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.
67 #[derive(Clone, Copy, PartialEq)]
73 #[derive(Clone, Copy)]
74 pub struct BreakToken {
79 #[derive(Clone, Copy)]
80 pub struct BeginToken {
87 String(String, isize),
95 pub fn is_eof(&self) -> bool {
102 pub fn is_hardbreak_tok(&self) -> bool {
104 Token::Break(BreakToken {
107 }) if bs == SIZE_INFINITY =>
115 pub fn tok_str(token: &Token) -> String {
117 Token::String(ref s, len) => format!("STR({},{})", s, len),
118 Token::Break(_) => "BREAK".to_string(),
119 Token::Begin(_) => "BEGIN".to_string(),
120 Token::End => "END".to_string(),
121 Token::Eof => "EOF".to_string()
125 pub fn buf_str(toks: &[Token],
132 assert_eq!(n, szs.len());
135 let mut s = string::String::from("[");
136 while i != right && l != 0 {
141 s.push_str(&format!("{}={}",
151 #[derive(Copy, Clone)]
152 pub enum PrintStackBreak {
157 #[derive(Copy, Clone)]
158 pub struct PrintStackElem {
160 pbreak: PrintStackBreak
163 const SIZE_INFINITY: isize = 0xffff;
165 pub fn mk_printer<'a>(out: Box<io::Write+'a>, linewidth: usize) -> Printer<'a> {
166 // Yes 3, it makes the ring buffers big enough to never
168 let n: usize = 3 * linewidth;
169 debug!("mk_printer {}", linewidth);
170 let token = vec![Token::Eof; n];
171 let size = vec![0_isize; n];
172 let scan_stack = vec![0_usize; n];
176 margin: linewidth as isize,
177 space: linewidth as isize,
184 scan_stack: scan_stack,
185 scan_stack_empty: true,
188 print_stack: Vec::new(),
189 pending_indentation: 0
194 /// In case you do not have the paper, here is an explanation of what's going
197 /// There is a stream of input tokens flowing through this printer.
199 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
200 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
201 /// case every token worth buffering is 1 char long, so it's ok.
203 /// Tokens are String, Break, and Begin/End to delimit blocks.
205 /// Begin tokens can carry an offset, saying "how far to indent when you break
206 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
207 /// breaking. Consistent breaking means that after the first break, no attempt
208 /// will be made to flow subsequent breaks together onto lines. Inconsistent
209 /// is the opposite. Inconsistent breaking example would be, say:
211 /// foo(hello, there, good, friends)
213 /// breaking inconsistently to become
218 /// whereas a consistent breaking would yield:
225 /// That is, in the consistent-break blocks we value vertical alignment
226 /// more than the ability to cram stuff onto a line. But in all cases if it
227 /// can make a block a one-liner, it'll do so.
229 /// Carrying on with high-level logic:
231 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
232 /// 'right' indices denote the active portion of the ring buffer as well as
233 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
234 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
235 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
236 /// and point-in-infinite-stream senses freely.
238 /// There is a parallel ring buffer, 'size', that holds the calculated size of
239 /// each token. Why calculated? Because for Begin/End pairs, the "size"
240 /// includes everything between the pair. That is, the "size" of Begin is
241 /// actually the sum of the sizes of everything between Begin and the paired
242 /// End that follows. Since that is arbitrarily far in the future, 'size' is
243 /// being rewritten regularly while the printer runs; in fact most of the
244 /// machinery is here to work out 'size' entries on the fly (and give up when
245 /// they're so obviously over-long that "infinity" is a good enough
246 /// approximation for purposes of line breaking).
248 /// The "input side" of the printer is managed as an abstract process called
249 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
250 /// manage calculating 'size'. SCAN is, in other words, the process of
251 /// calculating 'size' entries.
253 /// The "output side" of the printer is managed by an abstract process called
254 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
255 /// do with each token/size pair it consumes as it goes. It's trying to consume
256 /// the entire buffered window, but can't output anything until the size is >=
257 /// 0 (sizes are set to negative while they're pending calculation).
259 /// So SCAN takes input and buffers tokens and pending calculations, while
260 /// PRINT gobbles up completed calculations and tokens from the buffer. The
261 /// theory is that the two can never get more than 3N tokens apart, because
262 /// once there's "obviously" too much data to fit on a line, in a size
263 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
266 /// In this implementation (following the paper, again) the SCAN process is
267 /// the method called 'pretty_print', and the 'PRINT' process is the method
269 pub struct Printer<'a> {
270 pub out: Box<io::Write+'a>,
272 /// Width of lines we're constrained to
274 /// Number of spaces left on line
276 /// Index of left side of input stream
278 /// Index of right side of input stream
280 /// Ring-buffer stream goes through
282 /// Ring-buffer of calculated sizes
284 /// Running size of stream "...left"
286 /// Running size of stream "...right"
288 /// Pseudo-stack, really a ring too. Holds the
289 /// primary-ring-buffers index of the Begin that started the
290 /// current block, possibly with the most recent Break after that
291 /// Begin (if there is any) on top of it. Stuff is flushed off the
292 /// bottom as it becomes irrelevant due to the primary ring-buffer
294 scan_stack: Vec<usize> ,
295 /// Top==bottom disambiguator
296 scan_stack_empty: bool,
297 /// Index of top of scan_stack
299 /// Index of bottom of scan_stack
301 /// Stack of blocks-in-progress being flushed by print
302 print_stack: Vec<PrintStackElem> ,
303 /// Buffered indentation to avoid writing trailing whitespace
304 pending_indentation: isize,
307 impl<'a> Printer<'a> {
308 pub fn last_token(&mut self) -> Token {
309 self.token[self.right].clone()
311 // be very careful with this!
312 pub fn replace_last_token(&mut self, t: Token) {
313 self.token[self.right] = t;
315 pub fn pretty_print(&mut self, token: Token) -> io::Result<()> {
316 debug!("pp Vec<{},{}>", self.left, self.right);
319 if !self.scan_stack_empty {
321 try!(self.advance_left());
327 if self.scan_stack_empty {
329 self.right_total = 1;
332 } else { self.advance_right(); }
333 debug!("pp Begin({})/buffer Vec<{},{}>",
334 b.offset, self.left, self.right);
335 self.token[self.right] = token;
336 self.size[self.right] = -self.right_total;
337 let right = self.right;
338 self.scan_push(right);
342 if self.scan_stack_empty {
343 debug!("pp End/print Vec<{},{}>", self.left, self.right);
346 debug!("pp End/buffer Vec<{},{}>", self.left, self.right);
347 self.advance_right();
348 self.token[self.right] = token;
349 self.size[self.right] = -1;
350 let right = self.right;
351 self.scan_push(right);
356 if self.scan_stack_empty {
358 self.right_total = 1;
361 } else { self.advance_right(); }
362 debug!("pp Break({})/buffer Vec<{},{}>",
363 b.offset, self.left, self.right);
365 let right = self.right;
366 self.scan_push(right);
367 self.token[self.right] = token;
368 self.size[self.right] = -self.right_total;
369 self.right_total += b.blank_space;
372 Token::String(s, len) => {
373 if self.scan_stack_empty {
374 debug!("pp String('{}')/print Vec<{},{}>",
375 s, self.left, self.right);
376 self.print(Token::String(s, len), len)
378 debug!("pp String('{}')/buffer Vec<{},{}>",
379 s, self.left, self.right);
380 self.advance_right();
381 self.token[self.right] = Token::String(s, len);
382 self.size[self.right] = len;
383 self.right_total += len;
389 pub fn check_stream(&mut self) -> io::Result<()> {
390 debug!("check_stream Vec<{}, {}> with left_total={}, right_total={}",
391 self.left, self.right, self.left_total, self.right_total);
392 if self.right_total - self.left_total > self.space {
393 debug!("scan window is {}, longer than space on line ({})",
394 self.right_total - self.left_total, self.space);
395 if !self.scan_stack_empty {
396 if self.left == self.scan_stack[self.bottom] {
397 debug!("setting {} to infinity and popping", self.left);
398 let scanned = self.scan_pop_bottom();
399 self.size[scanned] = SIZE_INFINITY;
402 try!(self.advance_left());
403 if self.left != self.right {
404 try!(self.check_stream());
409 pub fn scan_push(&mut self, x: usize) {
410 debug!("scan_push {}", x);
411 if self.scan_stack_empty {
412 self.scan_stack_empty = false;
415 self.top %= self.buf_len;
416 assert!((self.top != self.bottom));
418 self.scan_stack[self.top] = x;
420 pub fn scan_pop(&mut self) -> usize {
421 assert!((!self.scan_stack_empty));
422 let x = self.scan_stack[self.top];
423 if self.top == self.bottom {
424 self.scan_stack_empty = true;
426 self.top += self.buf_len - 1; self.top %= self.buf_len;
430 pub fn scan_top(&mut self) -> usize {
431 assert!((!self.scan_stack_empty));
432 return self.scan_stack[self.top];
434 pub fn scan_pop_bottom(&mut self) -> usize {
435 assert!((!self.scan_stack_empty));
436 let x = self.scan_stack[self.bottom];
437 if self.top == self.bottom {
438 self.scan_stack_empty = true;
440 self.bottom += 1; self.bottom %= self.buf_len;
444 pub fn advance_right(&mut self) {
446 self.right %= self.buf_len;
447 assert!((self.right != self.left));
449 pub fn advance_left(&mut self) -> io::Result<()> {
450 debug!("advance_left Vec<{},{}>, sizeof({})={}", self.left, self.right,
451 self.left, self.size[self.left]);
453 let mut left_size = self.size[self.left];
455 while left_size >= 0 {
456 let left = self.token[self.left].clone();
458 let len = match left {
459 Token::Break(b) => b.blank_space,
460 Token::String(_, len) => {
461 assert_eq!(len, left_size);
467 try!(self.print(left, left_size));
469 self.left_total += len;
471 if self.left == self.right {
476 self.left %= self.buf_len;
478 left_size = self.size[self.left];
483 pub fn check_stack(&mut self, k: isize) {
484 if !self.scan_stack_empty {
485 let x = self.scan_top();
486 match self.token[x] {
489 let popped = self.scan_pop();
490 self.size[popped] = self.size[x] + self.right_total;
491 self.check_stack(k - 1);
495 // paper says + not =, but that makes no sense.
496 let popped = self.scan_pop();
497 self.size[popped] = 1;
498 self.check_stack(k + 1);
501 let popped = self.scan_pop();
502 self.size[popped] = self.size[x] + self.right_total;
510 pub fn print_newline(&mut self, amount: isize) -> io::Result<()> {
511 debug!("NEWLINE {}", amount);
512 let ret = write!(self.out, "\n");
513 self.pending_indentation = 0;
517 pub fn indent(&mut self, amount: isize) {
518 debug!("INDENT {}", amount);
519 self.pending_indentation += amount;
521 pub fn get_top(&mut self) -> PrintStackElem {
522 let print_stack = &mut self.print_stack;
523 let n = print_stack.len();
525 (*print_stack)[n - 1]
529 pbreak: PrintStackBreak::Broken(Breaks::Inconsistent)
533 pub fn print_str(&mut self, s: &str) -> io::Result<()> {
534 while self.pending_indentation > 0 {
535 try!(write!(self.out, " "));
536 self.pending_indentation -= 1;
538 write!(self.out, "{}", s)
540 pub fn print(&mut self, token: Token, l: isize) -> io::Result<()> {
541 debug!("print {} {} (remaining line space={})", tok_str(&token), l,
543 debug!("{}", buf_str(&self.token,
551 let col = self.margin - self.space + b.offset;
552 debug!("print Begin -> push broken block at col {}", col);
553 self.print_stack.push(PrintStackElem {
555 pbreak: PrintStackBreak::Broken(b.breaks)
558 debug!("print Begin -> push fitting block");
559 self.print_stack.push(PrintStackElem {
561 pbreak: PrintStackBreak::Fits
567 debug!("print End -> pop End");
568 let print_stack = &mut self.print_stack;
569 assert!((!print_stack.is_empty()));
570 print_stack.pop().unwrap();
574 let top = self.get_top();
576 PrintStackBreak::Fits => {
577 debug!("print Break({}) in fitting block", b.blank_space);
578 self.space -= b.blank_space;
579 self.indent(b.blank_space);
582 PrintStackBreak::Broken(Breaks::Consistent) => {
583 debug!("print Break({}+{}) in consistent block",
584 top.offset, b.offset);
585 let ret = self.print_newline(top.offset + b.offset);
586 self.space = self.margin - (top.offset + b.offset);
589 PrintStackBreak::Broken(Breaks::Inconsistent) => {
591 debug!("print Break({}+{}) w/ newline in inconsistent",
592 top.offset, b.offset);
593 let ret = self.print_newline(top.offset + b.offset);
594 self.space = self.margin - (top.offset + b.offset);
597 debug!("print Break({}) w/o newline in inconsistent",
599 self.indent(b.blank_space);
600 self.space -= b.blank_space;
606 Token::String(s, len) => {
607 debug!("print String({})", s);
609 // assert!(l <= space);
611 self.print_str(&s[..])
614 // Eof should never get here.
621 // Convenience functions to talk to the printer.
624 pub fn rbox(p: &mut Printer, indent: usize, b: Breaks) -> io::Result<()> {
625 p.pretty_print(Token::Begin(BeginToken {
626 offset: indent as isize,
631 pub fn ibox(p: &mut Printer, indent: usize) -> io::Result<()> {
632 rbox(p, indent, Breaks::Inconsistent)
635 pub fn cbox(p: &mut Printer, indent: usize) -> io::Result<()> {
636 rbox(p, indent, Breaks::Consistent)
639 pub fn break_offset(p: &mut Printer, n: usize, off: isize) -> io::Result<()> {
640 p.pretty_print(Token::Break(BreakToken {
642 blank_space: n as isize
646 pub fn end(p: &mut Printer) -> io::Result<()> {
647 p.pretty_print(Token::End)
650 pub fn eof(p: &mut Printer) -> io::Result<()> {
651 p.pretty_print(Token::Eof)
654 pub fn word(p: &mut Printer, wrd: &str) -> io::Result<()> {
655 p.pretty_print(Token::String(/* bad */ wrd.to_string(), wrd.len() as isize))
658 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::Result<()> {
659 p.pretty_print(Token::String(/* bad */ wrd.to_string(), SIZE_INFINITY))
662 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::Result<()> {
663 p.pretty_print(Token::String(/* bad */ wrd.to_string(), 0))
666 pub fn spaces(p: &mut Printer, n: usize) -> io::Result<()> {
667 break_offset(p, n, 0)
670 pub fn zerobreak(p: &mut Printer) -> io::Result<()> {
674 pub fn space(p: &mut Printer) -> io::Result<()> {
678 pub fn hardbreak(p: &mut Printer) -> io::Result<()> {
679 spaces(p, SIZE_INFINITY as usize)
682 pub fn hardbreak_tok_offset(off: isize) -> Token {
683 Token::Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
686 pub fn hardbreak_tok() -> Token {
687 hardbreak_tok_offset(0)