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.
12 * This pretty-printer is a direct reimplementation of Philip Karlton's
13 * 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.
18 * The algorithm's aim is to break a stream into as few lines as possible
19 * while respecting the indentation-consistency requirements of the enclosing
20 * block, and avoiding breaking at silly places on block boundaries, for
21 * example, between "x" and ")" in "x)".
23 * I am implementing this algorithm because it comes with 20 pages of
24 * documentation explaining its theory, and because it addresses the set of
25 * concerns I've seen other pretty-printers fall down on. Weirdly. Even though
26 * it's 32 years old. What can I say?
28 * Despite some redundancies and quirks in the way it's implemented in that
29 * paper, I've opted to keep the implementation here as similar as I can,
30 * changing only what was blatantly wrong, a typo, or sufficiently
31 * non-idiomatic rust that it really stuck out.
33 * In particular you'll see a certain amount of churn related to INTEGER vs.
34 * CARDINAL in the Mesa implementation. Mesa apparently interconverts the two
35 * somewhat readily? In any case, I've used uint for indices-in-buffers and
36 * ints for character-sizes-and-indentation-offsets. This respects the need
37 * for ints to "go negative" while carrying a pending-calculation balance, and
38 * helps differentiate all the numbers flying around internally (slightly).
40 * I also inverted the indentation arithmetic used in the print stack, since
41 * the Mesa implementation (somewhat randomly) stores the offset on the print
42 * stack in terms of margin-col rather than col itself. I store col.
44 * I also implemented a small change in the String token, in that I store an
45 * explicit length for the string. For most tokens this is just the length of
46 * the accompanying string. But it's necessary to permit it to differ, for
47 * encoding things that are supposed to "go on their own line" -- certain
48 * classes of comment and blank-line -- where relying on adjacent
49 * hardbreak-like Break tokens with long blankness indication doesn't actually
50 * work. To see why, consider when there is a "thing that should be on its own
51 * line" between two long blocks, say functions. If you put a hardbreak after
52 * each function (or before each) and the breaking algorithm decides to break
53 * there anyways (because the functions themselves are long) you wind up with
54 * extra blank lines. If you don't put hardbreaks you can wind up with the
55 * "thing which should be on its own line" not getting its own line in the
56 * rare case of "really small functions" or such. This re-occurs with comments
57 * and explicit blank lines. So in those cases we use a string with a payload
58 * we want isolated to a line and an explicit length that's huge, surrounded
59 * by two zero-length breaks. The algorithm will try its best to fit it on a
60 * line (which it can't) and so naturally place the content on its own line to
61 * avoid combining it with other lines and making matters even worse.
65 use std::string::String;
67 #[deriving(Clone, PartialEq)]
74 pub struct BreakToken {
80 pub struct BeginToken {
95 pub fn is_eof(&self) -> bool {
96 match *self { Eof => true, _ => false }
99 pub fn is_hardbreak_tok(&self) -> bool {
104 }) if bs == SIZE_INFINITY =>
112 pub fn tok_str(t: Token) -> String {
114 String(s, len) => return format!("STR({},{})", s, len).to_string(),
115 Break(_) => return "BREAK".to_string(),
116 Begin(_) => return "BEGIN".to_string(),
117 End => return "END".to_string(),
118 Eof => return "EOF".to_string()
122 pub fn buf_str(toks: Vec<Token>,
129 assert_eq!(n, szs.len());
132 let mut s = String::from_str("[");
133 while i != right && l != 0u {
138 s.push_str(format!("{}={}",
140 tok_str(toks.get(i).clone())).as_slice());
145 return s.into_string();
148 pub enum PrintStackBreak {
153 pub struct PrintStackElem {
155 pbreak: PrintStackBreak
158 static SIZE_INFINITY: int = 0xffff;
160 pub fn mk_printer(out: Box<io::Writer>, linewidth: uint) -> Printer {
161 // Yes 3, it makes the ring buffers big enough to never
163 let n: uint = 3 * linewidth;
164 debug!("mk_printer {}", linewidth);
165 let token: Vec<Token> = Vec::from_elem(n, Eof);
166 let size: Vec<int> = Vec::from_elem(n, 0i);
167 let scan_stack: Vec<uint> = Vec::from_elem(n, 0u);
171 margin: linewidth as int,
172 space: linewidth as int,
179 scan_stack: scan_stack,
180 scan_stack_empty: true,
183 print_stack: Vec::new(),
184 pending_indentation: 0
190 * In case you do not have the paper, here is an explanation of what's going
193 * There is a stream of input tokens flowing through this printer.
195 * The printer buffers up to 3N tokens inside itself, where N is linewidth.
196 * Yes, linewidth is chars and tokens are multi-char, but in the worst
197 * case every token worth buffering is 1 char long, so it's ok.
199 * Tokens are String, Break, and Begin/End to delimit blocks.
201 * Begin tokens can carry an offset, saying "how far to indent when you break
202 * inside here", as well as a flag indicating "consistent" or "inconsistent"
203 * breaking. Consistent breaking means that after the first break, no attempt
204 * will be made to flow subsequent breaks together onto lines. Inconsistent
205 * is the opposite. Inconsistent breaking example would be, say:
207 * foo(hello, there, good, friends)
209 * breaking inconsistently to become
214 * whereas a consistent breaking would yield:
221 * That is, in the consistent-break blocks we value vertical alignment
222 * more than the ability to cram stuff onto a line. But in all cases if it
223 * can make a block a one-liner, it'll do so.
225 * Carrying on with high-level logic:
227 * The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
228 * 'right' indices denote the active portion of the ring buffer as well as
229 * describing hypothetical points-in-the-infinite-stream at most 3N tokens
230 * apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
231 * between using 'left' and 'right' terms to denote the wrapepd-to-ring-buffer
232 * and point-in-infinite-stream senses freely.
234 * There is a parallel ring buffer, 'size', that holds the calculated size of
235 * each token. Why calculated? Because for Begin/End pairs, the "size"
236 * includes everything between the pair. That is, the "size" of Begin is
237 * actually the sum of the sizes of everything between Begin and the paired
238 * End that follows. Since that is arbitrarily far in the future, 'size' is
239 * being rewritten regularly while the printer runs; in fact most of the
240 * machinery is here to work out 'size' entries on the fly (and give up when
241 * they're so obviously over-long that "infinity" is a good enough
242 * approximation for purposes of line breaking).
244 * The "input side" of the printer is managed as an abstract process called
245 * SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
246 * manage calculating 'size'. SCAN is, in other words, the process of
247 * calculating 'size' entries.
249 * The "output side" of the printer is managed by an abstract process called
250 * PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
251 * do with each token/size pair it consumes as it goes. It's trying to consume
252 * the entire buffered window, but can't output anything until the size is >=
253 * 0 (sizes are set to negative while they're pending calculation).
255 * So SCAN takes input and buffers tokens and pending calculations, while
256 * PRINT gobbles up completed calculations and tokens from the buffer. The
257 * theory is that the two can never get more than 3N tokens apart, because
258 * once there's "obviously" too much data to fit on a line, in a size
259 * calculation, SCAN will write "infinity" to the size and let PRINT consume
262 * In this implementation (following the paper, again) the SCAN process is
263 * the method called 'pretty_print', and the 'PRINT' process is the method
267 pub out: Box<io::Writer>,
269 margin: int, // width of lines we're constrained to
270 space: int, // number of spaces left on line
271 left: uint, // index of left side of input stream
272 right: uint, // index of right side of input stream
273 token: Vec<Token> , // ring-buffr stream goes through
274 size: Vec<int> , // ring-buffer of calculated sizes
275 left_total: int, // running size of stream "...left"
276 right_total: int, // running size of stream "...right"
277 // pseudo-stack, really a ring too. Holds the
278 // primary-ring-buffers index of the Begin that started the
279 // current block, possibly with the most recent Break after that
280 // Begin (if there is any) on top of it. Stuff is flushed off the
281 // bottom as it becomes irrelevant due to the primary ring-buffer
283 scan_stack: Vec<uint> ,
284 scan_stack_empty: bool, // top==bottom disambiguator
285 top: uint, // index of top of scan_stack
286 bottom: uint, // index of bottom of scan_stack
287 // stack of blocks-in-progress being flushed by print
288 print_stack: Vec<PrintStackElem> ,
289 // buffered indentation to avoid writing trailing whitespace
290 pending_indentation: int,
294 pub fn last_token(&mut self) -> Token {
295 (*self.token.get(self.right)).clone()
297 // be very careful with this!
298 pub fn replace_last_token(&mut self, t: Token) {
299 *self.token.get_mut(self.right) = t;
301 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
302 debug!("pp ~[{},{}]", self.left, self.right);
305 if !self.scan_stack_empty {
307 let left = (*self.token.get(self.left)).clone();
308 let left_size = *self.size.get(self.left);
309 try!(self.advance_left(left, left_size));
315 if self.scan_stack_empty {
317 self.right_total = 1;
320 } else { self.advance_right(); }
321 debug!("pp Begin({})/buffer ~[{},{}]",
322 b.offset, self.left, self.right);
323 *self.token.get_mut(self.right) = t;
324 *self.size.get_mut(self.right) = -self.right_total;
325 let right = self.right;
326 self.scan_push(right);
330 if self.scan_stack_empty {
331 debug!("pp End/print ~[{},{}]", self.left, self.right);
334 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
335 self.advance_right();
336 *self.token.get_mut(self.right) = t;
337 *self.size.get_mut(self.right) = -1;
338 let right = self.right;
339 self.scan_push(right);
344 if self.scan_stack_empty {
346 self.right_total = 1;
349 } else { self.advance_right(); }
350 debug!("pp Break({})/buffer ~[{},{}]",
351 b.offset, self.left, self.right);
353 let right = self.right;
354 self.scan_push(right);
355 *self.token.get_mut(self.right) = t;
356 *self.size.get_mut(self.right) = -self.right_total;
357 self.right_total += b.blank_space;
360 String(ref s, len) => {
361 if self.scan_stack_empty {
362 debug!("pp String('{}')/print ~[{},{}]",
363 *s, self.left, self.right);
364 self.print(t.clone(), len)
366 debug!("pp String('{}')/buffer ~[{},{}]",
367 *s, self.left, self.right);
368 self.advance_right();
369 *self.token.get_mut(self.right) = t.clone();
370 *self.size.get_mut(self.right) = len;
371 self.right_total += len;
377 pub fn check_stream(&mut self) -> io::IoResult<()> {
378 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
379 self.left, self.right, self.left_total, self.right_total);
380 if self.right_total - self.left_total > self.space {
381 debug!("scan window is {}, longer than space on line ({})",
382 self.right_total - self.left_total, self.space);
383 if !self.scan_stack_empty {
384 if self.left == *self.scan_stack.get(self.bottom) {
385 debug!("setting {} to infinity and popping", self.left);
386 let scanned = self.scan_pop_bottom();
387 *self.size.get_mut(scanned) = SIZE_INFINITY;
390 let left = (*self.token.get(self.left)).clone();
391 let left_size = *self.size.get(self.left);
392 try!(self.advance_left(left, left_size));
393 if self.left != self.right {
394 try!(self.check_stream());
399 pub fn scan_push(&mut self, x: uint) {
400 debug!("scan_push {}", x);
401 if self.scan_stack_empty {
402 self.scan_stack_empty = false;
405 self.top %= self.buf_len;
406 assert!((self.top != self.bottom));
408 *self.scan_stack.get_mut(self.top) = x;
410 pub fn scan_pop(&mut self) -> uint {
411 assert!((!self.scan_stack_empty));
412 let x = *self.scan_stack.get(self.top);
413 if self.top == self.bottom {
414 self.scan_stack_empty = true;
416 self.top += self.buf_len - 1u; self.top %= self.buf_len;
420 pub fn scan_top(&mut self) -> uint {
421 assert!((!self.scan_stack_empty));
422 return *self.scan_stack.get(self.top);
424 pub fn scan_pop_bottom(&mut self) -> uint {
425 assert!((!self.scan_stack_empty));
426 let x = *self.scan_stack.get(self.bottom);
427 if self.top == self.bottom {
428 self.scan_stack_empty = true;
430 self.bottom += 1u; self.bottom %= self.buf_len;
434 pub fn advance_right(&mut self) {
436 self.right %= self.buf_len;
437 assert!((self.right != self.left));
439 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
440 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
443 let ret = self.print(x.clone(), l);
445 Break(b) => self.left_total += b.blank_space,
447 assert_eq!(len, l); self.left_total += len;
451 if self.left != self.right {
453 self.left %= self.buf_len;
454 let left = (*self.token.get(self.left)).clone();
455 let left_size = *self.size.get(self.left);
456 try!(self.advance_left(left, left_size));
463 pub fn check_stack(&mut self, k: int) {
464 if !self.scan_stack_empty {
465 let x = self.scan_top();
466 match self.token.get(x) {
469 let popped = self.scan_pop();
470 *self.size.get_mut(popped) = *self.size.get(x) +
472 self.check_stack(k - 1);
476 // paper says + not =, but that makes no sense.
477 let popped = self.scan_pop();
478 *self.size.get_mut(popped) = 1;
479 self.check_stack(k + 1);
482 let popped = self.scan_pop();
483 *self.size.get_mut(popped) = *self.size.get(x) +
492 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
493 debug!("NEWLINE {}", amount);
494 let ret = write!(self.out, "\n");
495 self.pending_indentation = 0;
499 pub fn indent(&mut self, amount: int) {
500 debug!("INDENT {}", amount);
501 self.pending_indentation += amount;
503 pub fn get_top(&mut self) -> PrintStackElem {
504 let print_stack = &mut self.print_stack;
505 let n = print_stack.len();
507 *print_stack.get(n - 1u)
511 pbreak: Broken(Inconsistent)
515 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
516 while self.pending_indentation > 0 {
517 try!(write!(self.out, " "));
518 self.pending_indentation -= 1;
520 write!(self.out, "{}", s)
522 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
523 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
525 debug!("{}", buf_str(self.token.clone(),
533 let col = self.margin - self.space + b.offset;
534 debug!("print Begin -> push broken block at col {}", col);
535 self.print_stack.push(PrintStackElem {
537 pbreak: Broken(b.breaks)
540 debug!("print Begin -> push fitting block");
541 self.print_stack.push(PrintStackElem {
549 debug!("print End -> pop End");
550 let print_stack = &mut self.print_stack;
551 assert!((print_stack.len() != 0u));
552 print_stack.pop().unwrap();
556 let top = self.get_top();
559 debug!("print Break({}) in fitting block", b.blank_space);
560 self.space -= b.blank_space;
561 self.indent(b.blank_space);
564 Broken(Consistent) => {
565 debug!("print Break({}+{}) in consistent block",
566 top.offset, b.offset);
567 let ret = self.print_newline(top.offset + b.offset);
568 self.space = self.margin - (top.offset + b.offset);
571 Broken(Inconsistent) => {
573 debug!("print Break({}+{}) w/ newline in inconsistent",
574 top.offset, b.offset);
575 let ret = self.print_newline(top.offset + b.offset);
576 self.space = self.margin - (top.offset + b.offset);
579 debug!("print Break({}) w/o newline in inconsistent",
581 self.indent(b.blank_space);
582 self.space -= b.blank_space;
589 debug!("print String({})", s);
591 // assert!(l <= space);
593 self.print_str(s.as_slice())
596 // Eof should never get here.
603 // Convenience functions to talk to the printer.
606 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
607 p.pretty_print(Begin(BeginToken {
608 offset: indent as int,
613 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
614 rbox(p, indent, Inconsistent)
617 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
618 rbox(p, indent, Consistent)
621 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
622 p.pretty_print(Break(BreakToken {
624 blank_space: n as int
628 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
630 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
632 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
633 p.pretty_print(String(/* bad */ wrd.to_string(), wrd.len() as int))
636 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
637 p.pretty_print(String(/* bad */ wrd.to_string(), SIZE_INFINITY))
640 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
641 p.pretty_print(String(/* bad */ wrd.to_string(), 0))
644 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
645 break_offset(p, n, 0)
648 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
652 pub fn space(p: &mut Printer) -> io::IoResult<()> {
656 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
657 spaces(p, SIZE_INFINITY as uint)
660 pub fn hardbreak_tok_offset(off: int) -> Token {
661 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
664 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }