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.
63 use std::string::String;
65 #[deriving(Clone, PartialEq)]
72 pub struct BreakToken {
78 pub struct BeginToken {
93 pub fn is_eof(&self) -> bool {
94 match *self { Eof => true, _ => false }
97 pub fn is_hardbreak_tok(&self) -> bool {
102 }) if bs == SIZE_INFINITY =>
110 pub fn tok_str(t: Token) -> String {
112 String(s, len) => return format!("STR({},{})", s, len),
113 Break(_) => return "BREAK".to_string(),
114 Begin(_) => return "BEGIN".to_string(),
115 End => return "END".to_string(),
116 Eof => return "EOF".to_string()
120 pub fn buf_str(toks: Vec<Token>,
127 assert_eq!(n, szs.len());
130 let mut s = String::from_str("[");
131 while i != right && l != 0u {
136 s.push_str(format!("{}={}",
138 tok_str(toks.get(i).clone())).as_slice());
143 return s.into_string();
146 pub enum PrintStackBreak {
151 pub struct PrintStackElem {
153 pbreak: PrintStackBreak
156 static SIZE_INFINITY: int = 0xffff;
158 pub fn mk_printer(out: Box<io::Writer+'static>, linewidth: uint) -> Printer {
159 // Yes 3, it makes the ring buffers big enough to never
161 let n: uint = 3 * linewidth;
162 debug!("mk_printer {}", linewidth);
163 let token: Vec<Token> = Vec::from_elem(n, Eof);
164 let size: Vec<int> = Vec::from_elem(n, 0i);
165 let scan_stack: Vec<uint> = Vec::from_elem(n, 0u);
169 margin: linewidth as int,
170 space: linewidth as int,
177 scan_stack: scan_stack,
178 scan_stack_empty: true,
181 print_stack: Vec::new(),
182 pending_indentation: 0
187 /// In case you do not have the paper, here is an explanation of what's going
190 /// There is a stream of input tokens flowing through this printer.
192 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
193 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
194 /// case every token worth buffering is 1 char long, so it's ok.
196 /// Tokens are String, Break, and Begin/End to delimit blocks.
198 /// Begin tokens can carry an offset, saying "how far to indent when you break
199 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
200 /// breaking. Consistent breaking means that after the first break, no attempt
201 /// will be made to flow subsequent breaks together onto lines. Inconsistent
202 /// is the opposite. Inconsistent breaking example would be, say:
204 /// foo(hello, there, good, friends)
206 /// breaking inconsistently to become
211 /// whereas a consistent breaking would yield:
218 /// That is, in the consistent-break blocks we value vertical alignment
219 /// more than the ability to cram stuff onto a line. But in all cases if it
220 /// can make a block a one-liner, it'll do so.
222 /// Carrying on with high-level logic:
224 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
225 /// 'right' indices denote the active portion of the ring buffer as well as
226 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
227 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
228 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
229 /// and point-in-infinite-stream senses freely.
231 /// There is a parallel ring buffer, 'size', that holds the calculated size of
232 /// each token. Why calculated? Because for Begin/End pairs, the "size"
233 /// includes everything between the pair. That is, the "size" of Begin is
234 /// actually the sum of the sizes of everything between Begin and the paired
235 /// End that follows. Since that is arbitrarily far in the future, 'size' is
236 /// being rewritten regularly while the printer runs; in fact most of the
237 /// machinery is here to work out 'size' entries on the fly (and give up when
238 /// they're so obviously over-long that "infinity" is a good enough
239 /// approximation for purposes of line breaking).
241 /// The "input side" of the printer is managed as an abstract process called
242 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
243 /// manage calculating 'size'. SCAN is, in other words, the process of
244 /// calculating 'size' entries.
246 /// The "output side" of the printer is managed by an abstract process called
247 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
248 /// do with each token/size pair it consumes as it goes. It's trying to consume
249 /// the entire buffered window, but can't output anything until the size is >=
250 /// 0 (sizes are set to negative while they're pending calculation).
252 /// So SCAN takes input and buffers tokens and pending calculations, while
253 /// PRINT gobbles up completed calculations and tokens from the buffer. The
254 /// theory is that the two can never get more than 3N tokens apart, because
255 /// once there's "obviously" too much data to fit on a line, in a size
256 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
259 /// In this implementation (following the paper, again) the SCAN process is
260 /// the method called 'pretty_print', and the 'PRINT' process is the method
263 pub out: Box<io::Writer+'static>,
265 /// Width of lines we're constrained to
267 /// Number of spaces left on line
269 /// Index of left side of input stream
271 /// Index of right side of input stream
273 /// Ring-buffer stream goes through
275 /// Ring-buffer of 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: Vec<uint> ,
288 /// Top==bottom disambiguator
289 scan_stack_empty: bool,
290 /// Index of top of scan_stack
292 /// Index of bottom of scan_stack
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: int,
301 pub fn last_token(&mut self) -> Token {
302 (*self.token.get(self.right)).clone()
304 // be very careful with this!
305 pub fn replace_last_token(&mut self, t: Token) {
306 *self.token.get_mut(self.right) = t;
308 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
309 debug!("pp ~[{},{}]", self.left, self.right);
312 if !self.scan_stack_empty {
314 let left = (*self.token.get(self.left)).clone();
315 let left_size = *self.size.get(self.left);
316 try!(self.advance_left(left, left_size));
322 if self.scan_stack_empty {
324 self.right_total = 1;
327 } else { self.advance_right(); }
328 debug!("pp Begin({})/buffer ~[{},{}]",
329 b.offset, self.left, self.right);
330 *self.token.get_mut(self.right) = t;
331 *self.size.get_mut(self.right) = -self.right_total;
332 let right = self.right;
333 self.scan_push(right);
337 if self.scan_stack_empty {
338 debug!("pp End/print ~[{},{}]", self.left, self.right);
341 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
342 self.advance_right();
343 *self.token.get_mut(self.right) = t;
344 *self.size.get_mut(self.right) = -1;
345 let right = self.right;
346 self.scan_push(right);
351 if self.scan_stack_empty {
353 self.right_total = 1;
356 } else { self.advance_right(); }
357 debug!("pp Break({})/buffer ~[{},{}]",
358 b.offset, self.left, self.right);
360 let right = self.right;
361 self.scan_push(right);
362 *self.token.get_mut(self.right) = t;
363 *self.size.get_mut(self.right) = -self.right_total;
364 self.right_total += b.blank_space;
367 String(ref s, len) => {
368 if self.scan_stack_empty {
369 debug!("pp String('{}')/print ~[{},{}]",
370 *s, self.left, self.right);
371 self.print(t.clone(), len)
373 debug!("pp String('{}')/buffer ~[{},{}]",
374 *s, self.left, self.right);
375 self.advance_right();
376 *self.token.get_mut(self.right) = t.clone();
377 *self.size.get_mut(self.right) = len;
378 self.right_total += len;
384 pub fn check_stream(&mut self) -> io::IoResult<()> {
385 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
386 self.left, self.right, self.left_total, self.right_total);
387 if self.right_total - self.left_total > self.space {
388 debug!("scan window is {}, longer than space on line ({})",
389 self.right_total - self.left_total, self.space);
390 if !self.scan_stack_empty {
391 if self.left == *self.scan_stack.get(self.bottom) {
392 debug!("setting {} to infinity and popping", self.left);
393 let scanned = self.scan_pop_bottom();
394 *self.size.get_mut(scanned) = SIZE_INFINITY;
397 let left = (*self.token.get(self.left)).clone();
398 let left_size = *self.size.get(self.left);
399 try!(self.advance_left(left, left_size));
400 if self.left != self.right {
401 try!(self.check_stream());
406 pub fn scan_push(&mut self, x: uint) {
407 debug!("scan_push {}", x);
408 if self.scan_stack_empty {
409 self.scan_stack_empty = false;
412 self.top %= self.buf_len;
413 assert!((self.top != self.bottom));
415 *self.scan_stack.get_mut(self.top) = x;
417 pub fn scan_pop(&mut self) -> uint {
418 assert!((!self.scan_stack_empty));
419 let x = *self.scan_stack.get(self.top);
420 if self.top == self.bottom {
421 self.scan_stack_empty = true;
423 self.top += self.buf_len - 1u; self.top %= self.buf_len;
427 pub fn scan_top(&mut self) -> uint {
428 assert!((!self.scan_stack_empty));
429 return *self.scan_stack.get(self.top);
431 pub fn scan_pop_bottom(&mut self) -> uint {
432 assert!((!self.scan_stack_empty));
433 let x = *self.scan_stack.get(self.bottom);
434 if self.top == self.bottom {
435 self.scan_stack_empty = true;
437 self.bottom += 1u; self.bottom %= self.buf_len;
441 pub fn advance_right(&mut self) {
443 self.right %= self.buf_len;
444 assert!((self.right != self.left));
446 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
447 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
450 let ret = self.print(x.clone(), l);
452 Break(b) => self.left_total += b.blank_space,
454 assert_eq!(len, l); self.left_total += len;
458 if self.left != self.right {
460 self.left %= self.buf_len;
461 let left = (*self.token.get(self.left)).clone();
462 let left_size = *self.size.get(self.left);
463 try!(self.advance_left(left, left_size));
470 pub fn check_stack(&mut self, k: int) {
471 if !self.scan_stack_empty {
472 let x = self.scan_top();
473 match self.token.get(x) {
476 let popped = self.scan_pop();
477 *self.size.get_mut(popped) = *self.size.get(x) +
479 self.check_stack(k - 1);
483 // paper says + not =, but that makes no sense.
484 let popped = self.scan_pop();
485 *self.size.get_mut(popped) = 1;
486 self.check_stack(k + 1);
489 let popped = self.scan_pop();
490 *self.size.get_mut(popped) = *self.size.get(x) +
499 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
500 debug!("NEWLINE {}", amount);
501 let ret = write!(self.out, "\n");
502 self.pending_indentation = 0;
506 pub fn indent(&mut self, amount: int) {
507 debug!("INDENT {}", amount);
508 self.pending_indentation += amount;
510 pub fn get_top(&mut self) -> PrintStackElem {
511 let print_stack = &mut self.print_stack;
512 let n = print_stack.len();
514 *print_stack.get(n - 1u)
518 pbreak: Broken(Inconsistent)
522 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
523 while self.pending_indentation > 0 {
524 try!(write!(self.out, " "));
525 self.pending_indentation -= 1;
527 write!(self.out, "{}", s)
529 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
530 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
532 debug!("{}", buf_str(self.token.clone(),
540 let col = self.margin - self.space + b.offset;
541 debug!("print Begin -> push broken block at col {}", col);
542 self.print_stack.push(PrintStackElem {
544 pbreak: Broken(b.breaks)
547 debug!("print Begin -> push fitting block");
548 self.print_stack.push(PrintStackElem {
556 debug!("print End -> pop End");
557 let print_stack = &mut self.print_stack;
558 assert!((print_stack.len() != 0u));
559 print_stack.pop().unwrap();
563 let top = self.get_top();
566 debug!("print Break({}) in fitting block", b.blank_space);
567 self.space -= b.blank_space;
568 self.indent(b.blank_space);
571 Broken(Consistent) => {
572 debug!("print Break({}+{}) in consistent block",
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 Broken(Inconsistent) => {
580 debug!("print Break({}+{}) w/ newline in inconsistent",
581 top.offset, b.offset);
582 let ret = self.print_newline(top.offset + b.offset);
583 self.space = self.margin - (top.offset + b.offset);
586 debug!("print Break({}) w/o newline in inconsistent",
588 self.indent(b.blank_space);
589 self.space -= b.blank_space;
596 debug!("print String({})", s);
598 // assert!(l <= space);
600 self.print_str(s.as_slice())
603 // Eof should never get here.
610 // Convenience functions to talk to the printer.
613 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
614 p.pretty_print(Begin(BeginToken {
615 offset: indent as int,
620 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
621 rbox(p, indent, Inconsistent)
624 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
625 rbox(p, indent, Consistent)
628 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
629 p.pretty_print(Break(BreakToken {
631 blank_space: n as int
635 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
637 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
639 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
640 p.pretty_print(String(/* bad */ wrd.to_string(), wrd.len() as int))
643 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
644 p.pretty_print(String(/* bad */ wrd.to_string(), SIZE_INFINITY))
647 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
648 p.pretty_print(String(/* bad */ wrd.to_string(), 0))
651 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
652 break_offset(p, n, 0)
655 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
659 pub fn space(p: &mut Printer) -> io::IoResult<()> {
663 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
664 spaces(p, SIZE_INFINITY as uint)
667 pub fn hardbreak_tok_offset(off: int) -> Token {
668 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
671 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }