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)]
75 impl Copy for Breaks {}
78 pub struct BreakToken {
83 impl Copy for BreakToken {}
86 pub struct BeginToken {
91 impl Copy for BeginToken {}
95 String(string::String, int),
103 pub fn is_eof(&self) -> bool {
104 match *self { Eof => true, _ => false }
107 pub fn is_hardbreak_tok(&self) -> bool {
112 }) if bs == SIZE_INFINITY =>
120 pub fn tok_str(t: Token) -> string::String {
122 String(s, len) => return format!("STR({},{})", s, len),
123 Break(_) => return "BREAK".to_string(),
124 Begin(_) => return "BEGIN".to_string(),
125 End => return "END".to_string(),
126 Eof => return "EOF".to_string()
130 pub fn buf_str(toks: Vec<Token>,
137 assert_eq!(n, szs.len());
140 let mut s = string::String::from_str("[");
141 while i != right && l != 0u {
146 s.push_str(format!("{}={}",
148 tok_str(toks[i].clone())).as_slice());
153 return s.into_string();
156 pub enum PrintStackBreak {
161 impl Copy for PrintStackBreak {}
163 pub struct PrintStackElem {
165 pbreak: PrintStackBreak
168 impl Copy for PrintStackElem {}
170 static SIZE_INFINITY: int = 0xffff;
172 pub fn mk_printer(out: Box<io::Writer+'static>, linewidth: uint) -> Printer {
173 // Yes 3, it makes the ring buffers big enough to never
175 let n: uint = 3 * linewidth;
176 debug!("mk_printer {}", linewidth);
177 let token: Vec<Token> = Vec::from_elem(n, Eof);
178 let size: Vec<int> = Vec::from_elem(n, 0i);
179 let scan_stack: Vec<uint> = Vec::from_elem(n, 0u);
183 margin: linewidth as int,
184 space: linewidth as int,
191 scan_stack: scan_stack,
192 scan_stack_empty: true,
195 print_stack: Vec::new(),
196 pending_indentation: 0
201 /// In case you do not have the paper, here is an explanation of what's going
204 /// There is a stream of input tokens flowing through this printer.
206 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
207 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
208 /// case every token worth buffering is 1 char long, so it's ok.
210 /// Tokens are String, Break, and Begin/End to delimit blocks.
212 /// Begin tokens can carry an offset, saying "how far to indent when you break
213 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
214 /// breaking. Consistent breaking means that after the first break, no attempt
215 /// will be made to flow subsequent breaks together onto lines. Inconsistent
216 /// is the opposite. Inconsistent breaking example would be, say:
218 /// foo(hello, there, good, friends)
220 /// breaking inconsistently to become
225 /// whereas a consistent breaking would yield:
232 /// That is, in the consistent-break blocks we value vertical alignment
233 /// more than the ability to cram stuff onto a line. But in all cases if it
234 /// can make a block a one-liner, it'll do so.
236 /// Carrying on with high-level logic:
238 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
239 /// 'right' indices denote the active portion of the ring buffer as well as
240 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
241 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
242 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
243 /// and point-in-infinite-stream senses freely.
245 /// There is a parallel ring buffer, 'size', that holds the calculated size of
246 /// each token. Why calculated? Because for Begin/End pairs, the "size"
247 /// includes everything between the pair. That is, the "size" of Begin is
248 /// actually the sum of the sizes of everything between Begin and the paired
249 /// End that follows. Since that is arbitrarily far in the future, 'size' is
250 /// being rewritten regularly while the printer runs; in fact most of the
251 /// machinery is here to work out 'size' entries on the fly (and give up when
252 /// they're so obviously over-long that "infinity" is a good enough
253 /// approximation for purposes of line breaking).
255 /// The "input side" of the printer is managed as an abstract process called
256 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
257 /// manage calculating 'size'. SCAN is, in other words, the process of
258 /// calculating 'size' entries.
260 /// The "output side" of the printer is managed by an abstract process called
261 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
262 /// do with each token/size pair it consumes as it goes. It's trying to consume
263 /// the entire buffered window, but can't output anything until the size is >=
264 /// 0 (sizes are set to negative while they're pending calculation).
266 /// So SCAN takes input and buffers tokens and pending calculations, while
267 /// PRINT gobbles up completed calculations and tokens from the buffer. The
268 /// theory is that the two can never get more than 3N tokens apart, because
269 /// once there's "obviously" too much data to fit on a line, in a size
270 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
273 /// In this implementation (following the paper, again) the SCAN process is
274 /// the method called 'pretty_print', and the 'PRINT' process is the method
277 pub out: Box<io::Writer+'static>,
279 /// Width of lines we're constrained to
281 /// Number of spaces left on line
283 /// Index of left side of input stream
285 /// Index of right side of input stream
287 /// Ring-buffer stream goes through
289 /// Ring-buffer of calculated sizes
291 /// Running size of stream "...left"
293 /// Running size of stream "...right"
295 /// Pseudo-stack, really a ring too. Holds the
296 /// primary-ring-buffers index of the Begin that started the
297 /// current block, possibly with the most recent Break after that
298 /// Begin (if there is any) on top of it. Stuff is flushed off the
299 /// bottom as it becomes irrelevant due to the primary ring-buffer
301 scan_stack: Vec<uint> ,
302 /// Top==bottom disambiguator
303 scan_stack_empty: bool,
304 /// Index of top of scan_stack
306 /// Index of bottom of scan_stack
308 /// Stack of blocks-in-progress being flushed by print
309 print_stack: Vec<PrintStackElem> ,
310 /// Buffered indentation to avoid writing trailing whitespace
311 pending_indentation: int,
315 pub fn last_token(&mut self) -> Token {
316 self.token[self.right].clone()
318 // be very careful with this!
319 pub fn replace_last_token(&mut self, t: Token) {
320 self.token[self.right] = t;
322 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
323 debug!("pp ~[{},{}]", self.left, self.right);
326 if !self.scan_stack_empty {
328 let left = self.token[self.left].clone();
329 let left_size = self.size[self.left];
330 try!(self.advance_left(left, left_size));
336 if self.scan_stack_empty {
338 self.right_total = 1;
341 } else { self.advance_right(); }
342 debug!("pp Begin({})/buffer ~[{},{}]",
343 b.offset, self.left, self.right);
344 self.token[self.right] = t;
345 self.size[self.right] = -self.right_total;
346 let right = self.right;
347 self.scan_push(right);
351 if self.scan_stack_empty {
352 debug!("pp End/print ~[{},{}]", self.left, self.right);
355 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
356 self.advance_right();
357 self.token[self.right] = t;
358 self.size[self.right] = -1;
359 let right = self.right;
360 self.scan_push(right);
365 if self.scan_stack_empty {
367 self.right_total = 1;
370 } else { self.advance_right(); }
371 debug!("pp Break({})/buffer ~[{},{}]",
372 b.offset, self.left, self.right);
374 let right = self.right;
375 self.scan_push(right);
376 self.token[self.right] = t;
377 self.size[self.right] = -self.right_total;
378 self.right_total += b.blank_space;
381 String(ref s, len) => {
382 if self.scan_stack_empty {
383 debug!("pp String('{}')/print ~[{},{}]",
384 *s, self.left, self.right);
385 self.print(t.clone(), len)
387 debug!("pp String('{}')/buffer ~[{},{}]",
388 *s, self.left, self.right);
389 self.advance_right();
390 self.token[self.right] = t.clone();
391 self.size[self.right] = len;
392 self.right_total += len;
398 pub fn check_stream(&mut self) -> io::IoResult<()> {
399 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
400 self.left, self.right, self.left_total, self.right_total);
401 if self.right_total - self.left_total > self.space {
402 debug!("scan window is {}, longer than space on line ({})",
403 self.right_total - self.left_total, self.space);
404 if !self.scan_stack_empty {
405 if self.left == self.scan_stack[self.bottom] {
406 debug!("setting {} to infinity and popping", self.left);
407 let scanned = self.scan_pop_bottom();
408 self.size[scanned] = SIZE_INFINITY;
411 let left = self.token[self.left].clone();
412 let left_size = self.size[self.left];
413 try!(self.advance_left(left, left_size));
414 if self.left != self.right {
415 try!(self.check_stream());
420 pub fn scan_push(&mut self, x: uint) {
421 debug!("scan_push {}", x);
422 if self.scan_stack_empty {
423 self.scan_stack_empty = false;
426 self.top %= self.buf_len;
427 assert!((self.top != self.bottom));
429 self.scan_stack[self.top] = x;
431 pub fn scan_pop(&mut self) -> uint {
432 assert!((!self.scan_stack_empty));
433 let x = self.scan_stack[self.top];
434 if self.top == self.bottom {
435 self.scan_stack_empty = true;
437 self.top += self.buf_len - 1u; self.top %= self.buf_len;
441 pub fn scan_top(&mut self) -> uint {
442 assert!((!self.scan_stack_empty));
443 return self.scan_stack[self.top];
445 pub fn scan_pop_bottom(&mut self) -> uint {
446 assert!((!self.scan_stack_empty));
447 let x = self.scan_stack[self.bottom];
448 if self.top == self.bottom {
449 self.scan_stack_empty = true;
451 self.bottom += 1u; self.bottom %= self.buf_len;
455 pub fn advance_right(&mut self) {
457 self.right %= self.buf_len;
458 assert!((self.right != self.left));
460 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
461 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
464 let ret = self.print(x.clone(), l);
466 Break(b) => self.left_total += b.blank_space,
468 assert_eq!(len, l); self.left_total += len;
472 if self.left != self.right {
474 self.left %= self.buf_len;
475 let left = self.token[self.left].clone();
476 let left_size = self.size[self.left];
477 try!(self.advance_left(left, left_size));
484 pub fn check_stack(&mut self, k: int) {
485 if !self.scan_stack_empty {
486 let x = self.scan_top();
487 match self.token[x] {
490 let popped = self.scan_pop();
491 self.size[popped] = self.size[x] + self.right_total;
492 self.check_stack(k - 1);
496 // paper says + not =, but that makes no sense.
497 let popped = self.scan_pop();
498 self.size[popped] = 1;
499 self.check_stack(k + 1);
502 let popped = self.scan_pop();
503 self.size[popped] = self.size[x] + self.right_total;
511 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
512 debug!("NEWLINE {}", amount);
513 let ret = write!(self.out, "\n");
514 self.pending_indentation = 0;
518 pub fn indent(&mut self, amount: int) {
519 debug!("INDENT {}", amount);
520 self.pending_indentation += amount;
522 pub fn get_top(&mut self) -> PrintStackElem {
523 let print_stack = &mut self.print_stack;
524 let n = print_stack.len();
526 (*print_stack)[n - 1]
530 pbreak: Broken(Inconsistent)
534 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
535 while self.pending_indentation > 0 {
536 try!(write!(self.out, " "));
537 self.pending_indentation -= 1;
539 write!(self.out, "{}", s)
541 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
542 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
544 debug!("{}", buf_str(self.token.clone(),
552 let col = self.margin - self.space + b.offset;
553 debug!("print Begin -> push broken block at col {}", col);
554 self.print_stack.push(PrintStackElem {
556 pbreak: Broken(b.breaks)
559 debug!("print Begin -> push fitting block");
560 self.print_stack.push(PrintStackElem {
568 debug!("print End -> pop End");
569 let print_stack = &mut self.print_stack;
570 assert!((print_stack.len() != 0u));
571 print_stack.pop().unwrap();
575 let top = self.get_top();
578 debug!("print Break({}) in fitting block", b.blank_space);
579 self.space -= b.blank_space;
580 self.indent(b.blank_space);
583 Broken(Consistent) => {
584 debug!("print Break({}+{}) in consistent block",
585 top.offset, b.offset);
586 let ret = self.print_newline(top.offset + b.offset);
587 self.space = self.margin - (top.offset + b.offset);
590 Broken(Inconsistent) => {
592 debug!("print Break({}+{}) w/ newline in inconsistent",
593 top.offset, b.offset);
594 let ret = self.print_newline(top.offset + b.offset);
595 self.space = self.margin - (top.offset + b.offset);
598 debug!("print Break({}) w/o newline in inconsistent",
600 self.indent(b.blank_space);
601 self.space -= b.blank_space;
608 debug!("print String({})", s);
610 // assert!(l <= space);
612 self.print_str(s.as_slice())
615 // Eof should never get here.
622 // Convenience functions to talk to the printer.
625 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
626 p.pretty_print(Begin(BeginToken {
627 offset: indent as int,
632 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
633 rbox(p, indent, Inconsistent)
636 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
637 rbox(p, indent, Consistent)
640 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
641 p.pretty_print(Break(BreakToken {
643 blank_space: n as int
647 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
649 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
651 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
652 p.pretty_print(String(/* bad */ wrd.to_string(), wrd.len() as int))
655 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
656 p.pretty_print(String(/* bad */ wrd.to_string(), SIZE_INFINITY))
659 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
660 p.pretty_print(String(/* bad */ wrd.to_string(), 0))
663 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
664 break_offset(p, n, 0)
667 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
671 pub fn space(p: &mut Printer) -> io::IoResult<()> {
675 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
676 spaces(p, SIZE_INFINITY as uint)
679 pub fn hardbreak_tok_offset(off: int) -> Token {
680 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
683 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }