1 // Copyright 2012 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
12 * This pretty-printer is a direct reimplementation of Philip Karlton's
13 * Mesa pretty-printer, as described in appendix A of
15 * STAN-CS-79-770: "Pretty Printing", by Derek C. Oppen.
16 * Stanford Department of Computer Science, 1979.
18 * The algorithm's aim is to break a stream into as few lines as possible
19 * while respecting the indentation-consistency requirements of the enclosing
20 * block, and avoiding breaking at silly places on block boundaries, for
21 * example, between "x" and ")" in "x)".
23 * I am implementing this algorithm because it comes with 20 pages of
24 * documentation explaining its theory, and because it addresses the set of
25 * concerns I've seen other pretty-printers fall down on. Weirdly. Even though
26 * it's 32 years old. What can I say?
28 * Despite some redundancies and quirks in the way it's implemented in that
29 * paper, I've opted to keep the implementation here as similar as I can,
30 * changing only what was blatantly wrong, a typo, or sufficiently
31 * non-idiomatic rust that it really stuck out.
33 * In particular you'll see a certain amount of churn related to INTEGER vs.
34 * CARDINAL in the Mesa implementation. Mesa apparently interconverts the two
35 * somewhat readily? In any case, I've used uint for indices-in-buffers and
36 * ints for character-sizes-and-indentation-offsets. This respects the need
37 * for ints to "go negative" while carrying a pending-calculation balance, and
38 * helps differentiate all the numbers flying around internally (slightly).
40 * I also inverted the indentation arithmetic used in the print stack, since
41 * the Mesa implementation (somewhat randomly) stores the offset on the print
42 * stack in terms of margin-col rather than col itself. I store col.
44 * I also implemented a small change in the String token, in that I store an
45 * explicit length for the string. For most tokens this is just the length of
46 * the accompanying string. But it's necessary to permit it to differ, for
47 * encoding things that are supposed to "go on their own line" -- certain
48 * classes of comment and blank-line -- where relying on adjacent
49 * hardbreak-like Break tokens with long blankness indication doesn't actually
50 * work. To see why, consider when there is a "thing that should be on its own
51 * line" between two long blocks, say functions. If you put a hardbreak after
52 * each function (or before each) and the breaking algorithm decides to break
53 * there anyways (because the functions themselves are long) you wind up with
54 * extra blank lines. If you don't put hardbreaks you can wind up with the
55 * "thing which should be on its own line" not getting its own line in the
56 * rare case of "really small functions" or such. This re-occurs with comments
57 * and explicit blank lines. So in those cases we use a string with a payload
58 * we want isolated to a line and an explicit length that's huge, surrounded
59 * by two zero-length breaks. The algorithm will try its best to fit it on a
60 * line (which it can't) and so naturally place the content on its own line to
61 * avoid combining it with other lines and making matters even worse.
67 #[deriving(Clone, Eq)]
74 pub struct BreakToken {
80 pub struct BeginToken {
95 pub fn is_eof(&self) -> bool {
96 match *self { Eof => true, _ => false }
99 pub fn is_hardbreak_tok(&self) -> bool {
104 }) if bs == SIZE_INFINITY =>
112 pub fn tok_str(t: Token) -> ~str {
114 String(s, len) => return format!("STR({},{})", s, len),
115 Break(_) => return ~"BREAK",
116 Begin(_) => return ~"BEGIN",
117 End => return ~"END",
122 pub fn buf_str(toks: Vec<Token> , szs: Vec<int> , left: uint, right: uint,
125 assert_eq!(n, szs.len());
129 while i != right && L != 0u {
134 s.push_str(format!("{}={}", szs[i], tok_str(toks[i].clone())));
142 enum PrintStackBreak {
147 struct PrintStackElem {
149 pbreak: PrintStackBreak
152 static SIZE_INFINITY: int = 0xffff;
154 pub fn mk_printer(out: ~io::Writer, linewidth: uint) -> Printer {
155 // Yes 3, it makes the ring buffers big enough to never
157 let n: uint = 3 * linewidth;
158 debug!("mk_printer {}", linewidth);
159 let token: Vec<Token> = vec::from_elem(n, Eof);
160 let size: Vec<int> = vec::from_elem(n, 0);
161 let scan_stack: Vec<uint> = vec::from_elem(n, 0u);
165 margin: linewidth as int,
166 space: linewidth as int,
173 scan_stack: scan_stack,
174 scan_stack_empty: true,
177 print_stack: Vec::new(),
178 pending_indentation: 0
184 * In case you do not have the paper, here is an explanation of what's going
187 * There is a stream of input tokens flowing through this printer.
189 * The printer buffers up to 3N tokens inside itself, where N is linewidth.
190 * Yes, linewidth is chars and tokens are multi-char, but in the worst
191 * case every token worth buffering is 1 char long, so it's ok.
193 * Tokens are String, Break, and Begin/End to delimit blocks.
195 * Begin tokens can carry an offset, saying "how far to indent when you break
196 * inside here", as well as a flag indicating "consistent" or "inconsistent"
197 * breaking. Consistent breaking means that after the first break, no attempt
198 * will be made to flow subsequent breaks together onto lines. Inconsistent
199 * is the opposite. Inconsistent breaking example would be, say:
201 * foo(hello, there, good, friends)
203 * breaking inconsistently to become
208 * whereas a consistent breaking would yield:
215 * That is, in the consistent-break blocks we value vertical alignment
216 * more than the ability to cram stuff onto a line. But in all cases if it
217 * can make a block a one-liner, it'll do so.
219 * Carrying on with high-level logic:
221 * The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
222 * 'right' indices denote the active portion of the ring buffer as well as
223 * describing hypothetical points-in-the-infinite-stream at most 3N tokens
224 * apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
225 * between using 'left' and 'right' terms to denote the wrapepd-to-ring-buffer
226 * and point-in-infinite-stream senses freely.
228 * There is a parallel ring buffer, 'size', that holds the calculated size of
229 * each token. Why calculated? Because for Begin/End pairs, the "size"
230 * includes everything betwen the pair. That is, the "size" of Begin is
231 * actually the sum of the sizes of everything between Begin and the paired
232 * End that follows. Since that is arbitrarily far in the future, 'size' is
233 * being rewritten regularly while the printer runs; in fact most of the
234 * machinery is here to work out 'size' entries on the fly (and give up when
235 * they're so obviously over-long that "infinity" is a good enough
236 * approximation for purposes of line breaking).
238 * The "input side" of the printer is managed as an abstract process called
239 * SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
240 * manage calculating 'size'. SCAN is, in other words, the process of
241 * calculating 'size' entries.
243 * The "output side" of the printer is managed by an abstract process called
244 * PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
245 * do with each token/size pair it consumes as it goes. It's trying to consume
246 * the entire buffered window, but can't output anything until the size is >=
247 * 0 (sizes are set to negative while they're pending calculation).
249 * So SCAN takes input and buffers tokens and pending calculations, while
250 * PRINT gobbles up completed calculations and tokens from the buffer. The
251 * theory is that the two can never get more than 3N tokens apart, because
252 * once there's "obviously" too much data to fit on a line, in a size
253 * calculation, SCAN will write "infinity" to the size and let PRINT consume
256 * In this implementation (following the paper, again) the SCAN process is
257 * the method called 'pretty_print', and the 'PRINT' process is the method
263 margin: int, // width of lines we're constrained to
264 space: int, // number of spaces left on line
265 left: uint, // index of left side of input stream
266 right: uint, // index of right side of input stream
267 token: Vec<Token> , // ring-buffr stream goes through
268 size: Vec<int> , // ring-buffer of calculated sizes
269 left_total: int, // running size of stream "...left"
270 right_total: int, // running size of stream "...right"
271 // pseudo-stack, really a ring too. Holds the
272 // primary-ring-buffers index of the Begin that started the
273 // current block, possibly with the most recent Break after that
274 // Begin (if there is any) on top of it. Stuff is flushed off the
275 // bottom as it becomes irrelevant due to the primary ring-buffer
277 scan_stack: Vec<uint> ,
278 scan_stack_empty: bool, // top==bottom disambiguator
279 top: uint, // index of top of scan_stack
280 bottom: uint, // index of bottom of scan_stack
281 // stack of blocks-in-progress being flushed by print
282 print_stack: Vec<PrintStackElem> ,
283 // buffered indentation to avoid writing trailing whitespace
284 pending_indentation: int,
288 pub fn last_token(&mut self) -> Token {
289 self.token[self.right].clone()
291 // be very careful with this!
292 pub fn replace_last_token(&mut self, t: Token) {
293 self.token[self.right] = t;
295 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
296 debug!("pp ~[{},{}]", self.left, self.right);
299 if !self.scan_stack_empty {
301 let left = self.token[self.left].clone();
302 try!(self.advance_left(left, self.size[self.left]));
308 if self.scan_stack_empty {
310 self.right_total = 1;
313 } else { self.advance_right(); }
314 debug!("pp Begin({})/buffer ~[{},{}]",
315 b.offset, self.left, self.right);
316 self.token[self.right] = t;
317 self.size[self.right] = -self.right_total;
318 self.scan_push(self.right);
322 if self.scan_stack_empty {
323 debug!("pp End/print ~[{},{}]", self.left, self.right);
326 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
327 self.advance_right();
328 self.token[self.right] = t;
329 self.size[self.right] = -1;
330 self.scan_push(self.right);
335 if self.scan_stack_empty {
337 self.right_total = 1;
340 } else { self.advance_right(); }
341 debug!("pp Break({})/buffer ~[{},{}]",
342 b.offset, self.left, self.right);
344 self.scan_push(self.right);
345 self.token[self.right] = t;
346 self.size[self.right] = -self.right_total;
347 self.right_total += b.blank_space;
350 String(ref s, len) => {
351 if self.scan_stack_empty {
352 debug!("pp String('{}')/print ~[{},{}]",
353 *s, self.left, self.right);
354 self.print(t.clone(), len)
356 debug!("pp String('{}')/buffer ~[{},{}]",
357 *s, self.left, self.right);
358 self.advance_right();
359 self.token[self.right] = t.clone();
360 self.size[self.right] = len;
361 self.right_total += len;
367 pub fn check_stream(&mut self) -> io::IoResult<()> {
368 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
369 self.left, self.right, self.left_total, self.right_total);
370 if self.right_total - self.left_total > self.space {
371 debug!("scan window is {}, longer than space on line ({})",
372 self.right_total - self.left_total, self.space);
373 if !self.scan_stack_empty {
374 if self.left == self.scan_stack[self.bottom] {
375 debug!("setting {} to infinity and popping", self.left);
376 self.size[self.scan_pop_bottom()] = SIZE_INFINITY;
379 let left = self.token[self.left].clone();
380 try!(self.advance_left(left, self.size[self.left]));
381 if self.left != self.right {
382 try!(self.check_stream());
387 pub fn scan_push(&mut self, x: uint) {
388 debug!("scan_push {}", x);
389 if self.scan_stack_empty {
390 self.scan_stack_empty = false;
393 self.top %= self.buf_len;
394 assert!((self.top != self.bottom));
396 self.scan_stack[self.top] = x;
398 pub fn scan_pop(&mut self) -> uint {
399 assert!((!self.scan_stack_empty));
400 let x = self.scan_stack[self.top];
401 if self.top == self.bottom {
402 self.scan_stack_empty = true;
403 } else { self.top += self.buf_len - 1u; self.top %= self.buf_len; }
406 pub fn scan_top(&mut self) -> uint {
407 assert!((!self.scan_stack_empty));
408 return self.scan_stack[self.top];
410 pub fn scan_pop_bottom(&mut self) -> uint {
411 assert!((!self.scan_stack_empty));
412 let x = self.scan_stack[self.bottom];
413 if self.top == self.bottom {
414 self.scan_stack_empty = true;
415 } else { self.bottom += 1u; self.bottom %= self.buf_len; }
418 pub fn advance_right(&mut self) {
420 self.right %= self.buf_len;
421 assert!((self.right != self.left));
423 pub fn advance_left(&mut self, x: Token, L: int) -> io::IoResult<()> {
424 debug!("advnce_left ~[{},{}], sizeof({})={}", self.left, self.right,
427 let ret = self.print(x.clone(), L);
429 Break(b) => self.left_total += b.blank_space,
431 assert_eq!(len, L); self.left_total += len;
435 if self.left != self.right {
437 self.left %= self.buf_len;
438 let left = self.token[self.left].clone();
439 try!(self.advance_left(left, self.size[self.left]));
446 pub fn check_stack(&mut self, k: int) {
447 if !self.scan_stack_empty {
448 let x = self.scan_top();
449 match self.token[x] {
452 self.size[self.scan_pop()] = self.size[x] +
454 self.check_stack(k - 1);
458 // paper says + not =, but that makes no sense.
459 self.size[self.scan_pop()] = 1;
460 self.check_stack(k + 1);
463 self.size[self.scan_pop()] = self.size[x] + self.right_total;
464 if k > 0 { self.check_stack(k); }
469 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
470 debug!("NEWLINE {}", amount);
471 let ret = write!(self.out, "\n");
472 self.pending_indentation = 0;
476 pub fn indent(&mut self, amount: int) {
477 debug!("INDENT {}", amount);
478 self.pending_indentation += amount;
480 pub fn get_top(&mut self) -> PrintStackElem {
481 let print_stack = &mut self.print_stack;
482 let n = print_stack.len();
488 pbreak: Broken(Inconsistent)
492 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
493 while self.pending_indentation > 0 {
494 try!(write!(self.out, " "));
495 self.pending_indentation -= 1;
497 write!(self.out, "{}", s)
499 pub fn print(&mut self, x: Token, L: int) -> io::IoResult<()> {
500 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), L,
502 debug!("{}", buf_str(self.token.clone(),
510 let col = self.margin - self.space + b.offset;
511 debug!("print Begin -> push broken block at col {}", col);
512 self.print_stack.push(PrintStackElem {
514 pbreak: Broken(b.breaks)
517 debug!("print Begin -> push fitting block");
518 self.print_stack.push(PrintStackElem {
526 debug!("print End -> pop End");
527 let print_stack = &mut self.print_stack;
528 assert!((print_stack.len() != 0u));
529 print_stack.pop().unwrap();
533 let top = self.get_top();
536 debug!("print Break({}) in fitting block", b.blank_space);
537 self.space -= b.blank_space;
538 self.indent(b.blank_space);
541 Broken(Consistent) => {
542 debug!("print Break({}+{}) in consistent block",
543 top.offset, b.offset);
544 let ret = self.print_newline(top.offset + b.offset);
545 self.space = self.margin - (top.offset + b.offset);
548 Broken(Inconsistent) => {
550 debug!("print Break({}+{}) w/ newline in inconsistent",
551 top.offset, b.offset);
552 let ret = self.print_newline(top.offset + b.offset);
553 self.space = self.margin - (top.offset + b.offset);
556 debug!("print Break({}) w/o newline in inconsistent",
558 self.indent(b.blank_space);
559 self.space -= b.blank_space;
566 debug!("print String({})", s);
568 // assert!(L <= space);
573 // Eof should never get here.
580 // Convenience functions to talk to the printer.
583 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
584 p.pretty_print(Begin(BeginToken {
585 offset: indent as int,
590 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
591 rbox(p, indent, Inconsistent)
594 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
595 rbox(p, indent, Consistent)
598 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
599 p.pretty_print(Break(BreakToken {
601 blank_space: n as int
605 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
607 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
609 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
610 p.pretty_print(String(/* bad */ wrd.to_str(), wrd.len() as int))
613 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
614 p.pretty_print(String(/* bad */ wrd.to_str(), SIZE_INFINITY))
617 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
618 p.pretty_print(String(/* bad */ wrd.to_str(), 0))
621 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
622 break_offset(p, n, 0)
625 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
629 pub fn space(p: &mut Printer) -> io::IoResult<()> {
633 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
634 spaces(p, SIZE_INFINITY as uint)
637 pub fn hardbreak_tok_offset(off: int) -> Token {
638 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
641 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }