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
14 //! STAN-CS-79-770: "Pretty Printing", by Derek C. Oppen.
15 //! Stanford Department of Computer Science, 1979.
17 //! The algorithm's aim is to break a stream into as few lines as possible
18 //! while respecting the indentation-consistency requirements of the enclosing
19 //! block, and avoiding breaking at silly places on block boundaries, for
20 //! example, between "x" and ")" in "x)".
22 //! I am implementing this algorithm because it comes with 20 pages of
23 //! documentation explaining its theory, and because it addresses the set of
24 //! concerns I've seen other pretty-printers fall down on. Weirdly. Even though
25 //! it's 32 years old. What can I say?
27 //! Despite some redundancies and quirks in the way it's implemented in that
28 //! paper, I've opted to keep the implementation here as similar as I can,
29 //! changing only what was blatantly wrong, a typo, or sufficiently
30 //! non-idiomatic rust that it really stuck out.
32 //! In particular you'll see a certain amount of churn related to INTEGER vs.
33 //! CARDINAL in the Mesa implementation. Mesa apparently interconverts the two
34 //! somewhat readily? In any case, I've used usize for indices-in-buffers and
35 //! ints for character-sizes-and-indentation-offsets. This respects the need
36 //! for ints to "go negative" while carrying a pending-calculation balance, and
37 //! helps differentiate all the numbers flying around internally (slightly).
39 //! I also inverted the indentation arithmetic used in the print stack, since
40 //! the Mesa implementation (somewhat randomly) stores the offset on the print
41 //! stack in terms of margin-col rather than col itself. I store col.
43 //! I also implemented a small change in the String token, in that I store an
44 //! explicit length for the string. For most tokens this is just the length of
45 //! the accompanying string. But it's necessary to permit it to differ, for
46 //! encoding things that are supposed to "go on their own line" -- certain
47 //! classes of comment and blank-line -- where relying on adjacent
48 //! hardbreak-like Break tokens with long blankness indication doesn't actually
49 //! work. To see why, consider when there is a "thing that should be on its own
50 //! line" between two long blocks, say functions. If you put a hardbreak after
51 //! each function (or before each) and the breaking algorithm decides to break
52 //! there anyways (because the functions themselves are long) you wind up with
53 //! extra blank lines. If you don't put hardbreaks you can wind up with the
54 //! "thing which should be on its own line" not getting its own line in the
55 //! rare case of "really small functions" or such. This re-occurs with comments
56 //! and explicit blank lines. So in those cases we use a string with a payload
57 //! we want isolated to a line and an explicit length that's huge, surrounded
58 //! by two zero-length breaks. The algorithm will try its best to fit it on a
59 //! line (which it can't) and so naturally place the content on its own line to
60 //! avoid combining it with other lines and making matters even worse.
64 use std::iter::repeat;
66 #[derive(Clone, Copy, PartialEq)]
72 #[derive(Clone, Copy)]
73 pub struct BreakToken {
78 #[derive(Clone, Copy)]
79 pub struct BeginToken {
86 String(String, isize),
94 pub fn is_eof(&self) -> bool {
101 pub fn is_hardbreak_tok(&self) -> bool {
103 Token::Break(BreakToken {
106 }) if bs == SIZE_INFINITY =>
114 pub fn tok_str(token: &Token) -> String {
116 Token::String(ref s, len) => format!("STR({},{})", s, len),
117 Token::Break(_) => "BREAK".to_string(),
118 Token::Begin(_) => "BEGIN".to_string(),
119 Token::End => "END".to_string(),
120 Token::Eof => "EOF".to_string()
124 pub fn buf_str(toks: &[Token],
131 assert_eq!(n, szs.len());
134 let mut s = string::String::from_str("[");
135 while i != right && l != 0 {
140 s.push_str(&format!("{}={}",
142 tok_str(&toks[i]))[]);
151 pub enum PrintStackBreak {
157 pub struct PrintStackElem {
159 pbreak: PrintStackBreak
162 static SIZE_INFINITY: isize = 0xffff;
164 pub fn mk_printer(out: Box<old_io::Writer+'static>, linewidth: usize) -> Printer {
165 // Yes 3, it makes the ring buffers big enough to never
167 let n: usize = 3 * linewidth;
168 debug!("mk_printer {}", linewidth);
169 let token: Vec<Token> = repeat(Token::Eof).take(n).collect();
170 let size: Vec<isize> = repeat(0).take(n).collect();
171 let scan_stack: Vec<usize> = repeat(0).take(n).collect();
175 margin: linewidth as isize,
176 space: linewidth as isize,
183 scan_stack: scan_stack,
184 scan_stack_empty: true,
187 print_stack: Vec::new(),
188 pending_indentation: 0
193 /// In case you do not have the paper, here is an explanation of what's going
196 /// There is a stream of input tokens flowing through this printer.
198 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
199 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
200 /// case every token worth buffering is 1 char long, so it's ok.
202 /// Tokens are String, Break, and Begin/End to delimit blocks.
204 /// Begin tokens can carry an offset, saying "how far to indent when you break
205 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
206 /// breaking. Consistent breaking means that after the first break, no attempt
207 /// will be made to flow subsequent breaks together onto lines. Inconsistent
208 /// is the opposite. Inconsistent breaking example would be, say:
210 /// foo(hello, there, good, friends)
212 /// breaking inconsistently to become
217 /// whereas a consistent breaking would yield:
224 /// That is, in the consistent-break blocks we value vertical alignment
225 /// more than the ability to cram stuff onto a line. But in all cases if it
226 /// can make a block a one-liner, it'll do so.
228 /// Carrying on with high-level logic:
230 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
231 /// 'right' indices denote the active portion of the ring buffer as well as
232 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
233 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
234 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
235 /// and point-in-infinite-stream senses freely.
237 /// There is a parallel ring buffer, 'size', that holds the calculated size of
238 /// each token. Why calculated? Because for Begin/End pairs, the "size"
239 /// includes everything between the pair. That is, the "size" of Begin is
240 /// actually the sum of the sizes of everything between Begin and the paired
241 /// End that follows. Since that is arbitrarily far in the future, 'size' is
242 /// being rewritten regularly while the printer runs; in fact most of the
243 /// machinery is here to work out 'size' entries on the fly (and give up when
244 /// they're so obviously over-long that "infinity" is a good enough
245 /// approximation for purposes of line breaking).
247 /// The "input side" of the printer is managed as an abstract process called
248 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
249 /// manage calculating 'size'. SCAN is, in other words, the process of
250 /// calculating 'size' entries.
252 /// The "output side" of the printer is managed by an abstract process called
253 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
254 /// do with each token/size pair it consumes as it goes. It's trying to consume
255 /// the entire buffered window, but can't output anything until the size is >=
256 /// 0 (sizes are set to negative while they're pending calculation).
258 /// So SCAN takes input and buffers tokens and pending calculations, while
259 /// PRINT gobbles up completed calculations and tokens from the buffer. The
260 /// theory is that the two can never get more than 3N tokens apart, because
261 /// once there's "obviously" too much data to fit on a line, in a size
262 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
265 /// In this implementation (following the paper, again) the SCAN process is
266 /// the method called 'pretty_print', and the 'PRINT' process is the method
269 pub out: Box<old_io::Writer+'static>,
271 /// Width of lines we're constrained to
273 /// Number of spaces left on line
275 /// Index of left side of input stream
277 /// Index of right side of input stream
279 /// Ring-buffer stream goes through
281 /// Ring-buffer of 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: Vec<usize> ,
294 /// Top==bottom disambiguator
295 scan_stack_empty: bool,
296 /// Index of top of scan_stack
298 /// Index of bottom of scan_stack
300 /// Stack of blocks-in-progress being flushed by print
301 print_stack: Vec<PrintStackElem> ,
302 /// Buffered indentation to avoid writing trailing whitespace
303 pending_indentation: isize,
307 pub fn last_token(&mut self) -> Token {
308 self.token[self.right].clone()
310 // be very careful with this!
311 pub fn replace_last_token(&mut self, t: Token) {
312 self.token[self.right] = t;
314 pub fn pretty_print(&mut self, token: Token) -> old_io::IoResult<()> {
315 debug!("pp ~[{},{}]", self.left, self.right);
318 if !self.scan_stack_empty {
320 try!(self.advance_left());
326 if self.scan_stack_empty {
328 self.right_total = 1;
331 } else { self.advance_right(); }
332 debug!("pp Begin({})/buffer ~[{},{}]",
333 b.offset, self.left, self.right);
334 self.token[self.right] = token;
335 self.size[self.right] = -self.right_total;
336 let right = self.right;
337 self.scan_push(right);
341 if self.scan_stack_empty {
342 debug!("pp End/print ~[{},{}]", self.left, self.right);
345 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
346 self.advance_right();
347 self.token[self.right] = token;
348 self.size[self.right] = -1;
349 let right = self.right;
350 self.scan_push(right);
355 if self.scan_stack_empty {
357 self.right_total = 1;
360 } else { self.advance_right(); }
361 debug!("pp Break({})/buffer ~[{},{}]",
362 b.offset, self.left, self.right);
364 let right = self.right;
365 self.scan_push(right);
366 self.token[self.right] = token;
367 self.size[self.right] = -self.right_total;
368 self.right_total += b.blank_space;
371 Token::String(s, len) => {
372 if self.scan_stack_empty {
373 debug!("pp String('{}')/print ~[{},{}]",
374 s, self.left, self.right);
375 self.print(Token::String(s, len), len)
377 debug!("pp String('{}')/buffer ~[{},{}]",
378 s, self.left, self.right);
379 self.advance_right();
380 self.token[self.right] = Token::String(s, len);
381 self.size[self.right] = len;
382 self.right_total += len;
388 pub fn check_stream(&mut self) -> old_io::IoResult<()> {
389 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
390 self.left, self.right, self.left_total, self.right_total);
391 if self.right_total - self.left_total > self.space {
392 debug!("scan window is {}, longer than space on line ({})",
393 self.right_total - self.left_total, self.space);
394 if !self.scan_stack_empty {
395 if self.left == self.scan_stack[self.bottom] {
396 debug!("setting {} to infinity and popping", self.left);
397 let scanned = self.scan_pop_bottom();
398 self.size[scanned] = SIZE_INFINITY;
401 try!(self.advance_left());
402 if self.left != self.right {
403 try!(self.check_stream());
408 pub fn scan_push(&mut self, x: usize) {
409 debug!("scan_push {}", x);
410 if self.scan_stack_empty {
411 self.scan_stack_empty = false;
414 self.top %= self.buf_len;
415 assert!((self.top != self.bottom));
417 self.scan_stack[self.top] = x;
419 pub fn scan_pop(&mut self) -> usize {
420 assert!((!self.scan_stack_empty));
421 let x = self.scan_stack[self.top];
422 if self.top == self.bottom {
423 self.scan_stack_empty = true;
425 self.top += self.buf_len - 1; self.top %= self.buf_len;
429 pub fn scan_top(&mut self) -> usize {
430 assert!((!self.scan_stack_empty));
431 return self.scan_stack[self.top];
433 pub fn scan_pop_bottom(&mut self) -> usize {
434 assert!((!self.scan_stack_empty));
435 let x = self.scan_stack[self.bottom];
436 if self.top == self.bottom {
437 self.scan_stack_empty = true;
439 self.bottom += 1; self.bottom %= self.buf_len;
443 pub fn advance_right(&mut self) {
445 self.right %= self.buf_len;
446 assert!((self.right != self.left));
448 pub fn advance_left(&mut self) -> old_io::IoResult<()> {
449 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
450 self.left, self.size[self.left]);
452 let mut left_size = self.size[self.left];
454 while left_size >= 0 {
455 let left = self.token[self.left].clone();
457 let len = match left {
458 Token::Break(b) => b.blank_space,
459 Token::String(_, len) => {
460 assert_eq!(len, left_size);
466 try!(self.print(left, left_size));
468 self.left_total += len;
470 if self.left == self.right {
475 self.left %= self.buf_len;
477 left_size = self.size[self.left];
482 pub fn check_stack(&mut self, k: isize) {
483 if !self.scan_stack_empty {
484 let x = self.scan_top();
485 match self.token[x] {
488 let popped = self.scan_pop();
489 self.size[popped] = self.size[x] + self.right_total;
490 self.check_stack(k - 1);
494 // paper says + not =, but that makes no sense.
495 let popped = self.scan_pop();
496 self.size[popped] = 1;
497 self.check_stack(k + 1);
500 let popped = self.scan_pop();
501 self.size[popped] = self.size[x] + self.right_total;
509 pub fn print_newline(&mut self, amount: isize) -> old_io::IoResult<()> {
510 debug!("NEWLINE {}", amount);
511 let ret = write!(self.out, "\n");
512 self.pending_indentation = 0;
516 pub fn indent(&mut self, amount: isize) {
517 debug!("INDENT {}", amount);
518 self.pending_indentation += amount;
520 pub fn get_top(&mut self) -> PrintStackElem {
521 let print_stack = &mut self.print_stack;
522 let n = print_stack.len();
524 (*print_stack)[n - 1]
528 pbreak: PrintStackBreak::Broken(Breaks::Inconsistent)
532 pub fn print_str(&mut self, s: &str) -> old_io::IoResult<()> {
533 while self.pending_indentation > 0 {
534 try!(write!(self.out, " "));
535 self.pending_indentation -= 1;
537 write!(self.out, "{}", s)
539 pub fn print(&mut self, token: Token, l: isize) -> old_io::IoResult<()> {
540 debug!("print {} {} (remaining line space={})", tok_str(&token), l,
542 debug!("{}", buf_str(&self.token,
550 let col = self.margin - self.space + b.offset;
551 debug!("print Begin -> push broken block at col {}", col);
552 self.print_stack.push(PrintStackElem {
554 pbreak: PrintStackBreak::Broken(b.breaks)
557 debug!("print Begin -> push fitting block");
558 self.print_stack.push(PrintStackElem {
560 pbreak: PrintStackBreak::Fits
566 debug!("print End -> pop End");
567 let print_stack = &mut self.print_stack;
568 assert!((print_stack.len() != 0));
569 print_stack.pop().unwrap();
573 let top = self.get_top();
575 PrintStackBreak::Fits => {
576 debug!("print Break({}) in fitting block", b.blank_space);
577 self.space -= b.blank_space;
578 self.indent(b.blank_space);
581 PrintStackBreak::Broken(Breaks::Consistent) => {
582 debug!("print Break({}+{}) in consistent block",
583 top.offset, b.offset);
584 let ret = self.print_newline(top.offset + b.offset);
585 self.space = self.margin - (top.offset + b.offset);
588 PrintStackBreak::Broken(Breaks::Inconsistent) => {
590 debug!("print Break({}+{}) w/ newline in inconsistent",
591 top.offset, b.offset);
592 let ret = self.print_newline(top.offset + b.offset);
593 self.space = self.margin - (top.offset + b.offset);
596 debug!("print Break({}) w/o newline in inconsistent",
598 self.indent(b.blank_space);
599 self.space -= b.blank_space;
605 Token::String(s, len) => {
606 debug!("print String({})", s);
608 // assert!(l <= space);
610 self.print_str(&s[..])
613 // Eof should never get here.
620 // Convenience functions to talk to the printer.
623 pub fn rbox(p: &mut Printer, indent: usize, b: Breaks) -> old_io::IoResult<()> {
624 p.pretty_print(Token::Begin(BeginToken {
625 offset: indent as isize,
630 pub fn ibox(p: &mut Printer, indent: usize) -> old_io::IoResult<()> {
631 rbox(p, indent, Breaks::Inconsistent)
634 pub fn cbox(p: &mut Printer, indent: usize) -> old_io::IoResult<()> {
635 rbox(p, indent, Breaks::Consistent)
638 pub fn break_offset(p: &mut Printer, n: usize, off: isize) -> old_io::IoResult<()> {
639 p.pretty_print(Token::Break(BreakToken {
641 blank_space: n as isize
645 pub fn end(p: &mut Printer) -> old_io::IoResult<()> {
646 p.pretty_print(Token::End)
649 pub fn eof(p: &mut Printer) -> old_io::IoResult<()> {
650 p.pretty_print(Token::Eof)
653 pub fn word(p: &mut Printer, wrd: &str) -> old_io::IoResult<()> {
654 p.pretty_print(Token::String(/* bad */ wrd.to_string(), wrd.len() as isize))
657 pub fn huge_word(p: &mut Printer, wrd: &str) -> old_io::IoResult<()> {
658 p.pretty_print(Token::String(/* bad */ wrd.to_string(), SIZE_INFINITY))
661 pub fn zero_word(p: &mut Printer, wrd: &str) -> old_io::IoResult<()> {
662 p.pretty_print(Token::String(/* bad */ wrd.to_string(), 0))
665 pub fn spaces(p: &mut Printer, n: usize) -> old_io::IoResult<()> {
666 break_offset(p, n, 0)
669 pub fn zerobreak(p: &mut Printer) -> old_io::IoResult<()> {
673 pub fn space(p: &mut Printer) -> old_io::IoResult<()> {
677 pub fn hardbreak(p: &mut Printer) -> old_io::IoResult<()> {
678 spaces(p, SIZE_INFINITY as usize)
681 pub fn hardbreak_tok_offset(off: isize) -> Token {
682 Token::Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
685 pub fn hardbreak_tok() -> Token {
686 hardbreak_tok_offset(0)