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
15 //! STAN-CS-79-770: "Pretty Printing", by Derek C. Oppen.
16 //! Stanford Department of Computer Science, 1979.
19 //! The algorithm's aim is to break a stream into as few lines as possible
20 //! while respecting the indentation-consistency requirements of the enclosing
21 //! block, and avoiding breaking at silly places on block boundaries, for
22 //! example, between "x" and ")" in "x)".
24 //! I am implementing this algorithm because it comes with 20 pages of
25 //! documentation explaining its theory, and because it addresses the set of
26 //! concerns I've seen other pretty-printers fall down on. Weirdly. Even though
27 //! it's 32 years old. What can I say?
29 //! Despite some redundancies and quirks in the way it's implemented in that
30 //! paper, I've opted to keep the implementation here as similar as I can,
31 //! changing only what was blatantly wrong, a typo, or sufficiently
32 //! non-idiomatic rust that it really stuck out.
34 //! In particular you'll see a certain amount of churn related to INTEGER vs.
35 //! CARDINAL in the Mesa implementation. Mesa apparently interconverts the two
36 //! somewhat readily? In any case, I've used usize for indices-in-buffers and
37 //! ints for character-sizes-and-indentation-offsets. This respects the need
38 //! for ints to "go negative" while carrying a pending-calculation balance, and
39 //! helps differentiate all the numbers flying around internally (slightly).
41 //! I also inverted the indentation arithmetic used in the print stack, since
42 //! the Mesa implementation (somewhat randomly) stores the offset on the print
43 //! stack in terms of margin-col rather than col itself. I store col.
45 //! I also implemented a small change in the String token, in that I store an
46 //! explicit length for the string. For most tokens this is just the length of
47 //! the accompanying string. But it's necessary to permit it to differ, for
48 //! encoding things that are supposed to "go on their own line" -- certain
49 //! classes of comment and blank-line -- where relying on adjacent
50 //! hardbreak-like Break tokens with long blankness indication doesn't actually
51 //! work. To see why, consider when there is a "thing that should be on its own
52 //! line" between two long blocks, say functions. If you put a hardbreak after
53 //! each function (or before each) and the breaking algorithm decides to break
54 //! there anyways (because the functions themselves are long) you wind up with
55 //! extra blank lines. If you don't put hardbreaks you can wind up with the
56 //! "thing which should be on its own line" not getting its own line in the
57 //! rare case of "really small functions" or such. This re-occurs with comments
58 //! and explicit blank lines. So in those cases we use a string with a payload
59 //! we want isolated to a line and an explicit length that's huge, surrounded
60 //! by two zero-length breaks. The algorithm will try its best to fit it on a
61 //! line (which it can't) and so naturally place the content on its own line to
62 //! avoid combining it with other lines and making matters even worse.
64 use std
::collections
::VecDeque
;
68 #[derive(Clone, Copy, PartialEq)]
74 #[derive(Clone, Copy)]
75 pub struct BreakToken
{
80 #[derive(Clone, Copy)]
81 pub struct BeginToken
{
88 String(String
, isize),
96 pub fn is_eof(&self) -> bool
{
103 pub fn is_hardbreak_tok(&self) -> bool
{
105 Token
::Break(BreakToken
{
108 }) if bs
== SIZE_INFINITY
=>
116 impl fmt
::Display
for Token
{
117 fn fmt(&self, f
: &mut fmt
::Formatter
) -> fmt
::Result
{
119 Token
::String(ref s
, len
) => write
!(f
, "STR({},{})", s
, len
),
120 Token
::Break(_
) => f
.write_str("BREAK"),
121 Token
::Begin(_
) => f
.write_str("BEGIN"),
122 Token
::End
=> f
.write_str("END"),
123 Token
::Eof
=> f
.write_str("EOF"),
128 fn buf_str(buf
: &[BufEntry
], left
: usize, right
: usize, lim
: usize) -> String
{
132 let mut s
= String
::from("[");
133 while i
!= right
&& l
!= 0 {
138 s
.push_str(&format
!("{}={}", buf
[i
].size
, &buf
[i
].token
));
146 #[derive(Copy, Clone)]
147 pub enum PrintStackBreak
{
152 #[derive(Copy, Clone)]
153 pub struct PrintStackElem
{
155 pbreak
: PrintStackBreak
158 const SIZE_INFINITY
: isize = 0xffff;
160 pub fn mk_printer
<'a
>(out
: Box
<io
::Write
+'a
>, linewidth
: usize) -> Printer
<'a
> {
161 // Yes 55, it makes the ring buffers big enough to never fall behind.
162 let n
: usize = 55 * linewidth
;
163 debug
!("mk_printer {}", linewidth
);
167 margin
: linewidth
as isize,
168 space
: linewidth
as isize,
171 buf
: vec
![BufEntry { token: Token::Eof, size: 0 }
; n
],
174 scan_stack
: VecDeque
::new(),
175 print_stack
: Vec
::new(),
176 pending_indentation
: 0
181 /// In case you do not have the paper, here is an explanation of what's going
184 /// There is a stream of input tokens flowing through this printer.
186 /// The printer buffers up to 3N tokens inside itself, where N is linewidth.
187 /// Yes, linewidth is chars and tokens are multi-char, but in the worst
188 /// case every token worth buffering is 1 char long, so it's ok.
190 /// Tokens are String, Break, and Begin/End to delimit blocks.
192 /// Begin tokens can carry an offset, saying "how far to indent when you break
193 /// inside here", as well as a flag indicating "consistent" or "inconsistent"
194 /// breaking. Consistent breaking means that after the first break, no attempt
195 /// will be made to flow subsequent breaks together onto lines. Inconsistent
196 /// is the opposite. Inconsistent breaking example would be, say:
198 /// foo(hello, there, good, friends)
200 /// breaking inconsistently to become
205 /// whereas a consistent breaking would yield:
212 /// That is, in the consistent-break blocks we value vertical alignment
213 /// more than the ability to cram stuff onto a line. But in all cases if it
214 /// can make a block a one-liner, it'll do so.
216 /// Carrying on with high-level logic:
218 /// The buffered tokens go through a ring-buffer, 'tokens'. The 'left' and
219 /// 'right' indices denote the active portion of the ring buffer as well as
220 /// describing hypothetical points-in-the-infinite-stream at most 3N tokens
221 /// apart (i.e. "not wrapped to ring-buffer boundaries"). The paper will switch
222 /// between using 'left' and 'right' terms to denote the wrapped-to-ring-buffer
223 /// and point-in-infinite-stream senses freely.
225 /// There is a parallel ring buffer, 'size', that holds the calculated size of
226 /// each token. Why calculated? Because for Begin/End pairs, the "size"
227 /// includes everything between the pair. That is, the "size" of Begin is
228 /// actually the sum of the sizes of everything between Begin and the paired
229 /// End that follows. Since that is arbitrarily far in the future, 'size' is
230 /// being rewritten regularly while the printer runs; in fact most of the
231 /// machinery is here to work out 'size' entries on the fly (and give up when
232 /// they're so obviously over-long that "infinity" is a good enough
233 /// approximation for purposes of line breaking).
235 /// The "input side" of the printer is managed as an abstract process called
236 /// SCAN, which uses 'scan_stack', to manage calculating 'size'. SCAN is, in
237 /// other words, the process of calculating 'size' entries.
239 /// The "output side" of the printer is managed by an abstract process called
240 /// PRINT, which uses 'print_stack', 'margin' and 'space' to figure out what to
241 /// do with each token/size pair it consumes as it goes. It's trying to consume
242 /// the entire buffered window, but can't output anything until the size is >=
243 /// 0 (sizes are set to negative while they're pending calculation).
245 /// So SCAN takes input and buffers tokens and pending calculations, while
246 /// PRINT gobbles up completed calculations and tokens from the buffer. The
247 /// theory is that the two can never get more than 3N tokens apart, because
248 /// once there's "obviously" too much data to fit on a line, in a size
249 /// calculation, SCAN will write "infinity" to the size and let PRINT consume
252 /// In this implementation (following the paper, again) the SCAN process is
253 /// the method called 'pretty_print', and the 'PRINT' process is the method
255 pub struct Printer
<'a
> {
256 pub out
: Box
<io
::Write
+'a
>,
258 /// Width of lines we're constrained to
260 /// Number of spaces left on line
262 /// Index of left side of input stream
264 /// Index of right side of input stream
266 /// Ring-buffer of tokens and calculated sizes
268 /// Running size of stream "...left"
270 /// Running size of stream "...right"
272 /// Pseudo-stack, really a ring too. Holds the
273 /// primary-ring-buffers index of the Begin that started the
274 /// current block, possibly with the most recent Break after that
275 /// Begin (if there is any) on top of it. Stuff is flushed off the
276 /// bottom as it becomes irrelevant due to the primary ring-buffer
278 scan_stack
: VecDeque
<usize>,
279 /// Stack of blocks-in-progress being flushed by print
280 print_stack
: Vec
<PrintStackElem
> ,
281 /// Buffered indentation to avoid writing trailing whitespace
282 pending_indentation
: isize,
291 impl<'a
> Printer
<'a
> {
292 pub fn last_token(&mut self) -> Token
{
293 self.buf
[self.right
].token
.clone()
295 // be very careful with this!
296 pub fn replace_last_token(&mut self, t
: Token
) {
297 self.buf
[self.right
].token
= t
;
299 pub fn pretty_print(&mut self, token
: Token
) -> io
::Result
<()> {
300 debug
!("pp Vec<{},{}>", self.left
, self.right
);
303 if !self.scan_stack
.is_empty() {
305 self.advance_left()?
;
311 if self.scan_stack
.is_empty() {
313 self.right_total
= 1;
316 } else { self.advance_right(); }
317 debug
!("pp Begin({})/buffer Vec<{},{}>",
318 b
.offset
, self.left
, self.right
);
319 self.buf
[self.right
] = BufEntry { token: token, size: -self.right_total }
;
320 let right
= self.right
;
321 self.scan_push(right
);
325 if self.scan_stack
.is_empty() {
326 debug
!("pp End/print Vec<{},{}>", self.left
, self.right
);
329 debug
!("pp End/buffer Vec<{},{}>", self.left
, self.right
);
330 self.advance_right();
331 self.buf
[self.right
] = BufEntry { token: token, size: -1 }
;
332 let right
= self.right
;
333 self.scan_push(right
);
338 if self.scan_stack
.is_empty() {
340 self.right_total
= 1;
343 } else { self.advance_right(); }
344 debug
!("pp Break({})/buffer Vec<{},{}>",
345 b
.offset
, self.left
, self.right
);
347 let right
= self.right
;
348 self.scan_push(right
);
349 self.buf
[self.right
] = BufEntry { token: token, size: -self.right_total }
;
350 self.right_total
+= b
.blank_space
;
353 Token
::String(s
, len
) => {
354 if self.scan_stack
.is_empty() {
355 debug
!("pp String('{}')/print Vec<{},{}>",
356 s
, self.left
, self.right
);
357 self.print(Token
::String(s
, len
), len
)
359 debug
!("pp String('{}')/buffer Vec<{},{}>",
360 s
, self.left
, self.right
);
361 self.advance_right();
362 self.buf
[self.right
] = BufEntry { token: Token::String(s, len), size: len }
;
363 self.right_total
+= len
;
369 pub fn check_stream(&mut self) -> io
::Result
<()> {
370 debug
!("check_stream Vec<{}, {}> with left_total={}, right_total={}",
371 self.left
, self.right
, self.left_total
, self.right_total
);
372 if self.right_total
- self.left_total
> self.space
{
373 debug
!("scan window is {}, longer than space on line ({})",
374 self.right_total
- self.left_total
, self.space
);
375 if Some(&self.left
) == self.scan_stack
.back() {
376 debug
!("setting {} to infinity and popping", self.left
);
377 let scanned
= self.scan_pop_bottom();
378 self.buf
[scanned
].size
= SIZE_INFINITY
;
380 self.advance_left()?
;
381 if self.left
!= self.right
{
382 self.check_stream()?
;
387 pub fn scan_push(&mut self, x
: usize) {
388 debug
!("scan_push {}", x
);
389 self.scan_stack
.push_front(x
);
391 pub fn scan_pop(&mut self) -> usize {
392 self.scan_stack
.pop_front().unwrap()
394 pub fn scan_top(&mut self) -> usize {
395 *self.scan_stack
.front().unwrap()
397 pub fn scan_pop_bottom(&mut self) -> usize {
398 self.scan_stack
.pop_back().unwrap()
400 pub fn advance_right(&mut self) {
402 self.right
%= self.buf_len
;
403 assert
!(self.right
!= self.left
);
405 pub fn advance_left(&mut self) -> io
::Result
<()> {
406 debug
!("advance_left Vec<{},{}>, sizeof({})={}", self.left
, self.right
,
407 self.left
, self.buf
[self.left
].size
);
409 let mut left_size
= self.buf
[self.left
].size
;
411 while left_size
>= 0 {
412 let left
= self.buf
[self.left
].token
.clone();
414 let len
= match left
{
415 Token
::Break(b
) => b
.blank_space
,
416 Token
::String(_
, len
) => {
417 assert_eq
!(len
, left_size
);
423 self.print(left
, left_size
)?
;
425 self.left_total
+= len
;
427 if self.left
== self.right
{
432 self.left
%= self.buf_len
;
434 left_size
= self.buf
[self.left
].size
;
439 pub fn check_stack(&mut self, k
: isize) {
440 if !self.scan_stack
.is_empty() {
441 let x
= self.scan_top();
442 match self.buf
[x
].token
{
445 let popped
= self.scan_pop();
446 self.buf
[popped
].size
= self.buf
[x
].size
+ self.right_total
;
447 self.check_stack(k
- 1);
451 // paper says + not =, but that makes no sense.
452 let popped
= self.scan_pop();
453 self.buf
[popped
].size
= 1;
454 self.check_stack(k
+ 1);
457 let popped
= self.scan_pop();
458 self.buf
[popped
].size
= self.buf
[x
].size
+ self.right_total
;
466 pub fn print_newline(&mut self, amount
: isize) -> io
::Result
<()> {
467 debug
!("NEWLINE {}", amount
);
468 let ret
= write
!(self.out
, "\n");
469 self.pending_indentation
= 0;
473 pub fn indent(&mut self, amount
: isize) {
474 debug
!("INDENT {}", amount
);
475 self.pending_indentation
+= amount
;
477 pub fn get_top(&mut self) -> PrintStackElem
{
478 match self.print_stack
.last() {
480 None
=> PrintStackElem
{
482 pbreak
: PrintStackBreak
::Broken(Breaks
::Inconsistent
)
486 pub fn print_str(&mut self, s
: &str) -> io
::Result
<()> {
487 while self.pending_indentation
> 0 {
488 write
!(self.out
, " ")?
;
489 self.pending_indentation
-= 1;
491 write
!(self.out
, "{}", s
)
493 pub fn print(&mut self, token
: Token
, l
: isize) -> io
::Result
<()> {
494 debug
!("print {} {} (remaining line space={})", token
, l
,
496 debug
!("{}", buf_str(&self.buf
,
503 let col
= self.margin
- self.space
+ b
.offset
;
504 debug
!("print Begin -> push broken block at col {}", col
);
505 self.print_stack
.push(PrintStackElem
{
507 pbreak
: PrintStackBreak
::Broken(b
.breaks
)
510 debug
!("print Begin -> push fitting block");
511 self.print_stack
.push(PrintStackElem
{
513 pbreak
: PrintStackBreak
::Fits
519 debug
!("print End -> pop End");
520 let print_stack
= &mut self.print_stack
;
521 assert
!(!print_stack
.is_empty());
522 print_stack
.pop().unwrap();
526 let top
= self.get_top();
528 PrintStackBreak
::Fits
=> {
529 debug
!("print Break({}) in fitting block", b
.blank_space
);
530 self.space
-= b
.blank_space
;
531 self.indent(b
.blank_space
);
534 PrintStackBreak
::Broken(Breaks
::Consistent
) => {
535 debug
!("print Break({}+{}) in consistent block",
536 top
.offset
, b
.offset
);
537 let ret
= self.print_newline(top
.offset
+ b
.offset
);
538 self.space
= self.margin
- (top
.offset
+ b
.offset
);
541 PrintStackBreak
::Broken(Breaks
::Inconsistent
) => {
543 debug
!("print Break({}+{}) w/ newline in inconsistent",
544 top
.offset
, b
.offset
);
545 let ret
= self.print_newline(top
.offset
+ b
.offset
);
546 self.space
= self.margin
- (top
.offset
+ b
.offset
);
549 debug
!("print Break({}) w/o newline in inconsistent",
551 self.indent(b
.blank_space
);
552 self.space
-= b
.blank_space
;
558 Token
::String(ref s
, len
) => {
559 debug
!("print String({})", s
);
561 // assert!(l <= space);
566 // Eof should never get here.
573 // Convenience functions to talk to the printer.
576 pub fn rbox(p
: &mut Printer
, indent
: usize, b
: Breaks
) -> io
::Result
<()> {
577 p
.pretty_print(Token
::Begin(BeginToken
{
578 offset
: indent
as isize,
583 pub fn ibox(p
: &mut Printer
, indent
: usize) -> io
::Result
<()> {
584 rbox(p
, indent
, Breaks
::Inconsistent
)
587 pub fn cbox(p
: &mut Printer
, indent
: usize) -> io
::Result
<()> {
588 rbox(p
, indent
, Breaks
::Consistent
)
591 pub fn break_offset(p
: &mut Printer
, n
: usize, off
: isize) -> io
::Result
<()> {
592 p
.pretty_print(Token
::Break(BreakToken
{
594 blank_space
: n
as isize
598 pub fn end(p
: &mut Printer
) -> io
::Result
<()> {
599 p
.pretty_print(Token
::End
)
602 pub fn eof(p
: &mut Printer
) -> io
::Result
<()> {
603 p
.pretty_print(Token
::Eof
)
606 pub fn word(p
: &mut Printer
, wrd
: &str) -> io
::Result
<()> {
607 p
.pretty_print(Token
::String(wrd
.to_string(), wrd
.len() as isize))
610 pub fn huge_word(p
: &mut Printer
, wrd
: &str) -> io
::Result
<()> {
611 p
.pretty_print(Token
::String(wrd
.to_string(), SIZE_INFINITY
))
614 pub fn zero_word(p
: &mut Printer
, wrd
: &str) -> io
::Result
<()> {
615 p
.pretty_print(Token
::String(wrd
.to_string(), 0))
618 pub fn spaces(p
: &mut Printer
, n
: usize) -> io
::Result
<()> {
619 break_offset(p
, n
, 0)
622 pub fn zerobreak(p
: &mut Printer
) -> io
::Result
<()> {
626 pub fn space(p
: &mut Printer
) -> io
::Result
<()> {
630 pub fn hardbreak(p
: &mut Printer
) -> io
::Result
<()> {
631 spaces(p
, SIZE_INFINITY
as usize)
634 pub fn hardbreak_tok_offset(off
: isize) -> Token
{
635 Token
::Break(BreakToken {offset: off, blank_space: SIZE_INFINITY}
)
638 pub fn hardbreak_tok() -> Token
{
639 hardbreak_tok_offset(0)