rest, tab, base, State{seq, Split.comments(n, sp)})
1685
1686def should_drop(parts: List<&2, Part>) -> Bool:
1687 match parts:
1688 case Con{Part{d, True{}, c}, Con{p, ps}}:
1689 True{}
1690 case _:
1691 False{}
1692
1693# Drops the trailing parts that hold only newlines, keeping one part.
1694def drop_parts(parts: List<&2, Part>, go: Bool) -> List<&2, Part>:
1695 match parts go:
1696 case Con{p, +rest} True{}:
1697 drop_parts(rest, should_drop(rest))
1698 case ps False{}:
1699 ps
1700 case Nil{} _:
1701 Nil{}
1702
1703def content.sep(comment: Bool, acc: List<&2, Doc>) -> List<&2, Doc>:
1704 match comment:
1705 case True{}:
1706 Con{hard(0), acc}
1707 case False{}:
1708 Con{soft(" ", 1), Con{text(",", 1), acc}}
1709
1710def content.go(parts: List<&2, Part>, comment: Bool, acc: List<&2, Doc>) -> List<&2, Doc>:
1711 match parts:
1712 case Nil{}:
1713 acc
1714 case Con{Part{d, newlines, next}, rest}:
1715 content.go(rest, next, Con{d, content.sep(comment, acc)})
1716
1717# The parts' docs with their separators, last first.
1718def content(parts: List<&2, Part>) -> List<&2, Doc>:
1719 match parts:
1720 case Nil{}:
1721 []
1722 case Con{Part{d, newlines, comment}, rest}:
1723 content.go(rest, comment, [d])
1724
1725def opening(comments: Bool) -> Doc:
1726 match comments:
1727 case True{}:
1728 hard(0)
1729 case False{}:
1730 soft("", 0)
1731
1732# A generic's closing > stays glued to the term before it: Bend parses a
1733# spaced > as a comparison, even across a newline.
1734def closing(close: Token, comments: Bool) -> Doc:
1735 Bool.pick(Doc, is_sym(Token.sym(close), Gt{}), nothing(), opening(comments))
1736
1737def Split.doc(open: Token, +close: Token, tab: U32, st: State) -> Doc:
1738 match st:
1739 case State{seq, Split{+parts, newlines, comment, attach, +comments}}:
1740 grp(
1741 cat(
1742 [
1743 code(open),
1744 nest(
1745 cat(
1746 [
1747 opening(comments),
1748 cat_rev(content(List.reverse(&2, Part, drop_parts(parts, should_drop(parts)))))
1749 ]
1750 ),
1751 tab
1752 ),
1753 closing(close, comments),
1754 code(close)
1755 ]
1756 )
1757 )
1758
1759# Build
1760# -----
1761def mode_of(open: Token, close: Token, children: List<&2, Tree>, block: Bool) -> Mode:
1762 match children block:
1763 case Nil{} _:
1764 Bare{open, close}
1765 case _ True{}:
1766 Block{open, close}
1767 case _ False{}:
1768 Parts{open, close}
1769
1770def start(base: U32) -> State:
1771 State{fresh(base), Split.fresh()}
1772
1773def visit(
1774 +mode: Mode,
1775 n: Tree,
1776 d: Doc,
1777 rest: List<&2, Tree>,
1778 tab: U32,
1779 base: U32,
1780 st: State
1781) -> State:
1782 match mode st:
1783 case Plain{} State{seq, sp}:
1784 State{Seq.visit(n, d, rest, tab, False{}, seq), sp}
1785 case Block{open, close} State{seq, sp}:
1786 State{Seq.visit(n, d, rest, tab, True{}, seq), sp}
1787 case Parts{open, close} st2:
1788 Split.visit(n, d, rest, tab, base, st2)
1789 case Bare{open, close} st2:
1790 st2
1791
1792def finish(mode: Mode, st: State, +tab: U32, base: U32) -> Doc:
1793 match mode st:
1794 case Plain{} State{seq, sp}:
1795 Seq.doc(seq, tab)
1796 case Block{open, close} State{seq, sp}:
1797 cat([code(open), Seq.doc(seq, tab), code(close)])
1798 case Parts{open, close} st2:
1799 Split.doc(open, close, tab, Split.end(tab, base, st2))
1800 case Bare{open, close} st2:
1801 cat([code(open), code(close)])
1802
1803# The doc of a list of trees, read as mode says. A group's children are
1804# built first, by the same def, as a smaller part of the list.
1805def build(nodes: List<&2, Tree>, +tab: U32, +base: U32, +mode: Mode, st: State) -> Doc:
1806 match nodes:
1807 case Nil{}:
1808 finish(mode, st, tab, base)
1809 case Con{Leaf{+t}, +rest}:
1810 build(rest, tab, base, mode, visit(mode, Leaf{t}, leaf_doc(t), rest, tab, base, st))
1811 case Con{Group{+open, +close, +children, +block}, +rest}:
1812 build(
1813 rest,
1814 tab,
1815 base,
1816 mode,
1817 visit(
1818 mode,
1819 Group{open, close, children, block},
1820 build(children, tab, base, mode_of(open, close, children, block), start(base)),
1821 rest,
1822 tab,
1823 base,
1824 st
1825 )
1826 )
1827
1828def sequence(nodes: List<&2, Tree>, tab: U32, +base: U32) -> Doc:
1829 build(nodes, tab, base, Plain{}, start(base))
1830
1831# Printer
1832# -------
1833# Jobs for a Cat's docs, last first, so its first doc ends on top.
1834def push(docs: List<&2, Doc>, +indent: U32, +flat: Bool, jobs: List<&2, Job>) -> List<&2, Job>:
1835 match docs:
1836 case Nil{}:
1837 jobs
1838 case Con{d, rest}:
1839 push(rest, indent, flat, Con{Job{d, indent, flat}, jobs})
1840
1841def fits.flat.of(over: Bool, brk: Bool, jobs: List<&2, Job>, left: U32) -> Fit:
1842 match over brk:
1843 case True{} _:
1844 Stop{False{}}
1845 case False{} True{}:
1846 Stop{True{}}
1847 case False{} False{}:
1848 Go{jobs, left}
1849
1850# A doc printed flat, whole: past the width, or up to a break within it.
1851def fits.flat(f: Flat, jobs: List<&2, Job>, +left: U32) -> Fit:
1852 match f:
1853 case Flat{+w, brk}:
1854 fits.flat.of(U32.is_gt(w, left), brk, jobs, (left - w: U32))
1855
1856def fits.width(stop: Bool, +n: U32, jobs: List<&2, Job>, +left: U32) -> Fit:
1857 match stop:
1858 case True{}:
1859 Stop{True{}}
1860 case False{}:
1861 Bool.pick(Fit, U32.is_gt(n, left), Stop{False{}}, Go{jobs, (left - n: U32)})
1862
1863def fits.step(jobs: List<&2, Job>, left: U32) -> Fit:
1864 match jobs:
1865 case Nil{}:
1866 Stop{True{}}
1867 case Con{Job{Str{body, blank, width, tail}, indent, flat}, rest}:
1868 fits.width(Maybe.is_some(&2, U32, tail), width, rest, left)
1869 case Con{Job{Line{text, width, hard, offset}, indent, flat}, rest}:
1870 fits.width(hard || Bool.not(flat), width, rest, left)
1871 case Con{Job{Grp{doc, f}, indent, flat}, rest}:
1872 fits.flat(f, rest, left)
1873 case Con{Job{Nest{doc, amount, f}, indent, flat}, rest}:
1874 fits.flat(f, rest, left)
1875 case Con{Job{Cat{docs, f}, indent, True{}}, rest}:
1876 fits.flat(f, rest, left)
1877 case Con{Job{Cat{docs, f}, indent, False{}}, rest}:
1878 Go{push(docs, indent, False{}, rest), left}
1879
1880def fits.go(fuel: Nat, f: Fit) -> Bool:
1881 match fuel f:
1882 case _ Stop{b}:
1883 b
1884 case 0n Go{jobs, left}:
1885 True{}
1886 case 1n+k Go{jobs, left}:
1887 fits.go(k, fits.step(jobs, left))
1888
1889def fits.from(over: Bool, fuel: Nat, left: U32, jobs: List<&2, Job>) -> Bool:
1890 match over:
1891 case True{}:
1892 False{}
1893 case False{}:
1894 fits.go(fuel, Go{jobs, left})
1895
1896# Whether the jobs fit before their first break, from col.
1897def fits(fuel: Nat, +width: U32, +col: U32, jobs: List<&2, Job>) -> Bool:
1898 fits.from(U32.is_gt(col, width), fuel, (width - col: U32), jobs)
1899
1900def col_of(tail: Maybe<&2, U32>, col: U32, width: U32) -> U32:
1901 match tail:
1902 case Some{t}:
1903 t
1904 case None{}:
1905 (col + width: U32)
1906
1907# Text goes out after the pending whitespace; its own trailing whitespace
1908# becomes pending.
1909def print.str(
1910 body: String,
1911 keep: String,
1912 blank: String,
1913 jobs: List<&2, Job>,
1914 out: List<&2, String>,
1915 pending: String,
1916 col: U32
1917) -> Print:
1918 match body:
1919 case SNil{}:
1920 Print{jobs, out, pending ++ blank, col}
1921 case _:
1922 Print{jobs, Con{keep, Con{pending, out}}, blank, col}
1923
1924def print.line(
1925 flat: Bool,
1926 text: String,
1927 +at: U32,
1928 jobs: List<&2, Job>,
1929 out: List<&2, String>,
1930 pending: String,
1931 col: U32,
1932 width: U32,
1933 newline: String
1934) -> Print:
1935 match flat:
1936 case True{}:
1937 Print{jobs, out, pending ++ text, (col + width: U32)}
1938 case False{}:
1939 Print{jobs, out, newline ++ pad(U32.to_nat(at), ""), at}
1940
1941def print.is_flat(
1942 flat: Bool,
1943 fuel: Nat,
1944 +width: Maybe<&2, U32>,
1945 col: U32,
1946 jobs: List<&2, Job>
1947) -> Bool:
1948 match flat width:
1949 case True{} _:
1950 True{}
1951 case False{} None{}:
1952 True{}
1953 case False{} Some{w}:
1954 fits(fuel, w, col, jobs)
1955
1956def is_unbounded(width: Maybe<&2, U32>) -> Bool:
1957 match width:
1958 case None{}:
1959 True{}
1960 case Some{w}:
1961 False{}
1962
1963def print.step(
1964 jobs: List<&2, Job>,
1965 out: List<&2, String>,
1966 pending: String,
1967 +col: U32,
1968 width: Maybe<&2, U32>,
1969 newline: String,
1970 fuel: Nat
1971) -> Print:
1972 match jobs:
1973 case Nil{}:
1974 Print{Nil{}, out, pending, col}
1975 case Con{Job{Str{+body, blank, w, tail}, indent, flat}, rest}:
1976 print.str(body, body, blank, rest, out, pending, col_of(tail, col, w))
1977 case Con{Job{Line{text, w, hard, offset}, +indent, +flat}, rest}:
1978 print.line(
1979 Bool.not(hard) && (flat || is_unbounded(width)),
1980 text,
1981 (indent + offset: U32),
1982 rest,
1983 out,
1984 pending,
1985 col,
1986 w,
1987 newline
1988 )
1989 case Con{Job{Nest{doc, amount, f}, indent, flat}, rest}:
1990 Print{Con{Job{doc, (indent + amount: U32), flat}, rest}, out, pending, col}
1991 case Con{Job{Grp{+doc, f}, +indent, +flat}, +rest}:
1992 Print{
1993 Con{
1994 Job{
1995 doc,
1996 indent,
1997 print.is_flat(flat, fuel, width, col, Con{Job{doc, indent, True{}}, rest})
1998 },
1999 rest
2000 },
2001 out,
2002 pending,
2003 col
2004 }
2005 case Con{Job{Cat{docs, f}, indent, flat}, rest}:
2006 Print{push(docs, indent, flat, rest), out, pending, col}
2007
2008# The text so far; pending whitespace is never written at the end.
2009def print.go(fuel: Nat, p: Print, +width: Maybe<&2, U32>, +newline: String, +budget: Nat) -> String:
2010 match fuel p:
2011 case _ Print{Nil{}, out, pending, col}:
2012 Strings.onto(out, "")
2013 case 0n Print{jobs, out, pending, col}:
2014 Strings.onto(out, "")
2015 case 1n+k Print{jobs, out, pending, col}:
2016 print.go(
2017 k,
2018 print.step(jobs, out, pending, col, width, newline, budget),
2019 width,
2020 newline,
2021 budget
2022 )
2023
2024# The doc at indent; fuel bounds the steps of the printer and of each fit.
2025# A group is all flat if it fits; else its separators break together.
2026# Nesting is relative, never aligned under a function name. The width is a
2027# target, not a license to split atoms.
2028def print(doc: Doc, width: Maybe<&2, U32>, +indent: U32, newline: String, +fuel: Nat) -> String:
2029 print.go(
2030 fuel,
2031 Print{[Job{doc, indent, False{}}], [], pad(U32.to_nat(indent), ""), indent},
2032 width,
2033 newline,
2034 fuel
2035 )
2036
2037# Records
2038# -------
2039def joins.of(+p: Tree, +next: Tree, following: Maybe<&2, Tree>) -> Bool:
2040 Bool.not(
2041 is_comment(last(p))
2042 )
2043 && Bool.not(is_comment(first(next)))
2044 && (
2045 (
2046 Bool.not(
2047 is_glued_prefix(first(next), following)
2048 )
2049 && is_leaf(next)
2050 && is_operator(Token.sym(first(next)))
2051 )
2052 || is_continued(Token.sym(last(p)))
2053 )
2054
2055# Whether a newline between previous and rest continues the record.
2056def joins(previous: Maybe<&2, Tree>, rest: List<&2, Tree>) -> Bool:
2057 match previous rest:
2058 case Some{p} Con{next, more}:
2059 joins.of(p, next, head(more))
2060 case _ _:
2061 False{}
2062
2063def records.newline(join: Bool, done: List<&2, List<&2, Tree>>, current: List<&2, Tree>) -> Recs:
2064 match join:
2065 case True{}:
2066 Recs{done, current}
2067 case False{}:
2068 Recs{Con{List.reverse(&2, Tree, current), done}, Nil{}}
2069
2070def records.step(
2071 newline: Bool,
2072 n: Tree,
2073 rest: List<&2, Tree>,
2074 done: List<&2, List<&2, Tree>>,
2075 +current: List<&2, Tree>
2076) -> Recs:
2077 match newline:
2078 case False{}:
2079 Recs{done, Con{n, current}}
2080 case True{}:
2081 records.newline(joins(head(current), rest), done, current)
2082
2083def records(nodes: List<&2, Tree>, r: Recs) -> List<&2, List<&2, Tree>>:
2084 match nodes r:
2085 case Nil{} Recs{done, current}:
2086 List.reverse(&2, List<&2, Tree>, Con{List.reverse(&2, Tree, current), done})
2087 case Con{+n, +rest} Recs{done, current}:
2088 records(rest, records.step(is_newline(n), n, rest, done, current))
2089
2090def cut.step(+n: Tree, c: Cut) -> Cut:
2091 match c:
2092 case Cut{before, after, True{}}:
2093 Cut{before, Con{n, after}, True{}}
2094 case Cut{before, after, False{}}:
2095 Cut{Con{n, before}, after, is_leaf_of(n, Colon{})}
2096
2097def cut.go(xs: List<&2, Tree>, c: Cut) -> Cut:
2098 match xs c:
2099 case Nil{} Cut{before, after, found}:
2100 Cut{List.reverse(&2, Tree, before), List.reverse(&2, Tree, after), found}
2101 case Con{n, rest} c2:
2102 cut.go(rest, cut.step(n, c2))
2103
2104# Whether the cut has a body after its colon that is not a comment.
2105def cut.is_inline(c: Cut) -> Bool:
2106 match c:
2107 case Cut{before, Con{x, xs}, True{}}:
2108 Bool.not(is_comment(first(x)))
2109 case _:
2110 False{}
2111
2112def is_not_hash(s: String) -> Bool:
2113 match s:
2114 case SNil{}:
2115 False{}
2116 case SCon{'#', rest}:
2117 False{}
2118 case _:
2119 True{}
2120
2121def split_hash.of(r: Run) -> Hash:
2122 match r:
2123 case Run{before, n, SNil{}, last}:
2124 Hash{before, None{}}
2125 case Run{before, n, rest, last}:
2126 Hash{before, Some{rest}}
2127
2128def split_hash(+s: String) -> Hash:
2129 split_hash.of(scan(~is_not_hash, s, s, is_not_hash(s), "", 0, '\0'))
2130
2131def import_space.of(+rest: String, s: String) -> String:
2132 Bool.pick(String, head_passes(~is_blank, rest), "import " ++ drop_blank(rest), s)
2133
2134def import_space(+s: String) -> String:
2135 import_space.of(String.drop(s, 6n), s)
2136
2137def is_alias_at.as(s: String) -> Bool:
2138 match s:
2139 case SCon{'a', SCon{'s', rest}}:
2140 head_passes(~is_blank, rest)
2141 case _:
2142 False{}
2143
2144def is_alias_at(+s: String) -> Bool:
2145 head_passes(~is_blank, s) && is_alias_at.as(drop_blank(s))
2146
2147def alias.go(s: String, hit: Bool, acc: String) -> String:
2148 match s hit:
2149 case SNil{} _:
2150 String.reverse(acc)
2151 case SCon{c, rest} True{}:
2152 String.reverse(acc) ++ " as " ++ drop_blank(String.drop(drop_blank(SCon{c, rest}), 2n))
2153 case SCon{c, +rest} False{}:
2154 alias.go(rest, is_alias_at(rest), SCon{c, acc})
2155
2156def alias(+s: String) -> String:
2157 alias.go(s, is_alias_at(s), "")
2158
2159def import_doc.of(h: Hash) -> Doc:
2160 match h:
2161 case Hash{before, None{}}:
2162 str(alias(import_space(trim_end(before))))
2163 case Hash{before, Some{+comment}}:
2164 cat(
2165 [
2166 str(alias(import_space(trim_end(before)))),
2167 space(" ", 2),
2168 str(comment_text(comment, comment))
2169 ]
2170 )
2171
2172# Where the record that head starts ends.
2173def record_end(head: Token, record: List<&2, Tree>) -> U32:
2174 Token.end(last(Maybe.default(&2, Tree, last_of(record, None{}), Leaf{head})))
2175
2176# Paths and aliases have their own grammar: / and - are not operators there,
2177# so an import keeps its text and only its spacing changes.
2178def import_doc(+head: Token, record: List<&2, Tree>) -> Doc:
2179 import_doc.of(
2180 split_hash(
2181 trim_end(
2182 Strings.take(
2183 Token.from(head),
2184 U32.to_nat((record_end(head, record) - Token.beg(head): U32))
2185 )
2186 )
2187 )
2188 )
2189
2190def decl_doc.of(inline: Bool, c: Cut, record: List<&2, Tree>, +tab: U32, +col: U32) -> Doc:
2191 match inline c:
2192 case True{} Cut{before, after, found}:
2193 cat([sequence(before, tab, col), nest(cat([hard(0), sequence(after, tab, col)]), tab)])
2194 case False{} c2:
2195 sequence(record, tab, col)
2196
2197def decl_doc.cut(+c: Cut, record: List<&2, Tree>, declaration: Bool, tab: U32, +col: U32) -> Doc:
2198 decl_doc.of(U32.is_zero(col) && declaration && cut.is_inline(c), c, record, tab, col)
2199
2200# A declaration's body on the line of its colon moves to the next line,
2201# indented.
2202def decl_doc(+record: List<&2, Tree>, declaration: Bool, tab: U32, col: U32) -> Doc:
2203 decl_doc.cut(cut.go(record, Cut{[], [], False{}}), record, declaration, tab, col)
2204
2205def record_doc(
2206 is_import: Bool,
2207 head: Token,
2208 record: List<&2, Tree>,
2209 declaration: Bool,
2210 tab: U32,
2211 col: U32
2212) -> Doc:
2213 match is_import:
2214 case True{}:
2215 import_doc(head, record)
2216 case False{}:
2217 decl_doc(record, declaration, tab, col)
2218
2219# The text of the first tokens of `xs`, after `acc`, the texts so far, last first.
2220def joined(xs: List<&2, Tree>, acc: List<&2, String>) -> String:
2221 match xs:
2222 case Nil{}:
2223 Strings.onto(acc, "")
2224 case Con{x, rest}:
2225 joined(rest, Token.text(first(x)) <> acc)
2226
2227def is_unsafe(at: Bool, record: List<&2, Tree>) -> Bool:
2228 match at:
2229 case True{}:
2230 is_same(joined(record, []), "@unsafe")
2231 case False{}:
2232 False{}
2233
2234def gapped(gap: Bool, rows: List<&2, String>) -> List<&2, String>:
2235 match gap rows:
2236 case True{} Con{+row, +rest}:
2237 Bool.pick(List<&2, String>, String.is_empty(row), Con{row, rest}, Con{"", Con{row, rest}})
2238 case _ rows2:
2239 rows2
2240
2241def rows.line(+head: Token, +record: List<&2, Tree>, +set: Settings, r: Rows) -> Rows:
2242 match set r:
2243 case Settings{+tab, width, newline} Rows{levels, +blank, unsafe, prior, rows}:
2244 +col = Token.col(head)
2245 +levels2 = indented(levels, col)
2246 +declaration = is_declaration(Token.sym(head))
2247 +comment = is_comment(head)
2248 gap = Bool.pick(
2249 Bool,
2250 U32.is_zero(col),
2251 (declaration && Bool.not(prior) && Bool.not(unsafe)) || (comment && blank),
2252 blank
2253 )
2254 doc = record_doc(is_import(Token.sym(head)), head, record, declaration, tab, col)
2255 fuel = U32.to_nat(
2256 (fuel_per_point() * ((record_end(head, record) - Token.beg(head): U32) + 1): U32)
2257 )
2258 Rows{
2259 levels2,
2260 False{},
2261 is_unsafe(is_sym(Token.sym(head), At{}), record),
2262 comment,
2263 Con{print(doc, width, (depth(levels2) * tab: U32), newline, fuel), gapped(gap, rows)}
2264 }
2265
2266def rows.step(record: List<&2, Tree>, set: Settings, r: Rows) -> Rows:
2267 match record r:
2268 case Nil{} Rows{levels, blank, unsafe, comment, rows}:
2269 Rows{levels, True{}, unsafe, comment, rows}
2270 case Con{+x, xs} r2:
2271 rows.line(first(x), Con{x, xs}, set, r2)
2272
2273def rows.go(records: List<&2, List<&2, Tree>>, +set: Settings, r: Rows) -> Rows:
2274 match records:
2275 case Nil{}:
2276 r
2277 case Con{record, rest}:
2278 rows.go(rest, set, rows.step(record, set, r))
2279
2280def is_head_empty(rows: List<&2, String>) -> Bool:
2281 match rows:
2282 case Con{row, rest}:
2283 String.is_empty(row)
2284 case Nil{}:
2285 False{}
2286
2287def drop_empty(rows: List<&2, String>, go: Bool) -> List<&2, String>:
2288 match rows go:
2289 case Con{row, +rest} True{}:
2290 drop_empty(rest, is_head_empty(rest))
2291 case rest _:
2292 rest
2293
2294def join(rows: List<&2, String>, +newline: String, acc: String) -> String:
2295 match rows:
2296 case Nil{}:
2297 acc
2298 case Con{row, rest}:
2299 join(rest, newline, row ++ newline ++ acc)
2300
2301# The rows, last first, joined, without the empty rows at the end, with a
2302# newline. Each row has no trailing whitespace already.
2303def text_of.trimmed(rows: List<&2, String>, +newline: String) -> String:
2304 match rows:
2305 case Nil{}:
2306 newline
2307 case Con{row, rest}:
2308 join(rest, newline, row ++ newline)
2309
2310def text_of(+rows: List<&2, String>, newline: String) -> String:
2311 text_of.trimmed(drop_empty(rows, is_head_empty(rows)), newline)
2312
2313def format.text(r: Rows, newline: String) -> String:
2314 match r:
2315 case Rows{levels, blank, unsafe, comment, rows}:
2316 text_of(rows, newline)
2317
2318def format.of(nodes: Maybe<&2, List<&2, Tree>>, +set: Settings) -> Maybe<&2, String>:
2319 match nodes set:
2320 case None{} s:
2321 None{}
2322 case Some{ns} Settings{tab, width, +newline}:
2323 Some{
2324 format.text(
2325 rows.go(
2326 records(ns, Recs{[], []}),
2327 Settings{tab, width, newline},
2328 Rows{[0], False{}, False{}, False{}, []}
2329 ),
2330 newline
2331 )
2332 }
2333
2334# The formatted text, or None{} if its delimiters do not balance.
2335def format(source: String, set: Settings) -> Maybe<&2, String>:
2336 format.of(tree.go(tokens(source), None{}, Built{[], []}), set)
2337
2338# Rule
2339# ====
2340
2341# Where `a` and `b` first differ: `same` is whether the characters before `at` were the same.
2342def differ(a: String, b: String, at: U32, same: Bool) -> U32:
2343 match a b same:
2344 case _ _ False{}:
2345 (at - 1: U32)
2346 case SCon{x, xs} SCon{y, ys} True{}:
2347 differ(xs, ys, (at + 1: U32), Char.is_eq(x, y))
2348 case _ _ True{}:
2349 at
2350
2351# Where the line that `s` starts, at `i`, ends: before its newline, or at the end of the text.
2352def line_end(s: String, i: U32) -> U32:
2353 match s:
2354 case SNil{}:
2355 i
2356 case SCon{'\n', rest}:
2357 i
2358 case SCon{c, rest}:
2359 line_end(rest, (i + 1: U32))
2360
2361def judged(same: Bool, +out: String, +text: String, +path: String) -> List<&2, Lint.Finding>:
2362 match same:
2363 case True{}:
2364 []
2365 case False{}:
2366 +at = differ(out, text, 0, True{})
2367 [
2368 Lint.Finding{
2369 Lint.Warning{},
2370 Lint.Span{path, at, line_end(String.drop(text, U32.to_nat(at)), at)},
2371 "Apply canonical formatting.",
2372 [
2373 Lint.Fix{
2374 "Format the file",
2375 [Lint.Replace{Lint.Span{path, 0, U32.from_nat(String.length(text))}, out}]
2376 }
2377 ]
2378 }
2379 ]
2380
2381def judge(formatted: Maybe<&2, String>, +text: String, path: String) -> List<&2, Lint.Finding>:
2382 match formatted:
2383 case None{}:
2384 [
2385 Lint.Finding{
2386 Lint.Warning{},
2387 Lint.Span{path, 0, 0},
2388 "Cannot format this file: unbalanced delimiters.",
2389 []
2390 }
2391 ]
2392 case Some{+out}:
2393 judged(is_same(out, text), out, text, path)
2394
2395def first_newline(s: String) -> String:
2396 match s:
2397 case SNil{}:
2398 "\n"
2399 case SCon{'\r', SCon{'\n', rest}}:
2400 "\r\n"
2401 case SCon{'\n', rest}:
2402 "\n"
2403 case SCon{c, rest}:
2404 first_newline(rest)
2405
2406def newline.preserve(preserve: Bool, source: String) -> String:
2407 match preserve:
2408 case True{}:
2409 first_newline(source)
2410 case False{}:
2411 "\n"
2412
2413def newline.crlf(crlf: Bool, eol: String, source: String) -> String:
2414 match crlf:
2415 case True{}:
2416 "\r\n"
2417 case False{}:
2418 newline.preserve(String.eq(eol, "preserve"), source)
2419
2420def newline_of(+eol: String, source: String) -> String:
2421 newline.crlf(String.eq(eol, "crlf"), eol, source)
2422
2423def width_of(value: Maybe<&2, Lint.Value>) -> Maybe<&2, U32>:
2424 match value:
2425 case Some{Lint.Num{n}}:
2426 Some{n}
2427 case _:
2428 None{}
2429
2430def settings(+here: Lint.Here) -> Settings:
2431 Settings{
2432 Lint.Value.number(Lint.Here.option(here, "tabWidth")),
2433 width_of(Lint.Here.option(here, "wrapAtWidth")),
2434 newline_of(Lint.Value.text(Lint.Here.option(here, "endOfLine")), Lint.Here.text(here))
2435 }
2436
2437def run(+here: Lint.Here) -> Lint.Ask(List<&2, Lint.Finding>):
2438 Lint.done(
2439 judge(format(Lint.Here.text(here), settings(here)), Lint.Here.text(here), Lint.Here.path(here))
2440 )
2441
2442def options() -> List<&2, Lint.Declared>:
2443 [
2444 Lint.Declared{"tabWidth", Lint.Num{2}, Lint.NumberOption{1, 4294967295}},
2445 Lint.Declared{
2446 "wrapAtWidth",
2447 Lint.Num{100},
2448 Lint.AnyOption{[Lint.NumberOption{1, 4294967295}, Lint.TextOption{["never"]}]}
2449 },
2450 Lint.Declared{"endOfLine", Lint.Chars{"lf"}, Lint.TextOption{["lf", "crlf", "preserve"]}}
2451 ]
2452
2453def rule() -> Lint.Rule:
2454 Lint.Rule{Lint.Needs{False{}, False{}}, options(), Lint.OnFile{run}}