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 the
144 //! methods called `Printer::pretty_print_*`, and the 'PRINT' process is the
145 //! method called `Printer::print`.
147 use std::collections::VecDeque;
150 use std::borrow::Cow;
152 /// How to break. Described in more detail in the module docs.
153 #[derive(Clone, Copy, PartialEq)]
159 #[derive(Clone, Copy)]
160 pub struct BreakToken {
165 #[derive(Clone, Copy)]
166 pub struct BeginToken {
173 // In practice a string token contains either a `&'static str` or a
174 // `String`. `Cow` is overkill for this because we never modify the data,
175 // but it's more convenient than rolling our own more specialized type.
176 String(Cow<'static, str>, isize),
184 pub fn is_eof(&self) -> bool {
191 pub fn is_hardbreak_tok(&self) -> bool {
193 Token::Break(BreakToken {
196 }) if bs == SIZE_INFINITY =>
204 impl fmt::Display for Token {
205 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
207 Token::String(ref s, len) => write!(f, "STR({},{})", s, len),
208 Token::Break(_) => f.write_str("BREAK"),
209 Token::Begin(_) => f.write_str("BEGIN"),
210 Token::End => f.write_str("END"),
211 Token::Eof => f.write_str("EOF"),
216 fn buf_str(buf: &[BufEntry], left: usize, right: usize, lim: usize) -> String {
220 let mut s = String::from("[");
221 while i != right && l != 0 {
226 s.push_str(&format!("{}={}", buf[i].size, &buf[i].token));
234 #[derive(Copy, Clone)]
235 pub enum PrintStackBreak {
240 #[derive(Copy, Clone)]
241 pub struct PrintStackElem {
243 pbreak: PrintStackBreak
246 const SIZE_INFINITY: isize = 0xffff;
248 pub fn mk_printer<'a>(out: Box<dyn io::Write+'a>, linewidth: usize) -> Printer<'a> {
249 // Yes 55, it makes the ring buffers big enough to never fall behind.
250 let n: usize = 55 * linewidth;
251 debug!("mk_printer {}", linewidth);
255 margin: linewidth as isize,
256 space: linewidth as isize,
259 // Initialize a single entry; advance_right() will extend it on demand
260 // up to `buf_max_len` elements.
261 buf: vec![BufEntry::default()],
264 scan_stack: VecDeque::new(),
265 print_stack: Vec::new(),
266 pending_indentation: 0
270 pub struct Printer<'a> {
271 out: Box<dyn io::Write+'a>,
273 /// Width of lines we're constrained to
275 /// Number of spaces left on line
277 /// Index of left side of input stream
279 /// Index of right side of input stream
281 /// Ring-buffer of tokens and calculated sizes
283 /// Running size of stream "...left"
285 /// Running size of stream "...right"
287 /// Pseudo-stack, really a ring too. Holds the
288 /// primary-ring-buffers index of the Begin that started the
289 /// current block, possibly with the most recent Break after that
290 /// Begin (if there is any) on top of it. Stuff is flushed off the
291 /// bottom as it becomes irrelevant due to the primary ring-buffer
293 scan_stack: VecDeque<usize>,
294 /// Stack of blocks-in-progress being flushed by print
295 print_stack: Vec<PrintStackElem> ,
296 /// Buffered indentation to avoid writing trailing whitespace
297 pending_indentation: isize,
306 impl Default for BufEntry {
307 fn default() -> Self {
308 BufEntry { token: Token::Eof, size: 0 }
312 impl<'a> Printer<'a> {
313 pub fn last_token(&mut self) -> Token {
314 self.buf[self.right].token.clone()
317 /// Be very careful with this!
318 pub fn replace_last_token(&mut self, t: Token) {
319 self.buf[self.right].token = t;
322 fn pretty_print_eof(&mut self) -> io::Result<()> {
323 if !self.scan_stack.is_empty() {
325 self.advance_left()?;
331 fn pretty_print_begin(&mut self, b: BeginToken) -> io::Result<()> {
332 if self.scan_stack.is_empty() {
334 self.right_total = 1;
338 self.advance_right();
340 debug!("pp Begin({})/buffer Vec<{},{}>",
341 b.offset, self.left, self.right);
342 self.buf[self.right] = BufEntry { token: Token::Begin(b), size: -self.right_total };
343 let right = self.right;
344 self.scan_push(right);
348 fn pretty_print_end(&mut self) -> io::Result<()> {
349 if self.scan_stack.is_empty() {
350 debug!("pp End/print Vec<{},{}>", self.left, self.right);
353 debug!("pp End/buffer Vec<{},{}>", self.left, self.right);
354 self.advance_right();
355 self.buf[self.right] = BufEntry { token: Token::End, size: -1 };
356 let right = self.right;
357 self.scan_push(right);
362 fn pretty_print_break(&mut self, b: BreakToken) -> io::Result<()> {
363 if self.scan_stack.is_empty() {
365 self.right_total = 1;
369 self.advance_right();
371 debug!("pp Break({})/buffer Vec<{},{}>",
372 b.offset, self.left, self.right);
374 let right = self.right;
375 self.scan_push(right);
376 self.buf[self.right] = BufEntry { token: Token::Break(b), size: -self.right_total };
377 self.right_total += b.blank_space;
381 fn pretty_print_string(&mut self, s: Cow<'static, str>, len: isize) -> io::Result<()> {
382 if self.scan_stack.is_empty() {
383 debug!("pp String('{}')/print Vec<{},{}>",
384 s, self.left, self.right);
385 self.print_string(s, len)
387 debug!("pp String('{}')/buffer Vec<{},{}>",
388 s, self.left, self.right);
389 self.advance_right();
390 self.buf[self.right] = BufEntry { token: Token::String(s, len), size: len };
391 self.right_total += len;
396 pub fn check_stream(&mut self) -> io::Result<()> {
397 debug!("check_stream Vec<{}, {}> with left_total={}, right_total={}",
398 self.left, self.right, self.left_total, self.right_total);
399 if self.right_total - self.left_total > self.space {
400 debug!("scan window is {}, longer than space on line ({})",
401 self.right_total - self.left_total, self.space);
402 if Some(&self.left) == self.scan_stack.back() {
403 debug!("setting {} to infinity and popping", self.left);
404 let scanned = self.scan_pop_bottom();
405 self.buf[scanned].size = SIZE_INFINITY;
407 self.advance_left()?;
408 if self.left != self.right {
409 self.check_stream()?;
415 pub fn scan_push(&mut self, x: usize) {
416 debug!("scan_push {}", x);
417 self.scan_stack.push_front(x);
420 pub fn scan_pop(&mut self) -> usize {
421 self.scan_stack.pop_front().unwrap()
424 pub fn scan_top(&mut self) -> usize {
425 *self.scan_stack.front().unwrap()
428 pub fn scan_pop_bottom(&mut self) -> usize {
429 self.scan_stack.pop_back().unwrap()
432 pub fn advance_right(&mut self) {
434 self.right %= self.buf_max_len;
435 // Extend the buf if necessary.
436 if self.right == self.buf.len() {
437 self.buf.push(BufEntry::default());
439 assert_ne!(self.right, self.left);
442 pub fn advance_left(&mut self) -> io::Result<()> {
443 debug!("advance_left Vec<{},{}>, sizeof({})={}", self.left, self.right,
444 self.left, self.buf[self.left].size);
446 let mut left_size = self.buf[self.left].size;
448 while left_size >= 0 {
449 let left = self.buf[self.left].token.clone();
451 let len = match left {
452 Token::Break(b) => b.blank_space,
453 Token::String(_, len) => {
454 assert_eq!(len, left_size);
460 self.print(left, left_size)?;
462 self.left_total += len;
464 if self.left == self.right {
469 self.left %= self.buf_max_len;
471 left_size = self.buf[self.left].size;
477 pub fn check_stack(&mut self, k: isize) {
478 if !self.scan_stack.is_empty() {
479 let x = self.scan_top();
480 match self.buf[x].token {
483 let popped = self.scan_pop();
484 self.buf[popped].size = self.buf[x].size + self.right_total;
485 self.check_stack(k - 1);
489 // paper says + not =, but that makes no sense.
490 let popped = self.scan_pop();
491 self.buf[popped].size = 1;
492 self.check_stack(k + 1);
495 let popped = self.scan_pop();
496 self.buf[popped].size = self.buf[x].size + self.right_total;
505 pub fn print_newline(&mut self, amount: isize) -> io::Result<()> {
506 debug!("NEWLINE {}", amount);
507 let ret = write!(self.out, "\n");
508 self.pending_indentation = 0;
513 pub fn indent(&mut self, amount: isize) {
514 debug!("INDENT {}", amount);
515 self.pending_indentation += amount;
518 pub fn get_top(&mut self) -> PrintStackElem {
519 match self.print_stack.last() {
521 None => PrintStackElem {
523 pbreak: PrintStackBreak::Broken(Breaks::Inconsistent)
528 pub fn print_begin(&mut self, b: BeginToken, l: isize) -> io::Result<()> {
530 let col = self.margin - self.space + b.offset;
531 debug!("print Begin -> push broken block at col {}", col);
532 self.print_stack.push(PrintStackElem {
534 pbreak: PrintStackBreak::Broken(b.breaks)
537 debug!("print Begin -> push fitting block");
538 self.print_stack.push(PrintStackElem {
540 pbreak: PrintStackBreak::Fits
546 pub fn print_end(&mut self) -> io::Result<()> {
547 debug!("print End -> pop End");
548 let print_stack = &mut self.print_stack;
549 assert!(!print_stack.is_empty());
550 print_stack.pop().unwrap();
554 pub fn print_break(&mut self, b: BreakToken, l: isize) -> io::Result<()> {
555 let top = self.get_top();
557 PrintStackBreak::Fits => {
558 debug!("print Break({}) in fitting block", b.blank_space);
559 self.space -= b.blank_space;
560 self.indent(b.blank_space);
563 PrintStackBreak::Broken(Breaks::Consistent) => {
564 debug!("print Break({}+{}) in consistent block",
565 top.offset, b.offset);
566 let ret = self.print_newline(top.offset + b.offset);
567 self.space = self.margin - (top.offset + b.offset);
570 PrintStackBreak::Broken(Breaks::Inconsistent) => {
572 debug!("print Break({}+{}) w/ newline in inconsistent",
573 top.offset, b.offset);
574 let ret = self.print_newline(top.offset + b.offset);
575 self.space = self.margin - (top.offset + b.offset);
578 debug!("print Break({}) w/o newline in inconsistent",
580 self.indent(b.blank_space);
581 self.space -= b.blank_space;
588 pub fn print_string(&mut self, s: Cow<'static, str>, len: isize) -> io::Result<()> {
589 debug!("print String({})", s);
590 // assert!(len <= space);
592 while self.pending_indentation > 0 {
593 write!(self.out, " ")?;
594 self.pending_indentation -= 1;
596 write!(self.out, "{}", s)
599 pub fn print(&mut self, token: Token, l: isize) -> io::Result<()> {
600 debug!("print {} {} (remaining line space={})", token, l,
602 debug!("{}", buf_str(&self.buf,
607 Token::Begin(b) => self.print_begin(b, l),
608 Token::End => self.print_end(),
609 Token::Break(b) => self.print_break(b, l),
610 Token::String(s, len) => {
612 self.print_string(s, len)
614 Token::Eof => panic!(), // Eof should never get here.
618 // Convenience functions to talk to the printer.
621 pub fn rbox(&mut self, indent: usize, b: Breaks) -> io::Result<()> {
622 self.pretty_print_begin(BeginToken {
623 offset: indent as isize,
628 /// Inconsistent breaking box
629 pub fn ibox(&mut self, indent: usize) -> io::Result<()> {
630 self.rbox(indent, Breaks::Inconsistent)
633 /// Consistent breaking box
634 pub fn cbox(&mut self, indent: usize) -> io::Result<()> {
635 self.rbox(indent, Breaks::Consistent)
638 pub fn break_offset(&mut self, n: usize, off: isize) -> io::Result<()> {
639 self.pretty_print_break(BreakToken {
641 blank_space: n as isize
645 pub fn end(&mut self) -> io::Result<()> {
646 self.pretty_print_end()
649 pub fn eof(&mut self) -> io::Result<()> {
650 self.pretty_print_eof()
653 pub fn word<S: Into<Cow<'static, str>>>(&mut self, wrd: S) -> io::Result<()> {
655 let len = s.len() as isize;
656 self.pretty_print_string(s, len)
659 fn spaces(&mut self, n: usize) -> io::Result<()> {
660 self.break_offset(n, 0)
663 pub fn zerobreak(&mut self) -> io::Result<()> {
667 pub fn space(&mut self) -> io::Result<()> {
671 pub fn hardbreak(&mut self) -> io::Result<()> {
672 self.spaces(SIZE_INFINITY as usize)
675 pub fn hardbreak_tok_offset(off: isize) -> Token {
676 Token::Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
679 pub fn hardbreak_tok() -> Token {
680 Self::hardbreak_tok_offset(0)