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.
11 //! This pretty-printer is a direct reimplementation of Philip Karlton's
12 //! 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.
19 //! The algorithm's aim is to break a stream into as few lines as possible
20 //! while respecting the indentation-consistency requirements of the enclosing
21 //! block, and avoiding breaking at silly places on block boundaries, for
22 //! example, between "x" and ")" in "x)".
24 //! I am implementing this algorithm because it comes with 20 pages of
25 //! documentation explaining its theory, and because it addresses the set of
26 //! concerns I've seen other pretty-printers fall down on. Weirdly. Even though
27 //! it's 32 years old. What can I say?
29 //! Despite some redundancies and quirks in the way it's implemented in that
30 //! paper, I've opted to keep the implementation here as similar as I can,
31 //! changing only what was blatantly wrong, a typo, or sufficiently
32 //! non-idiomatic rust that it really stuck out.
34 //! In particular you'll see a certain amount of churn related to INTEGER vs.
35 //! CARDINAL in the Mesa implementation. Mesa apparently interconverts the two
36 //! somewhat readily? In any case, I've used usize for indices-in-buffers and
37 //! ints for character-sizes-and-indentation-offsets. This respects the need
38 //! for ints to "go negative" while carrying a pending-calculation balance, and
39 //! helps differentiate all the numbers flying around internally (slightly).
41 //! I also inverted the indentation arithmetic used in the print stack, since
42 //! the Mesa implementation (somewhat randomly) stores the offset on the print
43 //! stack in terms of margin-col rather than col itself. I store col.
45 //! I also implemented a small change in the String token, in that I store an
46 //! explicit length for the string. For most tokens this is just the length of
47 //! the accompanying string. But it's necessary to permit it to differ, for
48 //! encoding things that are supposed to "go on their own line" -- certain
49 //! classes of comment and blank-line -- where relying on adjacent
50 //! hardbreak-like Break tokens with long blankness indication doesn't actually
51 //! work. To see why, consider when there is a "thing that should be on its own
52 //! line" between two long blocks, say functions. If you put a hardbreak after
53 //! each function (or before each) and the breaking algorithm decides to break
54 //! there anyways (because the functions themselves are long) you wind up with
55 //! extra blank lines. If you don't put hardbreaks you can wind up with the
56 //! "thing which should be on its own line" not getting its own line in the
57 //! rare case of "really small functions" or such. This re-occurs with comments
58 //! and explicit blank lines. So in those cases we use a string with a payload
59 //! we want isolated to a line and an explicit length that's huge, surrounded
60 //! by two zero-length breaks. The algorithm will try its best to fit it on a
61 //! line (which it can't) and so naturally place the content on its own line to
62 //! avoid combining it with other lines and making matters even worse.
66 //! In case you do not have the paper, here is an explanation of what's going
69 //! There is a stream of input tokens flowing through this printer.
71 //! The printer buffers up to 3N tokens inside itself, where N is linewidth.
72 //! Yes, linewidth is chars and tokens are multi-char, but in the worst
73 //! case every token worth buffering is 1 char long, so it's ok.
75 //! Tokens are String, Break, and Begin/End to delimit blocks.
77 //! Begin tokens can carry an offset, saying "how far to indent when you break
78 //! inside here", as well as a flag indicating "consistent" or "inconsistent"
79 //! breaking. Consistent breaking means that after the first break, no attempt
80 //! will be made to flow subsequent breaks together onto lines. Inconsistent
81 //! is the opposite. Inconsistent breaking example would be, say:
84 //! foo(hello, there, good, friends)
87 //! breaking inconsistently to become
94 //! whereas a consistent breaking would yield:
103 //! That is, in the consistent-break blocks we value vertical alignment
104 //! more than the ability to cram stuff onto a line. But in all cases if it
105 //! can make a block a one-liner, it'll do so.
107 //! Carrying on with high-level logic:
109 //! The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
110 //! 'right' indices denote the active portion of the ring buffer as well as
111 //! describing hypothetical points-in-the-infinite-stream at most 3N tokens
112 //! apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
113 //! between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
114 //! and point-in-infinite-stream senses freely.
116 //! There is a parallel ring buffer, 'size', that holds the calculated size of
117 //! each token. Why calculated? Because for Begin/End pairs, the "size"
118 //! includes everything between the pair. That is, the "size" of Begin is
119 //! actually the sum of the sizes of everything between Begin and the paired
120 //! End that follows. Since that is arbitrarily far in the future, 'size' is
121 //! being rewritten regularly while the printer runs; in fact most of the
122 //! machinery is here to work out 'size' entries on the fly (and give up when
123 //! they're so obviously over-long that "infinity" is a good enough
124 //! approximation for purposes of line breaking).
126 //! The "input side" of the printer is managed as an abstract process called
127 //! SCAN, which uses 'scan_stack', to manage calculating 'size'. SCAN is, in
128 //! other words, the process of calculating 'size' entries.
130 //! The "output side" of the printer is managed by an abstract process called
131 //! PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
132 //! do with each token/size pair it consumes as it goes. It's trying to consume
133 //! the entire buffered window, but can't output anything until the size is >=
134 //! 0 (sizes are set to negative while they're pending calculation).
136 //! So SCAN takes input and buffers tokens and pending calculations, while
137 //! PRINT gobbles up completed calculations and tokens from the buffer. The
138 //! theory is that the two can never get more than 3N tokens apart, because
139 //! once there's "obviously" too much data to fit on a line, in a size
140 //! calculation, SCAN will write "infinity" to the size and let PRINT consume
143 //! In this implementation (following the paper, again) the SCAN process is
144 //! the method called `Printer::pretty_print`, and the 'PRINT' process is the method
145 //! called `Printer::print`.
147 use std::collections::VecDeque;
151 /// How to break. Described in more detail in the module docs.
152 #[derive(Clone, Copy, PartialEq)]
158 #[derive(Clone, Copy)]
159 pub struct BreakToken {
164 #[derive(Clone, Copy)]
165 pub struct BeginToken {
172 String(String, isize),
180 pub fn is_eof(&self) -> bool {
187 pub fn is_hardbreak_tok(&self) -> bool {
189 Token::Break(BreakToken {
192 }) if bs == SIZE_INFINITY =>
200 impl fmt::Display for Token {
201 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
203 Token::String(ref s, len) => write!(f, "STR({},{})", s, len),
204 Token::Break(_) => f.write_str("BREAK"),
205 Token::Begin(_) => f.write_str("BEGIN"),
206 Token::End => f.write_str("END"),
207 Token::Eof => f.write_str("EOF"),
212 fn buf_str(buf: &[BufEntry], left: usize, right: usize, lim: usize) -> String {
216 let mut s = String::from("[");
217 while i != right && l != 0 {
222 s.push_str(&format!("{}={}", buf[i].size, &buf[i].token));
230 #[derive(Copy, Clone)]
231 pub enum PrintStackBreak {
236 #[derive(Copy, Clone)]
237 pub struct PrintStackElem {
239 pbreak: PrintStackBreak
242 const SIZE_INFINITY: isize = 0xffff;
244 pub fn mk_printer<'a>(out: Box<io::Write+'a>, linewidth: usize) -> Printer<'a> {
245 // Yes 55, it makes the ring buffers big enough to never fall behind.
246 let n: usize = 55 * linewidth;
247 debug!("mk_printer {}", linewidth);
251 margin: linewidth as isize,
252 space: linewidth as isize,
255 buf: vec![BufEntry { token: Token::Eof, size: 0 }; n],
258 scan_stack: VecDeque::new(),
259 print_stack: Vec::new(),
260 pending_indentation: 0
264 pub struct Printer<'a> {
265 pub out: Box<io::Write+'a>,
267 /// Width of lines we're constrained to
269 /// Number of spaces left on line
271 /// Index of left side of input stream
273 /// Index of right side of input stream
275 /// Ring-buffer of tokens and calculated sizes
277 /// Running size of stream "...left"
279 /// Running size of stream "...right"
281 /// Pseudo-stack, really a ring too. Holds the
282 /// primary-ring-buffers index of the Begin that started the
283 /// current block, possibly with the most recent Break after that
284 /// Begin (if there is any) on top of it. Stuff is flushed off the
285 /// bottom as it becomes irrelevant due to the primary ring-buffer
287 scan_stack: VecDeque<usize>,
288 /// Stack of blocks-in-progress being flushed by print
289 print_stack: Vec<PrintStackElem> ,
290 /// Buffered indentation to avoid writing trailing whitespace
291 pending_indentation: isize,
300 impl<'a> Printer<'a> {
301 pub fn last_token(&mut self) -> Token {
302 self.buf[self.right].token.clone()
304 /// be very careful with this!
305 pub fn replace_last_token(&mut self, t: Token) {
306 self.buf[self.right].token = t;
308 pub fn pretty_print(&mut self, token: Token) -> io::Result<()> {
309 debug!("pp Vec<{},{}>", self.left, self.right);
312 if !self.scan_stack.is_empty() {
314 self.advance_left()?;
320 if self.scan_stack.is_empty() {
322 self.right_total = 1;
325 } else { self.advance_right(); }
326 debug!("pp Begin({})/buffer Vec<{},{}>",
327 b.offset, self.left, self.right);
328 self.buf[self.right] = BufEntry { token: token, size: -self.right_total };
329 let right = self.right;
330 self.scan_push(right);
334 if self.scan_stack.is_empty() {
335 debug!("pp End/print Vec<{},{}>", self.left, self.right);
338 debug!("pp End/buffer Vec<{},{}>", self.left, self.right);
339 self.advance_right();
340 self.buf[self.right] = BufEntry { token: token, size: -1 };
341 let right = self.right;
342 self.scan_push(right);
347 if self.scan_stack.is_empty() {
349 self.right_total = 1;
352 } else { self.advance_right(); }
353 debug!("pp Break({})/buffer Vec<{},{}>",
354 b.offset, self.left, self.right);
356 let right = self.right;
357 self.scan_push(right);
358 self.buf[self.right] = BufEntry { token: token, size: -self.right_total };
359 self.right_total += b.blank_space;
362 Token::String(s, len) => {
363 if self.scan_stack.is_empty() {
364 debug!("pp String('{}')/print Vec<{},{}>",
365 s, self.left, self.right);
366 self.print(Token::String(s, len), len)
368 debug!("pp String('{}')/buffer Vec<{},{}>",
369 s, self.left, self.right);
370 self.advance_right();
371 self.buf[self.right] = BufEntry { token: Token::String(s, len), size: len };
372 self.right_total += len;
378 pub fn check_stream(&mut self) -> io::Result<()> {
379 debug!("check_stream Vec<{}, {}> with left_total={}, right_total={}",
380 self.left, self.right, self.left_total, self.right_total);
381 if self.right_total - self.left_total > self.space {
382 debug!("scan window is {}, longer than space on line ({})",
383 self.right_total - self.left_total, self.space);
384 if Some(&self.left) == self.scan_stack.back() {
385 debug!("setting {} to infinity and popping", self.left);
386 let scanned = self.scan_pop_bottom();
387 self.buf[scanned].size = SIZE_INFINITY;
389 self.advance_left()?;
390 if self.left != self.right {
391 self.check_stream()?;
396 pub fn scan_push(&mut self, x: usize) {
397 debug!("scan_push {}", x);
398 self.scan_stack.push_front(x);
400 pub fn scan_pop(&mut self) -> usize {
401 self.scan_stack.pop_front().unwrap()
403 pub fn scan_top(&mut self) -> usize {
404 *self.scan_stack.front().unwrap()
406 pub fn scan_pop_bottom(&mut self) -> usize {
407 self.scan_stack.pop_back().unwrap()
409 pub fn advance_right(&mut self) {
411 self.right %= self.buf_len;
412 assert!(self.right != self.left);
414 pub fn advance_left(&mut self) -> io::Result<()> {
415 debug!("advance_left Vec<{},{}>, sizeof({})={}", self.left, self.right,
416 self.left, self.buf[self.left].size);
418 let mut left_size = self.buf[self.left].size;
420 while left_size >= 0 {
421 let left = self.buf[self.left].token.clone();
423 let len = match left {
424 Token::Break(b) => b.blank_space,
425 Token::String(_, len) => {
426 assert_eq!(len, left_size);
432 self.print(left, left_size)?;
434 self.left_total += len;
436 if self.left == self.right {
441 self.left %= self.buf_len;
443 left_size = self.buf[self.left].size;
448 pub fn check_stack(&mut self, k: isize) {
449 if !self.scan_stack.is_empty() {
450 let x = self.scan_top();
451 match self.buf[x].token {
454 let popped = self.scan_pop();
455 self.buf[popped].size = self.buf[x].size + self.right_total;
456 self.check_stack(k - 1);
460 // paper says + not =, but that makes no sense.
461 let popped = self.scan_pop();
462 self.buf[popped].size = 1;
463 self.check_stack(k + 1);
466 let popped = self.scan_pop();
467 self.buf[popped].size = self.buf[x].size + self.right_total;
475 pub fn print_newline(&mut self, amount: isize) -> io::Result<()> {
476 debug!("NEWLINE {}", amount);
477 let ret = write!(self.out, "\n");
478 self.pending_indentation = 0;
482 pub fn indent(&mut self, amount: isize) {
483 debug!("INDENT {}", amount);
484 self.pending_indentation += amount;
486 pub fn get_top(&mut self) -> PrintStackElem {
487 match self.print_stack.last() {
489 None => PrintStackElem {
491 pbreak: PrintStackBreak::Broken(Breaks::Inconsistent)
495 pub fn print_str(&mut self, s: &str) -> io::Result<()> {
496 while self.pending_indentation > 0 {
497 write!(self.out, " ")?;
498 self.pending_indentation -= 1;
500 write!(self.out, "{}", s)
502 pub fn print(&mut self, token: Token, l: isize) -> io::Result<()> {
503 debug!("print {} {} (remaining line space={})", token, l,
505 debug!("{}", buf_str(&self.buf,
512 let col = self.margin - self.space + b.offset;
513 debug!("print Begin -> push broken block at col {}", col);
514 self.print_stack.push(PrintStackElem {
516 pbreak: PrintStackBreak::Broken(b.breaks)
519 debug!("print Begin -> push fitting block");
520 self.print_stack.push(PrintStackElem {
522 pbreak: PrintStackBreak::Fits
528 debug!("print End -> pop End");
529 let print_stack = &mut self.print_stack;
530 assert!(!print_stack.is_empty());
531 print_stack.pop().unwrap();
535 let top = self.get_top();
537 PrintStackBreak::Fits => {
538 debug!("print Break({}) in fitting block", b.blank_space);
539 self.space -= b.blank_space;
540 self.indent(b.blank_space);
543 PrintStackBreak::Broken(Breaks::Consistent) => {
544 debug!("print Break({}+{}) in consistent block",
545 top.offset, b.offset);
546 let ret = self.print_newline(top.offset + b.offset);
547 self.space = self.margin - (top.offset + b.offset);
550 PrintStackBreak::Broken(Breaks::Inconsistent) => {
552 debug!("print Break({}+{}) w/ newline in inconsistent",
553 top.offset, b.offset);
554 let ret = self.print_newline(top.offset + b.offset);
555 self.space = self.margin - (top.offset + b.offset);
558 debug!("print Break({}) w/o newline in inconsistent",
560 self.indent(b.blank_space);
561 self.space -= b.blank_space;
567 Token::String(ref s, len) => {
568 debug!("print String({})", s);
570 // assert!(l <= space);
575 // Eof should never get here.
582 // Convenience functions to talk to the printer.
585 pub fn rbox(p: &mut Printer, indent: usize, b: Breaks) -> io::Result<()> {
586 p.pretty_print(Token::Begin(BeginToken {
587 offset: indent as isize,
592 /// Inconsistent breaking box
593 pub fn ibox(p: &mut Printer, indent: usize) -> io::Result<()> {
594 rbox(p, indent, Breaks::Inconsistent)
597 /// Consistent breaking box
598 pub fn cbox(p: &mut Printer, indent: usize) -> io::Result<()> {
599 rbox(p, indent, Breaks::Consistent)
602 pub fn break_offset(p: &mut Printer, n: usize, off: isize) -> io::Result<()> {
603 p.pretty_print(Token::Break(BreakToken {
605 blank_space: n as isize
609 pub fn end(p: &mut Printer) -> io::Result<()> {
610 p.pretty_print(Token::End)
613 pub fn eof(p: &mut Printer) -> io::Result<()> {
614 p.pretty_print(Token::Eof)
617 pub fn word(p: &mut Printer, wrd: &str) -> io::Result<()> {
618 p.pretty_print(Token::String(wrd.to_string(), wrd.len() as isize))
621 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::Result<()> {
622 p.pretty_print(Token::String(wrd.to_string(), SIZE_INFINITY))
625 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::Result<()> {
626 p.pretty_print(Token::String(wrd.to_string(), 0))
629 pub fn spaces(p: &mut Printer, n: usize) -> io::Result<()> {
630 break_offset(p, n, 0)
633 pub fn zerobreak(p: &mut Printer) -> io::Result<()> {
637 pub fn space(p: &mut Printer) -> io::Result<()> {
641 pub fn hardbreak(p: &mut Printer) -> io::Result<()> {
642 spaces(p, SIZE_INFINITY as usize)
645 pub fn hardbreak_tok_offset(off: isize) -> Token {
646 Token::Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
649 pub fn hardbreak_tok() -> Token {
650 hardbreak_tok_offset(0)