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::strbuf::StrBuf;
67 #[deriving(Clone, Eq)]
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) -> ~str {
114 String(s, len) => return format!("STR({},{})", s, len),
115 Break(_) => return ~"BREAK",
116 Begin(_) => return ~"BEGIN",
117 End => return ~"END",
122 pub fn buf_str(toks: Vec<Token>,
129 assert_eq!(n, szs.len());
132 let mut s = StrBuf::from_str("[");
133 while i != right && l != 0u {
138 s.push_str(format!("{}={}", szs.get(i), tok_str(toks.get(i).clone())));
143 return s.into_owned();
146 pub enum PrintStackBreak {
151 pub struct PrintStackElem {
153 pbreak: PrintStackBreak
156 static SIZE_INFINITY: int = 0xffff;
158 pub fn mk_printer(out: ~io::Writer, 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, 0);
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
188 * In case you do not have the paper, here is an explanation of what's going
191 * There is a stream of input tokens flowing through this printer.
193 * The printer buffers up to 3N tokens inside itself, where N is linewidth.
194 * Yes, linewidth is chars and tokens are multi-char, but in the worst
195 * case every token worth buffering is 1 char long, so it's ok.
197 * Tokens are String, Break, and Begin/End to delimit blocks.
199 * Begin tokens can carry an offset, saying "how far to indent when you break
200 * inside here", as well as a flag indicating "consistent" or "inconsistent"
201 * breaking. Consistent breaking means that after the first break, no attempt
202 * will be made to flow subsequent breaks together onto lines. Inconsistent
203 * is the opposite. Inconsistent breaking example would be, say:
205 * foo(hello, there, good, friends)
207 * breaking inconsistently to become
212 * whereas a consistent breaking would yield:
219 * That is, in the consistent-break blocks we value vertical alignment
220 * more than the ability to cram stuff onto a line. But in all cases if it
221 * can make a block a one-liner, it'll do so.
223 * Carrying on with high-level logic:
225 * The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
226 * 'right' indices denote the active portion of the ring buffer as well as
227 * describing hypothetical points-in-the-infinite-stream at most 3N tokens
228 * apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
229 * between using 'left' and 'right' terms to denote the wrapepd-to-ring-buffer
230 * and point-in-infinite-stream senses freely.
232 * There is a parallel ring buffer, 'size', that holds the calculated size of
233 * each token. Why calculated? Because for Begin/End pairs, the "size"
234 * includes everything betwen the pair. That is, the "size" of Begin is
235 * actually the sum of the sizes of everything between Begin and the paired
236 * End that follows. Since that is arbitrarily far in the future, 'size' is
237 * being rewritten regularly while the printer runs; in fact most of the
238 * machinery is here to work out 'size' entries on the fly (and give up when
239 * they're so obviously over-long that "infinity" is a good enough
240 * approximation for purposes of line breaking).
242 * The "input side" of the printer is managed as an abstract process called
243 * SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
244 * manage calculating 'size'. SCAN is, in other words, the process of
245 * calculating 'size' entries.
247 * The "output side" of the printer is managed by an abstract process called
248 * PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
249 * do with each token/size pair it consumes as it goes. It's trying to consume
250 * the entire buffered window, but can't output anything until the size is >=
251 * 0 (sizes are set to negative while they're pending calculation).
253 * So SCAN takes input and buffers tokens and pending calculations, while
254 * PRINT gobbles up completed calculations and tokens from the buffer. The
255 * theory is that the two can never get more than 3N tokens apart, because
256 * once there's "obviously" too much data to fit on a line, in a size
257 * calculation, SCAN will write "infinity" to the size and let PRINT consume
260 * In this implementation (following the paper, again) the SCAN process is
261 * the method called 'pretty_print', and the 'PRINT' process is the method
265 pub out: ~io::Writer,
267 margin: int, // width of lines we're constrained to
268 space: int, // number of spaces left on line
269 left: uint, // index of left side of input stream
270 right: uint, // index of right side of input stream
271 token: Vec<Token> , // ring-buffr stream goes through
272 size: Vec<int> , // ring-buffer of calculated sizes
273 left_total: int, // running size of stream "...left"
274 right_total: int, // running size of stream "...right"
275 // pseudo-stack, really a ring too. Holds the
276 // primary-ring-buffers index of the Begin that started the
277 // current block, possibly with the most recent Break after that
278 // Begin (if there is any) on top of it. Stuff is flushed off the
279 // bottom as it becomes irrelevant due to the primary ring-buffer
281 scan_stack: Vec<uint> ,
282 scan_stack_empty: bool, // top==bottom disambiguator
283 top: uint, // index of top of scan_stack
284 bottom: uint, // index of bottom of scan_stack
285 // stack of blocks-in-progress being flushed by print
286 print_stack: Vec<PrintStackElem> ,
287 // buffered indentation to avoid writing trailing whitespace
288 pending_indentation: int,
292 pub fn last_token(&mut self) -> Token {
293 (*self.token.get(self.right)).clone()
295 // be very careful with this!
296 pub fn replace_last_token(&mut self, t: Token) {
297 *self.token.get_mut(self.right) = t;
299 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
300 debug!("pp ~[{},{}]", self.left, self.right);
303 if !self.scan_stack_empty {
305 let left = (*self.token.get(self.left)).clone();
306 let left_size = *self.size.get(self.left);
307 try!(self.advance_left(left, left_size));
313 if self.scan_stack_empty {
315 self.right_total = 1;
318 } else { self.advance_right(); }
319 debug!("pp Begin({})/buffer ~[{},{}]",
320 b.offset, self.left, self.right);
321 *self.token.get_mut(self.right) = t;
322 *self.size.get_mut(self.right) = -self.right_total;
323 self.scan_push(self.right);
327 if self.scan_stack_empty {
328 debug!("pp End/print ~[{},{}]", self.left, self.right);
331 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
332 self.advance_right();
333 *self.token.get_mut(self.right) = t;
334 *self.size.get_mut(self.right) = -1;
335 self.scan_push(self.right);
340 if self.scan_stack_empty {
342 self.right_total = 1;
345 } else { self.advance_right(); }
346 debug!("pp Break({})/buffer ~[{},{}]",
347 b.offset, self.left, self.right);
349 self.scan_push(self.right);
350 *self.token.get_mut(self.right) = t;
351 *self.size.get_mut(self.right) = -self.right_total;
352 self.right_total += b.blank_space;
355 String(ref s, len) => {
356 if self.scan_stack_empty {
357 debug!("pp String('{}')/print ~[{},{}]",
358 *s, self.left, self.right);
359 self.print(t.clone(), len)
361 debug!("pp String('{}')/buffer ~[{},{}]",
362 *s, self.left, self.right);
363 self.advance_right();
364 *self.token.get_mut(self.right) = t.clone();
365 *self.size.get_mut(self.right) = len;
366 self.right_total += len;
372 pub fn check_stream(&mut self) -> io::IoResult<()> {
373 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
374 self.left, self.right, self.left_total, self.right_total);
375 if self.right_total - self.left_total > self.space {
376 debug!("scan window is {}, longer than space on line ({})",
377 self.right_total - self.left_total, self.space);
378 if !self.scan_stack_empty {
379 if self.left == *self.scan_stack.get(self.bottom) {
380 debug!("setting {} to infinity and popping", self.left);
381 let scanned = self.scan_pop_bottom();
382 *self.size.get_mut(scanned) = SIZE_INFINITY;
385 let left = (*self.token.get(self.left)).clone();
386 let left_size = *self.size.get(self.left);
387 try!(self.advance_left(left, left_size));
388 if self.left != self.right {
389 try!(self.check_stream());
394 pub fn scan_push(&mut self, x: uint) {
395 debug!("scan_push {}", x);
396 if self.scan_stack_empty {
397 self.scan_stack_empty = false;
400 self.top %= self.buf_len;
401 assert!((self.top != self.bottom));
403 *self.scan_stack.get_mut(self.top) = x;
405 pub fn scan_pop(&mut self) -> uint {
406 assert!((!self.scan_stack_empty));
407 let x = *self.scan_stack.get(self.top);
408 if self.top == self.bottom {
409 self.scan_stack_empty = true;
411 self.top += self.buf_len - 1u; self.top %= self.buf_len;
415 pub fn scan_top(&mut self) -> uint {
416 assert!((!self.scan_stack_empty));
417 return *self.scan_stack.get(self.top);
419 pub fn scan_pop_bottom(&mut self) -> uint {
420 assert!((!self.scan_stack_empty));
421 let x = *self.scan_stack.get(self.bottom);
422 if self.top == self.bottom {
423 self.scan_stack_empty = true;
425 self.bottom += 1u; self.bottom %= self.buf_len;
429 pub fn advance_right(&mut self) {
431 self.right %= self.buf_len;
432 assert!((self.right != self.left));
434 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
435 debug!("advnce_left ~[{},{}], sizeof({})={}", self.left, self.right,
438 let ret = self.print(x.clone(), l);
440 Break(b) => self.left_total += b.blank_space,
442 assert_eq!(len, l); self.left_total += len;
446 if self.left != self.right {
448 self.left %= self.buf_len;
449 let left = (*self.token.get(self.left)).clone();
450 let left_size = *self.size.get(self.left);
451 try!(self.advance_left(left, left_size));
458 pub fn check_stack(&mut self, k: int) {
459 if !self.scan_stack_empty {
460 let x = self.scan_top();
461 match self.token.get(x) {
464 let popped = self.scan_pop();
465 *self.size.get_mut(popped) = *self.size.get(x) +
467 self.check_stack(k - 1);
471 // paper says + not =, but that makes no sense.
472 let popped = self.scan_pop();
473 *self.size.get_mut(popped) = 1;
474 self.check_stack(k + 1);
477 let popped = self.scan_pop();
478 *self.size.get_mut(popped) = *self.size.get(x) +
487 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
488 debug!("NEWLINE {}", amount);
489 let ret = write!(self.out, "\n");
490 self.pending_indentation = 0;
494 pub fn indent(&mut self, amount: int) {
495 debug!("INDENT {}", amount);
496 self.pending_indentation += amount;
498 pub fn get_top(&mut self) -> PrintStackElem {
499 let print_stack = &mut self.print_stack;
500 let n = print_stack.len();
502 *print_stack.get(n - 1u)
506 pbreak: Broken(Inconsistent)
510 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
511 while self.pending_indentation > 0 {
512 try!(write!(self.out, " "));
513 self.pending_indentation -= 1;
515 write!(self.out, "{}", s)
517 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
518 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
520 debug!("{}", buf_str(self.token.clone(),
528 let col = self.margin - self.space + b.offset;
529 debug!("print Begin -> push broken block at col {}", col);
530 self.print_stack.push(PrintStackElem {
532 pbreak: Broken(b.breaks)
535 debug!("print Begin -> push fitting block");
536 self.print_stack.push(PrintStackElem {
544 debug!("print End -> pop End");
545 let print_stack = &mut self.print_stack;
546 assert!((print_stack.len() != 0u));
547 print_stack.pop().unwrap();
551 let top = self.get_top();
554 debug!("print Break({}) in fitting block", b.blank_space);
555 self.space -= b.blank_space;
556 self.indent(b.blank_space);
559 Broken(Consistent) => {
560 debug!("print Break({}+{}) in consistent block",
561 top.offset, b.offset);
562 let ret = self.print_newline(top.offset + b.offset);
563 self.space = self.margin - (top.offset + b.offset);
566 Broken(Inconsistent) => {
568 debug!("print Break({}+{}) w/ newline in inconsistent",
569 top.offset, b.offset);
570 let ret = self.print_newline(top.offset + b.offset);
571 self.space = self.margin - (top.offset + b.offset);
574 debug!("print Break({}) w/o newline in inconsistent",
576 self.indent(b.blank_space);
577 self.space -= b.blank_space;
584 debug!("print String({})", s);
586 // assert!(l <= space);
591 // Eof should never get here.
598 // Convenience functions to talk to the printer.
601 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
602 p.pretty_print(Begin(BeginToken {
603 offset: indent as int,
608 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
609 rbox(p, indent, Inconsistent)
612 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
613 rbox(p, indent, Consistent)
616 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
617 p.pretty_print(Break(BreakToken {
619 blank_space: n as int
623 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
625 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
627 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
628 p.pretty_print(String(/* bad */ wrd.to_str(), wrd.len() as int))
631 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
632 p.pretty_print(String(/* bad */ wrd.to_str(), SIZE_INFINITY))
635 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
636 p.pretty_print(String(/* bad */ wrd.to_str(), 0))
639 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
640 break_offset(p, n, 0)
643 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
647 pub fn space(p: &mut Printer) -> io::IoResult<()> {
651 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
652 spaces(p, SIZE_INFINITY as uint)
655 pub fn hardbreak_tok_offset(off: int) -> Token {
656 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
659 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }