1 // Copyright 2015 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 // Formatting and tools for comments.
13 use std::{self, iter};
15 use syntax::codemap::Span;
18 use rewrite::RewriteContext;
19 use shape::{Indent, Shape};
20 use string::{rewrite_string, StringFormat};
21 use utils::{count_newlines, first_line_width, last_line_width};
23 fn is_custom_comment(comment: &str) -> bool {
24 if !comment.starts_with("//") {
26 } else if let Some(c) = comment.chars().nth(2) {
27 !c.is_alphanumeric() && !c.is_whitespace()
33 #[derive(Copy, Clone, PartialEq, Eq)]
34 pub enum CommentStyle<'a> {
44 fn custom_opener(s: &str) -> &str {
45 s.lines().next().map_or("", |first_line| {
48 .map_or(first_line, |space_index| &first_line[0..space_index + 1])
52 impl<'a> CommentStyle<'a> {
53 pub fn opener(&self) -> &'a str {
55 CommentStyle::DoubleSlash => "// ",
56 CommentStyle::TripleSlash => "/// ",
57 CommentStyle::Doc => "//! ",
58 CommentStyle::SingleBullet => "/* ",
59 CommentStyle::DoubleBullet => "/** ",
60 CommentStyle::Exclamation => "/*! ",
61 CommentStyle::Custom(opener) => opener,
65 pub fn closer(&self) -> &'a str {
67 CommentStyle::DoubleSlash
68 | CommentStyle::TripleSlash
69 | CommentStyle::Custom(..)
70 | CommentStyle::Doc => "",
71 CommentStyle::DoubleBullet => " **/",
72 CommentStyle::SingleBullet | CommentStyle::Exclamation => " */",
76 pub fn line_start(&self) -> &'a str {
78 CommentStyle::DoubleSlash => "// ",
79 CommentStyle::TripleSlash => "/// ",
80 CommentStyle::Doc => "//! ",
81 CommentStyle::SingleBullet | CommentStyle::Exclamation => " * ",
82 CommentStyle::DoubleBullet => " ** ",
83 CommentStyle::Custom(opener) => opener,
87 pub fn to_str_tuplet(&self) -> (&'a str, &'a str, &'a str) {
88 (self.opener(), self.closer(), self.line_start())
91 pub fn line_with_same_comment_style(&self, line: &str, normalize_comments: bool) -> bool {
93 CommentStyle::DoubleSlash | CommentStyle::TripleSlash | CommentStyle::Doc => {
94 line.trim_left().starts_with(self.line_start().trim_left())
95 || comment_style(line, normalize_comments) == *self
97 CommentStyle::DoubleBullet | CommentStyle::SingleBullet | CommentStyle::Exclamation => {
98 line.trim_left().starts_with(self.closer().trim_left())
99 || line.trim_left().starts_with(self.line_start().trim_left())
100 || comment_style(line, normalize_comments) == *self
102 CommentStyle::Custom(opener) => line.trim_left().starts_with(opener.trim_right()),
107 fn comment_style(orig: &str, normalize_comments: bool) -> CommentStyle {
108 if !normalize_comments {
109 if orig.starts_with("/**") && !orig.starts_with("/**/") {
110 CommentStyle::DoubleBullet
111 } else if orig.starts_with("/*!") {
112 CommentStyle::Exclamation
113 } else if orig.starts_with("/*") {
114 CommentStyle::SingleBullet
115 } else if orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/') {
116 CommentStyle::TripleSlash
117 } else if orig.starts_with("//!") {
119 } else if is_custom_comment(orig) {
120 CommentStyle::Custom(custom_opener(orig))
122 CommentStyle::DoubleSlash
124 } else if (orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/'))
125 || (orig.starts_with("/**") && !orig.starts_with("/**/"))
127 CommentStyle::TripleSlash
128 } else if orig.starts_with("//!") || orig.starts_with("/*!") {
130 } else if is_custom_comment(orig) {
131 CommentStyle::Custom(custom_opener(orig))
133 CommentStyle::DoubleSlash
137 /// Combine `prev_str` and `next_str` into a single `String`. `span` may contain
138 /// comments between two strings. If there are such comments, then that will be
139 /// recovered. If `allow_extend` is true and there is no comment between the two
140 /// strings, then they will be put on a single line as long as doing so does not
141 /// exceed max width.
142 pub fn combine_strs_with_missing_comments(
143 context: &RewriteContext,
149 ) -> Option<String> {
150 let mut allow_one_line = !prev_str.contains('\n') && !next_str.contains('\n');
151 let first_sep = if prev_str.is_empty() || next_str.is_empty() {
156 let mut one_line_width =
157 last_line_width(prev_str) + first_line_width(next_str) + first_sep.len();
159 let indent_str = shape.indent.to_string(context.config);
160 let missing_comment = rewrite_missing_comment(span, shape, context)?;
162 if missing_comment.is_empty() {
163 if allow_extend && prev_str.len() + first_sep.len() + next_str.len() <= shape.width {
164 return Some(format!("{}{}{}", prev_str, first_sep, next_str));
166 let sep = if prev_str.is_empty() {
169 String::from("\n") + &indent_str
171 return Some(format!("{}{}{}", prev_str, sep, next_str));
175 // We have a missing comment between the first expression and the second expression.
177 // Peek the the original source code and find out whether there is a newline between the first
178 // expression and the second expression or the missing comment. We will preserve the original
179 // layout whenever possible.
180 let original_snippet = context.snippet(span);
181 let prefer_same_line = if let Some(pos) = original_snippet.chars().position(|c| c == '/') {
182 !original_snippet[..pos].contains('\n')
184 !original_snippet.contains('\n')
187 one_line_width -= first_sep.len();
188 let first_sep = if prev_str.is_empty() || missing_comment.is_empty() {
191 let one_line_width = last_line_width(prev_str) + first_line_width(&missing_comment) + 1;
192 if prefer_same_line && one_line_width <= shape.width {
195 format!("\n{}", indent_str)
198 let second_sep = if missing_comment.is_empty() || next_str.is_empty() {
200 } else if missing_comment.starts_with("//") {
201 format!("\n{}", indent_str)
203 one_line_width += missing_comment.len() + first_sep.len() + 1;
204 allow_one_line &= !missing_comment.starts_with("//") && !missing_comment.contains('\n');
205 if prefer_same_line && allow_one_line && one_line_width <= shape.width {
208 format!("\n{}", indent_str)
213 prev_str, first_sep, missing_comment, second_sep, next_str,
217 pub fn rewrite_doc_comment(orig: &str, shape: Shape, config: &Config) -> Option<String> {
218 _rewrite_comment(orig, false, shape, config, true)
221 pub fn rewrite_comment(
226 ) -> Option<String> {
227 _rewrite_comment(orig, block_style, shape, config, false)
235 is_doc_comment: bool,
236 ) -> Option<String> {
237 // If there are lines without a starting sigil, we won't format them correctly
238 // so in that case we won't even re-align (if !config.normalize_comments()) and
239 // we should stop now.
240 let num_bare_lines = orig.lines()
241 .map(|line| line.trim())
242 .filter(|l| !(l.starts_with('*') || l.starts_with("//") || l.starts_with("/*")))
244 if num_bare_lines > 0 && !config.normalize_comments() {
245 return Some(orig.to_owned());
247 if !config.normalize_comments() && !config.wrap_comments() {
248 return light_rewrite_comment(orig, shape.indent, config, is_doc_comment);
251 identify_comment(orig, block_style, shape, config)
259 ) -> Option<String> {
260 let style = comment_style(orig, false);
261 let first_group = orig.lines()
262 .take_while(|l| style.line_with_same_comment_style(l, false))
265 let rest = orig.lines()
266 .skip(first_group.lines().count())
270 let first_group_str = rewrite_comment_inner(&first_group, block_style, style, shape, config)?;
272 Some(first_group_str)
274 identify_comment(&rest, block_style, shape, config).map(|rest_str| {
278 shape.indent.to_string(config),
285 fn rewrite_comment_inner(
291 ) -> Option<String> {
292 let (opener, closer, line_start) = if block_style {
293 CommentStyle::SingleBullet.to_str_tuplet()
295 comment_style(orig, config.normalize_comments()).to_str_tuplet()
298 let max_chars = shape
300 .checked_sub(closer.len() + opener.len())
302 let indent_str = shape.indent.to_string(config);
303 let fmt_indent = shape.indent + (opener.len() - line_start.len());
304 let mut fmt = StringFormat {
309 shape: Shape::legacy(max_chars, fmt_indent),
314 let line_breaks = count_newlines(orig.trim_right());
315 let lines = orig.lines()
317 .map(|(i, mut line)| {
320 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
321 line = line[..(line.len() - 2)].trim_right();
326 .map(|s| left_trim_comment_line(s, &style))
327 .map(|(line, has_leading_whitespace)| {
328 if orig.starts_with("/*") && line_breaks == 0 {
331 has_leading_whitespace || config.normalize_comments(),
334 (line, has_leading_whitespace || config.normalize_comments())
338 let mut result = String::with_capacity(orig.len() * 2);
339 result.push_str(opener);
340 let mut code_block_buffer = String::with_capacity(128);
341 let mut is_prev_line_multi_line = false;
342 let mut inside_code_block = false;
343 let comment_line_separator = format!("\n{}{}", indent_str, line_start);
344 let join_code_block_with_comment_line_separator = |s: &str| {
345 let mut result = String::with_capacity(s.len() + 128);
346 let mut iter = s.lines().peekable();
347 while let Some(line) = iter.next() {
348 result.push_str(line);
349 result.push_str(match iter.peek() {
350 Some(next_line) if next_line.is_empty() => comment_line_separator.trim_right(),
351 Some(..) => &comment_line_separator,
358 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
359 let is_last = i == count_newlines(orig);
361 if inside_code_block {
362 if line.starts_with("```") {
363 inside_code_block = false;
364 result.push_str(&comment_line_separator);
365 let code_block = ::format_code_block(&code_block_buffer, config)
366 .unwrap_or_else(|| code_block_buffer.to_owned());
367 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
368 code_block_buffer.clear();
369 result.push_str(&comment_line_separator);
370 result.push_str(line);
372 code_block_buffer.push_str(line);
373 code_block_buffer.push('\n');
378 inside_code_block = line.starts_with("```");
380 if result == opener {
381 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
382 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
388 } else if is_prev_line_multi_line && !line.is_empty() {
390 } else if is_last && !closer.is_empty() && line.is_empty() {
392 result.push_str(&indent_str);
394 result.push_str(&comment_line_separator);
395 if !has_leading_whitespace && result.ends_with(' ') {
401 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
402 match rewrite_string(line, &fmt, Some(max_chars)) {
404 is_prev_line_multi_line = s.contains('\n');
407 None if is_prev_line_multi_line => {
408 // We failed to put the current `line` next to the previous `line`.
409 // Remove the trailing space, then start rewrite on the next line.
411 result.push_str(&comment_line_separator);
412 fmt.shape = Shape::legacy(max_chars, fmt_indent);
413 match rewrite_string(line, &fmt, Some(max_chars)) {
415 is_prev_line_multi_line = s.contains('\n');
419 is_prev_line_multi_line = false;
420 result.push_str(line);
425 is_prev_line_multi_line = false;
426 result.push_str(line);
430 fmt.shape = if is_prev_line_multi_line {
432 let offset = 1 + last_line_width(&result) - line_start.len();
434 width: max_chars.checked_sub(offset).unwrap_or(0),
436 offset: fmt.shape.offset + offset,
439 Shape::legacy(max_chars, fmt_indent)
442 if line.is_empty() && result.ends_with(' ') && !is_last {
443 // Remove space if this is an empty comment or a doc comment.
446 result.push_str(line);
447 fmt.shape = Shape::legacy(max_chars, fmt_indent);
448 is_prev_line_multi_line = false;
452 result.push_str(closer);
453 if result == opener && result.ends_with(' ') {
461 /// Returns true if the given string MAY include URLs or alike.
462 fn has_url(s: &str) -> bool {
463 // This function may return false positive, but should get its job done in most cases.
464 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
467 /// Given the span, rewrite the missing comment inside it if available.
468 /// Note that the given span must only include comments (or leading/trailing whitespaces).
469 pub fn rewrite_missing_comment(
472 context: &RewriteContext,
473 ) -> Option<String> {
474 let missing_snippet = context.snippet(span);
475 let trimmed_snippet = missing_snippet.trim();
476 if !trimmed_snippet.is_empty() {
477 rewrite_comment(trimmed_snippet, false, shape, context.config)
483 /// Recover the missing comments in the specified span, if available.
484 /// The layout of the comments will be preserved as long as it does not break the code
485 /// and its total width does not exceed the max width.
486 pub fn recover_missing_comment_in_span(
489 context: &RewriteContext,
491 ) -> Option<String> {
492 let missing_comment = rewrite_missing_comment(span, shape, context)?;
493 if missing_comment.is_empty() {
496 let missing_snippet = context.snippet(span);
497 let pos = missing_snippet.chars().position(|c| c == '/').unwrap_or(0);
499 let total_width = missing_comment.len() + used_width + 1;
500 let force_new_line_before_comment =
501 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
502 let sep = if force_new_line_before_comment {
503 format!("\n{}", shape.indent.to_string(context.config))
507 Some(format!("{}{}", sep, missing_comment))
511 /// Trim trailing whitespaces unless they consist of two whitespaces.
512 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
513 if is_doc_comment && s.ends_with(" ")
514 && !s.chars().rev().nth(2).map_or(true, char::is_whitespace)
522 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
523 fn light_rewrite_comment(
527 is_doc_comment: bool,
528 ) -> Option<String> {
529 let lines: Vec<&str> = orig.lines()
531 // This is basically just l.trim(), but in the case that a line starts
532 // with `*` we want to leave one space before it, so it aligns with the
534 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
535 let left_trimmed = if let Some(fnw) = first_non_whitespace {
536 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
544 // Preserve markdown's double-space line break syntax.
545 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
548 Some(lines.join(&format!("\n{}", offset.to_string(config))))
551 /// Trims comment characters and possibly a single space from the left of a string.
552 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
553 /// this function returns true.
554 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
555 if line.starts_with("//! ") || line.starts_with("/// ") || line.starts_with("/*! ")
556 || line.starts_with("/** ")
559 } else if let CommentStyle::Custom(opener) = *style {
560 if line.starts_with(opener) {
561 (&line[opener.len()..], true)
563 (&line[opener.trim_right().len()..], false)
565 } else if line.starts_with("/* ") || line.starts_with("// ") || line.starts_with("//!")
566 || line.starts_with("///") || line.starts_with("** ")
567 || line.starts_with("/*!")
568 || (line.starts_with("/**") && !line.starts_with("/**/"))
570 (&line[3..], line.chars().nth(2).unwrap() == ' ')
571 } else if line.starts_with("/*") || line.starts_with("* ") || line.starts_with("//")
572 || line.starts_with("**")
574 (&line[2..], line.chars().nth(1).unwrap() == ' ')
575 } else if line.starts_with('*') {
578 (line, line.starts_with(' '))
582 pub trait FindUncommented {
583 fn find_uncommented(&self, pat: &str) -> Option<usize>;
586 impl FindUncommented for str {
587 fn find_uncommented(&self, pat: &str) -> Option<usize> {
588 let mut needle_iter = pat.chars();
589 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
590 match needle_iter.next() {
592 return Some(i - pat.len());
594 Some(c) => match kind {
595 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
597 needle_iter = pat.chars();
603 // Handle case where the pattern is a suffix of the search string
604 match needle_iter.next() {
606 None => Some(self.len() - pat.len()),
611 // Returns the first byte position after the first comment. The given string
612 // is expected to be prefixed by a comment, including delimiters.
613 // Good: "/* /* inner */ outer */ code();"
614 // Bad: "code(); // hello\n world!"
615 pub fn find_comment_end(s: &str) -> Option<usize> {
616 let mut iter = CharClasses::new(s.char_indices());
617 for (kind, (i, _c)) in &mut iter {
618 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
623 // Handle case where the comment ends at the end of s.
624 if iter.status == CharClassesStatus::Normal {
631 /// Returns true if text contains any comment.
632 pub fn contains_comment(text: &str) -> bool {
633 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
636 /// Remove trailing spaces from the specified snippet. We do not remove spaces
637 /// inside strings or comments.
638 pub fn remove_trailing_white_spaces(text: &str) -> String {
639 let mut buffer = String::with_capacity(text.len());
640 let mut space_buffer = String::with_capacity(128);
641 for (char_kind, c) in CharClasses::new(text.chars()) {
644 if char_kind == FullCodeCharKind::InString {
645 buffer.push_str(&space_buffer);
647 space_buffer.clear();
650 _ if c.is_whitespace() => {
651 space_buffer.push(c);
654 if !space_buffer.is_empty() {
655 buffer.push_str(&space_buffer);
656 space_buffer.clear();
665 pub struct CharClasses<T>
670 base: iter::Peekable<T>,
671 status: CharClassesStatus,
675 fn get_char(&self) -> char;
678 impl RichChar for char {
679 fn get_char(&self) -> char {
684 impl RichChar for (usize, char) {
685 fn get_char(&self) -> char {
690 impl RichChar for (char, usize) {
691 fn get_char(&self) -> char {
696 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
697 enum CharClassesStatus {
703 // The u32 is the nesting deepness of the comment
705 // Status when the '/' has been consumed, but not yet the '*', deepness is
706 // the new deepness (after the comment opening).
707 BlockCommentOpening(u32),
708 // Status when the '*' has been consumed, but not yet the '/', deepness is
709 // the new deepness (after the comment closing).
710 BlockCommentClosing(u32),
714 /// Distinguish between functional part of code and comments
715 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
716 pub enum CodeCharKind {
721 /// Distinguish between functional part of code and comments,
722 /// describing opening and closing of comments for ease when chunking
723 /// code from tagged characters
724 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
725 pub enum FullCodeCharKind {
727 /// The first character of a comment, there is only one for a comment (always '/')
729 /// Any character inside a comment including the second character of comment
730 /// marks ("//", "/*")
732 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
738 impl FullCodeCharKind {
739 pub fn is_comment(&self) -> bool {
741 FullCodeCharKind::StartComment
742 | FullCodeCharKind::InComment
743 | FullCodeCharKind::EndComment => true,
748 pub fn is_string(&self) -> bool {
749 *self == FullCodeCharKind::InString
752 fn to_codecharkind(&self) -> CodeCharKind {
753 if self.is_comment() {
754 CodeCharKind::Comment
761 impl<T> CharClasses<T>
766 pub fn new(base: T) -> CharClasses<T> {
768 base: base.peekable(),
769 status: CharClassesStatus::Normal,
774 impl<T> Iterator for CharClasses<T>
779 type Item = (FullCodeCharKind, T::Item);
781 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
782 let item = self.base.next()?;
783 let chr = item.get_char();
784 let mut char_kind = FullCodeCharKind::Normal;
785 self.status = match self.status {
786 CharClassesStatus::LitString => match chr {
787 '"' => CharClassesStatus::Normal,
789 char_kind = FullCodeCharKind::InString;
790 CharClassesStatus::LitStringEscape
793 char_kind = FullCodeCharKind::InString;
794 CharClassesStatus::LitString
797 CharClassesStatus::LitStringEscape => {
798 char_kind = FullCodeCharKind::InString;
799 CharClassesStatus::LitString
801 CharClassesStatus::LitChar => match chr {
802 '\\' => CharClassesStatus::LitCharEscape,
803 '\'' => CharClassesStatus::Normal,
804 _ => CharClassesStatus::LitChar,
806 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
807 CharClassesStatus::Normal => match chr {
809 char_kind = FullCodeCharKind::InString;
810 CharClassesStatus::LitString
812 '\'' => CharClassesStatus::LitChar,
813 '/' => match self.base.peek() {
814 Some(next) if next.get_char() == '*' => {
815 self.status = CharClassesStatus::BlockCommentOpening(1);
816 return Some((FullCodeCharKind::StartComment, item));
818 Some(next) if next.get_char() == '/' => {
819 self.status = CharClassesStatus::LineComment;
820 return Some((FullCodeCharKind::StartComment, item));
822 _ => CharClassesStatus::Normal,
824 _ => CharClassesStatus::Normal,
826 CharClassesStatus::BlockComment(deepness) => {
827 assert_ne!(deepness, 0);
828 self.status = match self.base.peek() {
829 Some(next) if next.get_char() == '/' && chr == '*' => {
830 CharClassesStatus::BlockCommentClosing(deepness - 1)
832 Some(next) if next.get_char() == '*' && chr == '/' => {
833 CharClassesStatus::BlockCommentOpening(deepness + 1)
835 _ => CharClassesStatus::BlockComment(deepness),
837 return Some((FullCodeCharKind::InComment, item));
839 CharClassesStatus::BlockCommentOpening(deepness) => {
840 assert_eq!(chr, '*');
841 self.status = CharClassesStatus::BlockComment(deepness);
842 return Some((FullCodeCharKind::InComment, item));
844 CharClassesStatus::BlockCommentClosing(deepness) => {
845 assert_eq!(chr, '/');
847 self.status = CharClassesStatus::Normal;
848 return Some((FullCodeCharKind::EndComment, item));
850 self.status = CharClassesStatus::BlockComment(deepness);
851 return Some((FullCodeCharKind::InComment, item));
854 CharClassesStatus::LineComment => match chr {
856 self.status = CharClassesStatus::Normal;
857 return Some((FullCodeCharKind::EndComment, item));
860 self.status = CharClassesStatus::LineComment;
861 return Some((FullCodeCharKind::InComment, item));
865 Some((char_kind, item))
869 /// Iterator over functional and commented parts of a string. Any part of a string is either
870 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
871 /// comments is functional code. Line comments contain their ending newlines.
872 struct UngroupedCommentCodeSlices<'a> {
874 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
877 impl<'a> UngroupedCommentCodeSlices<'a> {
878 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
879 UngroupedCommentCodeSlices {
881 iter: CharClasses::new(code.char_indices()).peekable(),
886 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
887 type Item = (CodeCharKind, usize, &'a str);
889 fn next(&mut self) -> Option<Self::Item> {
890 let (kind, (start_idx, _)) = self.iter.next()?;
892 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
893 // Consume all the Normal code
894 while let Some(&(char_kind, _)) = self.iter.peek() {
895 if char_kind.is_comment() {
898 let _ = self.iter.next();
901 FullCodeCharKind::StartComment => {
902 // Consume the whole comment
903 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
907 let slice = match self.iter.peek() {
908 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
909 None => &self.slice[start_idx..],
912 if kind.is_comment() {
913 CodeCharKind::Comment
923 /// Iterator over an alternating sequence of functional and commented parts of
924 /// a string. The first item is always a, possibly zero length, subslice of
925 /// functional text. Line style comments contain their ending newlines.
926 pub struct CommentCodeSlices<'a> {
928 last_slice_kind: CodeCharKind,
929 last_slice_end: usize,
932 impl<'a> CommentCodeSlices<'a> {
933 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
936 last_slice_kind: CodeCharKind::Comment,
942 impl<'a> Iterator for CommentCodeSlices<'a> {
943 type Item = (CodeCharKind, usize, &'a str);
945 fn next(&mut self) -> Option<Self::Item> {
946 if self.last_slice_end == self.slice.len() {
950 let mut sub_slice_end = self.last_slice_end;
951 let mut first_whitespace = None;
952 let subslice = &self.slice[self.last_slice_end..];
953 let mut iter = CharClasses::new(subslice.char_indices());
955 for (kind, (i, c)) in &mut iter {
956 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
957 && &subslice[..2] == "//"
958 && [' ', '\t'].contains(&c);
960 if is_comment_connector && first_whitespace.is_none() {
961 first_whitespace = Some(i);
964 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
965 let last_index = match first_whitespace {
969 sub_slice_end = self.last_slice_end + last_index;
973 if !is_comment_connector {
974 first_whitespace = None;
978 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
979 // This was the last subslice.
980 sub_slice_end = match first_whitespace {
981 Some(i) => self.last_slice_end + i,
982 None => self.slice.len(),
986 let kind = match self.last_slice_kind {
987 CodeCharKind::Comment => CodeCharKind::Normal,
988 CodeCharKind::Normal => CodeCharKind::Comment,
993 &self.slice[self.last_slice_end..sub_slice_end],
995 self.last_slice_end = sub_slice_end;
996 self.last_slice_kind = kind;
1002 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1003 /// (if it fits in the width/offset, else return None), else return `new`
1004 pub fn recover_comment_removed(
1007 context: &RewriteContext,
1008 ) -> Option<String> {
1009 let snippet = context.snippet(span);
1010 if snippet != new && changed_comment_content(snippet, &new) {
1011 // We missed some comments. Keep the original text.
1012 Some(snippet.to_owned())
1018 /// Return true if the two strings of code have the same payload of comments.
1019 /// The payload of comments is everything in the string except:
1020 /// - actual code (not comments)
1021 /// - comment start/end marks
1023 /// - '*' at the beginning of lines in block comments
1024 fn changed_comment_content(orig: &str, new: &str) -> bool {
1025 // Cannot write this as a fn since we cannot return types containing closures
1026 let code_comment_content = |code| {
1027 let slices = UngroupedCommentCodeSlices::new(code);
1029 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1030 .flat_map(|(_, _, s)| CommentReducer::new(s))
1032 let res = code_comment_content(orig).ne(code_comment_content(new));
1034 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1038 code_comment_content(orig).collect::<String>(),
1039 code_comment_content(new).collect::<String>()
1044 /// Iterator over the 'payload' characters of a comment.
1045 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1046 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1048 struct CommentReducer<'a> {
1050 at_start_line: bool,
1051 iter: std::str::Chars<'a>,
1054 impl<'a> CommentReducer<'a> {
1055 fn new(comment: &'a str) -> CommentReducer<'a> {
1056 let is_block = comment.starts_with("/*");
1057 let comment = remove_comment_header(comment);
1060 at_start_line: false, // There are no supplementary '*' on the first line
1061 iter: comment.chars(),
1066 impl<'a> Iterator for CommentReducer<'a> {
1068 fn next(&mut self) -> Option<Self::Item> {
1070 let mut c = self.iter.next()?;
1071 if self.is_block && self.at_start_line {
1072 while c.is_whitespace() {
1073 c = self.iter.next()?;
1075 // Ignore leading '*'
1077 c = self.iter.next()?;
1079 } else if c == '\n' {
1080 self.at_start_line = true;
1082 if !c.is_whitespace() {
1089 fn remove_comment_header(comment: &str) -> &str {
1090 if comment.starts_with("///") || comment.starts_with("//!") {
1092 } else if comment.starts_with("//") {
1094 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1095 || comment.starts_with("/*!")
1097 &comment[3..comment.len() - 2]
1100 comment.starts_with("/*"),
1101 format!("string '{}' is not a comment", comment)
1103 &comment[2..comment.len() - 2]
1109 use super::{contains_comment, rewrite_comment, CharClasses, CodeCharKind, CommentCodeSlices,
1110 FindUncommented, FullCodeCharKind};
1111 use shape::{Indent, Shape};
1115 let mut iter = CharClasses::new("//\n\n".chars());
1117 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1118 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1119 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1120 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1121 assert_eq!(None, iter.next());
1125 fn comment_code_slices() {
1126 let input = "code(); /* test */ 1 + 1";
1127 let mut iter = CommentCodeSlices::new(input);
1129 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1131 (CodeCharKind::Comment, 8, "/* test */"),
1132 iter.next().unwrap()
1134 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1135 assert_eq!(None, iter.next());
1139 fn comment_code_slices_two() {
1140 let input = "// comment\n test();";
1141 let mut iter = CommentCodeSlices::new(input);
1143 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1145 (CodeCharKind::Comment, 0, "// comment\n"),
1146 iter.next().unwrap()
1149 (CodeCharKind::Normal, 11, " test();"),
1150 iter.next().unwrap()
1152 assert_eq!(None, iter.next());
1156 fn comment_code_slices_three() {
1157 let input = "1 // comment\n // comment2\n\n";
1158 let mut iter = CommentCodeSlices::new(input);
1160 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1162 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1163 iter.next().unwrap()
1165 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1166 assert_eq!(None, iter.next());
1170 #[cfg_attr(rustfmt, rustfmt_skip)]
1171 fn format_comments() {
1172 let mut config: ::config::Config = Default::default();
1173 config.set().wrap_comments(true);
1174 config.set().normalize_comments(true);
1176 let comment = rewrite_comment(" //test",
1178 Shape::legacy(100, Indent::new(0, 100)),
1180 assert_eq!("/* test */", comment);
1182 let comment = rewrite_comment("// comment on a",
1184 Shape::legacy(10, Indent::empty()),
1186 assert_eq!("// comment\n// on a", comment);
1188 let comment = rewrite_comment("// A multi line comment\n // between args.",
1190 Shape::legacy(60, Indent::new(0, 12)),
1192 assert_eq!("// A multi line comment\n // between args.", comment);
1194 let input = "// comment";
1197 let comment = rewrite_comment(input,
1199 Shape::legacy(9, Indent::new(0, 69)),
1201 assert_eq!(expected, comment);
1203 let comment = rewrite_comment("/* trimmed */",
1205 Shape::legacy(100, Indent::new(0, 100)),
1207 assert_eq!("/* trimmed */", comment);
1210 // This is probably intended to be a non-test fn, but it is not used. I'm
1211 // keeping it around unless it helps us test stuff.
1212 fn uncommented(text: &str) -> String {
1213 CharClasses::new(text.chars())
1214 .filter_map(|(s, c)| match s {
1215 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1222 fn test_uncommented() {
1223 assert_eq!(&uncommented("abc/*...*/"), "abc");
1225 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1228 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1232 fn test_contains_comment() {
1233 assert_eq!(contains_comment("abc"), false);
1234 assert_eq!(contains_comment("abc // qsdf"), true);
1235 assert_eq!(contains_comment("abc /* kqsdf"), true);
1236 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1240 fn test_find_uncommented() {
1241 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1242 assert_eq!(expected, haystack.find_uncommented(needle));
1245 check("/*/ */test", "test", Some(6));
1246 check("//test\ntest", "test", Some(7));
1247 check("/* comment only */", "whatever", None);
1249 "/* comment */ some text /* more commentary */ result",
1253 check("sup // sup", "p", Some(2));
1254 check("sup", "x", None);
1255 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1256 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1257 check("hel/*lohello*/lo", "hello", None);
1258 check("acb", "ab", None);
1259 check(",/*A*/ ", ",", Some(0));
1260 check("abc", "abc", Some(0));
1261 check("/* abc */", "abc", None);
1262 check("/**/abc/* */", "abc", Some(4));
1263 check("\"/* abc */\"", "abc", Some(4));
1264 check("\"/* abc", "abc", Some(4));