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::*;
68 use std::iter::repeat;
70 #[derive(Clone, Copy, PartialEq)]
76 #[derive(Clone, Copy)]
77 pub struct BreakToken {
82 #[derive(Clone, Copy)]
83 pub struct BeginToken {
90 String(string::String, int),
98 pub fn is_eof(&self) -> bool {
99 match *self { Eof => true, _ => false }
102 pub fn is_hardbreak_tok(&self) -> bool {
107 }) if bs == SIZE_INFINITY =>
115 pub fn tok_str(t: Token) -> string::String {
117 String(s, len) => return format!("STR({},{})", s, len),
118 Break(_) => return "BREAK".to_string(),
119 Begin(_) => return "BEGIN".to_string(),
120 End => return "END".to_string(),
121 Eof => return "EOF".to_string()
125 pub fn buf_str(toks: Vec<Token>,
132 assert_eq!(n, szs.len());
135 let mut s = string::String::from_str("[");
136 while i != right && l != 0u {
141 s.push_str(&format!("{}={}",
143 tok_str(toks[i].clone()))[]);
152 pub enum PrintStackBreak {
158 pub struct PrintStackElem {
160 pbreak: PrintStackBreak
163 static SIZE_INFINITY: int = 0xffff;
165 pub fn mk_printer(out: Box<io::Writer+'static>, linewidth: uint) -> Printer {
166 // Yes 3, it makes the ring buffers big enough to never
168 let n: uint = 3 * linewidth;
169 debug!("mk_printer {}", linewidth);
170 let token: Vec<Token> = repeat(Eof).take(n).collect();
171 let size: Vec<int> = repeat(0i).take(n).collect();
172 let scan_stack: Vec<uint> = repeat(0u).take(n).collect();
176 margin: linewidth as int,
177 space: linewidth as int,
184 scan_stack: scan_stack,
185 scan_stack_empty: true,
188 print_stack: Vec::new(),
189 pending_indentation: 0
194 /// In case you do not have the paper, here is an explanation of what's going
197 /// There is a stream of input tokens flowing through this printer.
199 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
200 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
201 /// case every token worth buffering is 1 char long, so it's ok.
203 /// Tokens are String, Break, and Begin/End to delimit blocks.
205 /// Begin tokens can carry an offset, saying "how far to indent when you break
206 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
207 /// breaking. Consistent breaking means that after the first break, no attempt
208 /// will be made to flow subsequent breaks together onto lines. Inconsistent
209 /// is the opposite. Inconsistent breaking example would be, say:
211 /// foo(hello, there, good, friends)
213 /// breaking inconsistently to become
218 /// whereas a consistent breaking would yield:
225 /// That is, in the consistent-break blocks we value vertical alignment
226 /// more than the ability to cram stuff onto a line. But in all cases if it
227 /// can make a block a one-liner, it'll do so.
229 /// Carrying on with high-level logic:
231 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
232 /// 'right' indices denote the active portion of the ring buffer as well as
233 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
234 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
235 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
236 /// and point-in-infinite-stream senses freely.
238 /// There is a parallel ring buffer, 'size', that holds the calculated size of
239 /// each token. Why calculated? Because for Begin/End pairs, the "size"
240 /// includes everything between the pair. That is, the "size" of Begin is
241 /// actually the sum of the sizes of everything between Begin and the paired
242 /// End that follows. Since that is arbitrarily far in the future, 'size' is
243 /// being rewritten regularly while the printer runs; in fact most of the
244 /// machinery is here to work out 'size' entries on the fly (and give up when
245 /// they're so obviously over-long that "infinity" is a good enough
246 /// approximation for purposes of line breaking).
248 /// The "input side" of the printer is managed as an abstract process called
249 /// SCAN, which uses 'scan_stack', 'scan_stack_empty', 'top' and 'bottom', to
250 /// manage calculating 'size'. SCAN is, in other words, the process of
251 /// calculating 'size' entries.
253 /// The "output side" of the printer is managed by an abstract process called
254 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
255 /// do with each token/size pair it consumes as it goes. It's trying to consume
256 /// the entire buffered window, but can't output anything until the size is >=
257 /// 0 (sizes are set to negative while they're pending calculation).
259 /// So SCAN takes input and buffers tokens and pending calculations, while
260 /// PRINT gobbles up completed calculations and tokens from the buffer. The
261 /// theory is that the two can never get more than 3N tokens apart, because
262 /// once there's "obviously" too much data to fit on a line, in a size
263 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
266 /// In this implementation (following the paper, again) the SCAN process is
267 /// the method called 'pretty_print', and the 'PRINT' process is the method
270 pub out: Box<io::Writer+'static>,
272 /// Width of lines we're constrained to
274 /// Number of spaces left on line
276 /// Index of left side of input stream
278 /// Index of right side of input stream
280 /// Ring-buffer stream goes through
282 /// Ring-buffer of calculated sizes
284 /// Running size of stream "...left"
286 /// Running size of stream "...right"
288 /// Pseudo-stack, really a ring too. Holds the
289 /// primary-ring-buffers index of the Begin that started the
290 /// current block, possibly with the most recent Break after that
291 /// Begin (if there is any) on top of it. Stuff is flushed off the
292 /// bottom as it becomes irrelevant due to the primary ring-buffer
294 scan_stack: Vec<uint> ,
295 /// Top==bottom disambiguator
296 scan_stack_empty: bool,
297 /// Index of top of scan_stack
299 /// Index of bottom of scan_stack
301 /// Stack of blocks-in-progress being flushed by print
302 print_stack: Vec<PrintStackElem> ,
303 /// Buffered indentation to avoid writing trailing whitespace
304 pending_indentation: int,
308 pub fn last_token(&mut self) -> Token {
309 self.token[self.right].clone()
311 // be very careful with this!
312 pub fn replace_last_token(&mut self, t: Token) {
313 self.token[self.right] = t;
315 pub fn pretty_print(&mut self, t: Token) -> io::IoResult<()> {
316 debug!("pp ~[{},{}]", self.left, self.right);
319 if !self.scan_stack_empty {
321 let left = self.token[self.left].clone();
322 let left_size = self.size[self.left];
323 try!(self.advance_left(left, left_size));
329 if self.scan_stack_empty {
331 self.right_total = 1;
334 } else { self.advance_right(); }
335 debug!("pp Begin({})/buffer ~[{},{}]",
336 b.offset, self.left, self.right);
337 self.token[self.right] = t;
338 self.size[self.right] = -self.right_total;
339 let right = self.right;
340 self.scan_push(right);
344 if self.scan_stack_empty {
345 debug!("pp End/print ~[{},{}]", self.left, self.right);
348 debug!("pp End/buffer ~[{},{}]", self.left, self.right);
349 self.advance_right();
350 self.token[self.right] = t;
351 self.size[self.right] = -1;
352 let right = self.right;
353 self.scan_push(right);
358 if self.scan_stack_empty {
360 self.right_total = 1;
363 } else { self.advance_right(); }
364 debug!("pp Break({})/buffer ~[{},{}]",
365 b.offset, self.left, self.right);
367 let right = self.right;
368 self.scan_push(right);
369 self.token[self.right] = t;
370 self.size[self.right] = -self.right_total;
371 self.right_total += b.blank_space;
374 String(ref s, len) => {
375 if self.scan_stack_empty {
376 debug!("pp String('{}')/print ~[{},{}]",
377 *s, self.left, self.right);
378 self.print(t.clone(), len)
380 debug!("pp String('{}')/buffer ~[{},{}]",
381 *s, self.left, self.right);
382 self.advance_right();
383 self.token[self.right] = t.clone();
384 self.size[self.right] = len;
385 self.right_total += len;
391 pub fn check_stream(&mut self) -> io::IoResult<()> {
392 debug!("check_stream ~[{}, {}] with left_total={}, right_total={}",
393 self.left, self.right, self.left_total, self.right_total);
394 if self.right_total - self.left_total > self.space {
395 debug!("scan window is {}, longer than space on line ({})",
396 self.right_total - self.left_total, self.space);
397 if !self.scan_stack_empty {
398 if self.left == self.scan_stack[self.bottom] {
399 debug!("setting {} to infinity and popping", self.left);
400 let scanned = self.scan_pop_bottom();
401 self.size[scanned] = SIZE_INFINITY;
404 let left = self.token[self.left].clone();
405 let left_size = self.size[self.left];
406 try!(self.advance_left(left, left_size));
407 if self.left != self.right {
408 try!(self.check_stream());
413 pub fn scan_push(&mut self, x: uint) {
414 debug!("scan_push {}", x);
415 if self.scan_stack_empty {
416 self.scan_stack_empty = false;
419 self.top %= self.buf_len;
420 assert!((self.top != self.bottom));
422 self.scan_stack[self.top] = x;
424 pub fn scan_pop(&mut self) -> uint {
425 assert!((!self.scan_stack_empty));
426 let x = self.scan_stack[self.top];
427 if self.top == self.bottom {
428 self.scan_stack_empty = true;
430 self.top += self.buf_len - 1u; self.top %= self.buf_len;
434 pub fn scan_top(&mut self) -> uint {
435 assert!((!self.scan_stack_empty));
436 return self.scan_stack[self.top];
438 pub fn scan_pop_bottom(&mut self) -> uint {
439 assert!((!self.scan_stack_empty));
440 let x = self.scan_stack[self.bottom];
441 if self.top == self.bottom {
442 self.scan_stack_empty = true;
444 self.bottom += 1u; self.bottom %= self.buf_len;
448 pub fn advance_right(&mut self) {
450 self.right %= self.buf_len;
451 assert!((self.right != self.left));
453 pub fn advance_left(&mut self, x: Token, l: int) -> io::IoResult<()> {
454 debug!("advance_left ~[{},{}], sizeof({})={}", self.left, self.right,
457 let ret = self.print(x.clone(), l);
459 Break(b) => self.left_total += b.blank_space,
461 assert_eq!(len, l); self.left_total += len;
465 if self.left != self.right {
467 self.left %= self.buf_len;
468 let left = self.token[self.left].clone();
469 let left_size = self.size[self.left];
470 try!(self.advance_left(left, left_size));
477 pub fn check_stack(&mut self, k: int) {
478 if !self.scan_stack_empty {
479 let x = self.scan_top();
480 match self.token[x] {
483 let popped = self.scan_pop();
484 self.size[popped] = self.size[x] + self.right_total;
485 self.check_stack(k - 1);
489 // paper says + not =, but that makes no sense.
490 let popped = self.scan_pop();
491 self.size[popped] = 1;
492 self.check_stack(k + 1);
495 let popped = self.scan_pop();
496 self.size[popped] = self.size[x] + self.right_total;
504 pub fn print_newline(&mut self, amount: int) -> io::IoResult<()> {
505 debug!("NEWLINE {}", amount);
506 let ret = write!(self.out, "\n");
507 self.pending_indentation = 0;
511 pub fn indent(&mut self, amount: int) {
512 debug!("INDENT {}", amount);
513 self.pending_indentation += amount;
515 pub fn get_top(&mut self) -> PrintStackElem {
516 let print_stack = &mut self.print_stack;
517 let n = print_stack.len();
519 (*print_stack)[n - 1]
523 pbreak: Broken(Inconsistent)
527 pub fn print_str(&mut self, s: &str) -> io::IoResult<()> {
528 while self.pending_indentation > 0 {
529 try!(write!(self.out, " "));
530 self.pending_indentation -= 1;
532 write!(self.out, "{}", s)
534 pub fn print(&mut self, x: Token, l: int) -> io::IoResult<()> {
535 debug!("print {} {} (remaining line space={})", tok_str(x.clone()), l,
537 debug!("{}", buf_str(self.token.clone(),
545 let col = self.margin - self.space + b.offset;
546 debug!("print Begin -> push broken block at col {}", col);
547 self.print_stack.push(PrintStackElem {
549 pbreak: Broken(b.breaks)
552 debug!("print Begin -> push fitting block");
553 self.print_stack.push(PrintStackElem {
561 debug!("print End -> pop End");
562 let print_stack = &mut self.print_stack;
563 assert!((print_stack.len() != 0u));
564 print_stack.pop().unwrap();
568 let top = self.get_top();
571 debug!("print Break({}) in fitting block", b.blank_space);
572 self.space -= b.blank_space;
573 self.indent(b.blank_space);
576 Broken(Consistent) => {
577 debug!("print Break({}+{}) in consistent block",
578 top.offset, b.offset);
579 let ret = self.print_newline(top.offset + b.offset);
580 self.space = self.margin - (top.offset + b.offset);
583 Broken(Inconsistent) => {
585 debug!("print Break({}+{}) w/ newline in inconsistent",
586 top.offset, b.offset);
587 let ret = self.print_newline(top.offset + b.offset);
588 self.space = self.margin - (top.offset + b.offset);
591 debug!("print Break({}) w/o newline in inconsistent",
593 self.indent(b.blank_space);
594 self.space -= b.blank_space;
601 debug!("print String({})", s);
603 // assert!(l <= space);
608 // Eof should never get here.
615 // Convenience functions to talk to the printer.
618 pub fn rbox(p: &mut Printer, indent: uint, b: Breaks) -> io::IoResult<()> {
619 p.pretty_print(Begin(BeginToken {
620 offset: indent as int,
625 pub fn ibox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
626 rbox(p, indent, Inconsistent)
629 pub fn cbox(p: &mut Printer, indent: uint) -> io::IoResult<()> {
630 rbox(p, indent, Consistent)
633 pub fn break_offset(p: &mut Printer, n: uint, off: int) -> io::IoResult<()> {
634 p.pretty_print(Break(BreakToken {
636 blank_space: n as int
640 pub fn end(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(End) }
642 pub fn eof(p: &mut Printer) -> io::IoResult<()> { p.pretty_print(Eof) }
644 pub fn word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
645 p.pretty_print(String(/* bad */ wrd.to_string(), wrd.len() as int))
648 pub fn huge_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
649 p.pretty_print(String(/* bad */ wrd.to_string(), SIZE_INFINITY))
652 pub fn zero_word(p: &mut Printer, wrd: &str) -> io::IoResult<()> {
653 p.pretty_print(String(/* bad */ wrd.to_string(), 0))
656 pub fn spaces(p: &mut Printer, n: uint) -> io::IoResult<()> {
657 break_offset(p, n, 0)
660 pub fn zerobreak(p: &mut Printer) -> io::IoResult<()> {
664 pub fn space(p: &mut Printer) -> io::IoResult<()> {
668 pub fn hardbreak(p: &mut Printer) -> io::IoResult<()> {
669 spaces(p, SIZE_INFINITY as uint)
672 pub fn hardbreak_tok_offset(off: int) -> Token {
673 Break(BreakToken {offset: off, blank_space: SIZE_INFINITY})
676 pub fn hardbreak_tok() -> Token { return hardbreak_tok_offset(0); }