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 uint 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.
62 pub use self::PrintStackBreak::*;
63 pub use self::Breaks::*;
64 pub use self::Token::*;
69 #[deriving(Clone, PartialEq)]
76 pub struct BreakToken {
82 pub struct BeginToken {
89 String(string::String, int),
97 pub fn is_eof(&self) -> bool {
98 match *self { Eof => true, _ => false }
101 pub fn is_hardbreak_tok(&self) -> bool {
106 }) if bs == SIZE_INFINITY =>
114 pub fn tok_str(t: Token) -> string::String {
116 String(s, len) => return format!("STR({},{})", s, len),
117 Break(_) => return "BREAK".to_string(),
118 Begin(_) => return "BEGIN".to_string(),
119 End => return "END".to_string(),
120 Eof => return "EOF".to_string()
124 pub fn buf_str(toks: Vec<Token>,
131 assert_eq!(n, szs.len());
134 let mut s = string::String::from_str("[");
135 while i != right && l != 0u {
140 s.push_str(format!("{}={}",
142 tok_str(toks[i].clone())).as_slice());
147 return s.into_string();
150 pub enum PrintStackBreak {
155 pub struct PrintStackElem {
157 pbreak: PrintStackBreak
160 static SIZE_INFINITY: int = 0xffff;
162 pub fn mk_printer(out: Box<io::Writer+'static>, linewidth: uint) -> Printer {
163 // Yes 3, it makes the ring buffers big enough to never
165 let n: uint = 3 * linewidth;
166 debug!("mk_printer {}", linewidth);
167 let token: Vec<Token> = Vec::from_elem(n, Eof);
168 let size: Vec<int> = Vec::from_elem(n, 0i);
169 let scan_stack: Vec<uint> = Vec::from_elem(n, 0u);
173 margin: linewidth as int,
174 space: linewidth as int,
181 scan_stack: scan_stack,
182 scan_stack_empty: true,
185 print_stack: Vec::new(),
186 pending_indentation: 0
191 /// In case you do not have the paper, here is an explanation of what's going
194 /// There is a stream of input tokens flowing through this printer.
196 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
197 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
198 /// case every token worth buffering is 1 char long, so it's ok.
200 /// Tokens are String, Break, and Begin/End to delimit blocks.
202 /// Begin tokens can carry an offset, saying "how far to indent when you break
203 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
204 /// breaking. Consistent breaking means that after the first break, no attempt
205 /// will be made to flow subsequent breaks together onto lines. Inconsistent
206 /// is the opposite. Inconsistent breaking example would be, say:
208 /// foo(hello, there, good, friends)
210 /// breaking inconsistently to become
215 /// whereas a consistent breaking would yield:
222 /// That is, in the consistent-break blocks we value vertical alignment
223 /// more than the ability to cram stuff onto a line. But in all cases if it
224 /// can make a block a one-liner, it'll do so.
226 /// Carrying on with high-level logic:
228 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
229 /// 'right' indices denote the active portion of the ring buffer as well as
230 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
231 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
232 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
233 /// and point-in-infinite-stream senses freely.
235 /// There is a parallel ring buffer, 'size', that holds the calculated size of
236 /// each token. Why calculated? Because for Begin/End pairs, the "size"
237 /// includes everything between the pair. That is, the "size" of Begin is
238 /// actually the sum of the sizes of everything between Begin and the paired
239 /// End that follows. Since that is arbitrarily far in the future, 'size' is
240 /// being rewritten regularly while the printer runs; in fact most of the
241 /// machinery is here to work out 'size' entries on the fly (and give up when
242 /// they're so obviously over-long that "infinity" is a good enough
243 /// approximation for purposes of line breaking).
245 /// The "input side" of the printer is managed as an abstract process called
246 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
247 /// manage calculating 'size'. SCAN is, in other words, the process of
248 /// calculating 'size' entries.
250 /// The "output side" of the printer is managed by an abstract process called
251 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
252 /// do with each token/size pair it consumes as it goes. It's trying to consume
253 /// the entire buffered window, but can't output anything until the size is >=
254 /// 0 (sizes are set to negative while they're pending calculation).
256 /// So SCAN takes input and buffers tokens and pending calculations, while
257 /// PRINT gobbles up completed calculations and tokens from the buffer. The
258 /// theory is that the two can never get more than 3N tokens apart, because
259 /// once there's "obviously" too much data to fit on a line, in a size
260 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
263 /// In this implementation (following the paper, again) the SCAN process is
264 /// the method called 'pretty_print', and the 'PRINT' process is the method
267 pub out: Box<io::Writer+'static>,
269 /// Width of lines we're constrained to
271 /// Number of spaces left on line
273 /// Index of left side of input stream
275 /// Index of right side of input stream
277 /// Ring-buffer stream goes through
279 /// Ring-buffer of calculated sizes
281 /// Running size of stream "...left"
283 /// Running size of stream "...right"
285 /// Pseudo-stack, really a ring too. Holds the
286 /// primary-ring-buffers index of the Begin that started the
287 /// current block, possibly with the most recent Break after that
288 /// Begin (if there is any) on top of it. Stuff is flushed off the
289 /// bottom as it becomes irrelevant due to the primary ring-buffer
291 scan_stack: Vec<uint> ,
292 /// Top==bottom disambiguator
293 scan_stack_empty: bool,
294 /// Index of top of scan_stack
296 /// Index of bottom of scan_stack
298 /// Stack of blocks-in-progress being flushed by print
299 print_stack: Vec<PrintStackElem> ,
300 /// Buffered indentation to avoid writing trailing whitespace
301 pending_indentation: int,
305 pub fn last_token(&mut self) -> Token {
306 self.token[self.right].clone()
308 // be very careful with this!
309 pub fn replace_last_token(&mut self, t: Token) {
310 self.token[self.right] = t;
312 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
313 debug!("pp ~[{},{}]", self.left, self.right);
316 if !self.scan_stack_empty {
318 let left = self.token[self.left].clone();
319 let left_size = self.size[self.left];
320 try!(self.advance_left(left, left_size));
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] = t;
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] = t;
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] = t;
367 self.size[self.right] = -self.right_total;
368 self.right_total += b.blank_space;
371 String(ref s, len) => {
372 if self.scan_stack_empty {
373 debug!("pp String('{}')/print ~[{},{}]",
374 *s, self.left, self.right);
375 self.print(t.clone(), len)
377 debug!("pp String('{}')/buffer ~[{},{}]",
378 *s, self.left, self.right);
379 self.advance_right();
380 self.token[self.right] = t.clone();
381 self.size[self.right] = len;
382 self.right_total += len;
388 pub fn check_stream(&mut self) -> 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 let left = self.token[self.left].clone();
402 let left_size = self.size[self.left];
403 try!(self.advance_left(left, left_size));
404 if self.left != self.right {
405 try!(self.check_stream());
410 pub fn scan_push(&mut self, x: uint) {
411 debug!("scan_push {}", x);
412 if self.scan_stack_empty {
413 self.scan_stack_empty = false;
416 self.top %= self.buf_len;
417 assert!((self.top != self.bottom));
419 self.scan_stack[self.top] = x;
421 pub fn scan_pop(&mut self) -> uint {
422 assert!((!self.scan_stack_empty));
423 let x = self.scan_stack[self.top];
424 if self.top == self.bottom {
425 self.scan_stack_empty = true;
427 self.top += self.buf_len - 1u; self.top %= self.buf_len;
431 pub fn scan_top(&mut self) -> uint {
432 assert!((!self.scan_stack_empty));
433 return self.scan_stack[self.top];
435 pub fn scan_pop_bottom(&mut self) -> uint {
436 assert!((!self.scan_stack_empty));
437 let x = self.scan_stack[self.bottom];
438 if self.top == self.bottom {
439 self.scan_stack_empty = true;
441 self.bottom += 1u; self.bottom %= self.buf_len;
445 pub fn advance_right(&mut self) {
447 self.right %= self.buf_len;
448 assert!((self.right != self.left));
450 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
451 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
454 let ret = self.print(x.clone(), l);
456 Break(b) => self.left_total += b.blank_space,
458 assert_eq!(len, l); self.left_total += len;
462 if self.left != self.right {
464 self.left %= self.buf_len;
465 let left = self.token[self.left].clone();
466 let left_size = self.size[self.left];
467 try!(self.advance_left(left, left_size));
474 pub fn check_stack(&mut self, k: int) {
475 if !self.scan_stack_empty {
476 let x = self.scan_top();
477 match self.token[x] {
480 let popped = self.scan_pop();
481 self.size[popped] = self.size[x] + self.right_total;
482 self.check_stack(k - 1);
486 // paper says + not =, but that makes no sense.
487 let popped = self.scan_pop();
488 self.size[popped] = 1;
489 self.check_stack(k + 1);
492 let popped = self.scan_pop();
493 self.size[popped] = self.size[x] + self.right_total;
501 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
502 debug!("NEWLINE {}", amount);
503 let ret = write!(self.out, "\n");
504 self.pending_indentation = 0;
508 pub fn indent(&mut self, amount: int) {
509 debug!("INDENT {}", amount);
510 self.pending_indentation += amount;
512 pub fn get_top(&mut self) -> PrintStackElem {
513 let print_stack = &mut self.print_stack;
514 let n = print_stack.len();
516 (*print_stack)[n - 1]
520 pbreak: Broken(Inconsistent)
524 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
525 while self.pending_indentation > 0 {
526 try!(write!(self.out, " "));
527 self.pending_indentation -= 1;
529 write!(self.out, "{}", s)
531 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
532 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
534 debug!("{}", buf_str(self.token.clone(),
542 let col = self.margin - self.space + b.offset;
543 debug!("print Begin -> push broken block at col {}", col);
544 self.print_stack.push(PrintStackElem {
546 pbreak: Broken(b.breaks)
549 debug!("print Begin -> push fitting block");
550 self.print_stack.push(PrintStackElem {
558 debug!("print End -> pop End");
559 let print_stack = &mut self.print_stack;
560 assert!((print_stack.len() != 0u));
561 print_stack.pop().unwrap();
565 let top = self.get_top();
568 debug!("print Break({}) in fitting block", b.blank_space);
569 self.space -= b.blank_space;
570 self.indent(b.blank_space);
573 Broken(Consistent) => {
574 debug!("print Break({}+{}) in consistent block",
575 top.offset, b.offset);
576 let ret = self.print_newline(top.offset + b.offset);
577 self.space = self.margin - (top.offset + b.offset);
580 Broken(Inconsistent) => {
582 debug!("print Break({}+{}) w/ newline in inconsistent",
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 debug!("print Break({}) w/o newline in inconsistent",
590 self.indent(b.blank_space);
591 self.space -= b.blank_space;
598 debug!("print String({})", s);
600 // assert!(l <= space);
602 self.print_str(s.as_slice())
605 // Eof should never get here.
612 // Convenience functions to talk to the printer.
615 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
616 p.pretty_print(Begin(BeginToken {
617 offset: indent as int,
622 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
623 rbox(p, indent, Inconsistent)
626 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
627 rbox(p, indent, Consistent)
630 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
631 p.pretty_print(Break(BreakToken {
633 blank_space: n as int
637 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
639 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
641 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
642 p.pretty_print(String(/* bad */ wrd.to_string(), wrd.len() as int))
645 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
646 p.pretty_print(String(/* bad */ wrd.to_string(), SIZE_INFINITY))
649 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
650 p.pretty_print(String(/* bad */ wrd.to_string(), 0))
653 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
654 break_offset(p, n, 0)
657 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
661 pub fn space(p: &mut Printer) -> io::IoResult<()> {
665 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
666 spaces(p, SIZE_INFINITY as uint)
669 pub fn hardbreak_tok_offset(off: int) -> Token {
670 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
673 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }