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 pub fn combine_strs_with_missing_comments(
138 context: &RewriteContext,
144 ) -> Option<String> {
145 let mut allow_one_line = !prev_str.contains('\n') && !next_str.contains('\n');
146 let first_sep = if prev_str.is_empty() || next_str.is_empty() {
151 let mut one_line_width =
152 last_line_width(prev_str) + first_line_width(next_str) + first_sep.len();
154 let indent_str = shape.indent.to_string(context.config);
155 let missing_comment = rewrite_missing_comment(span, shape, context)?;
157 if missing_comment.is_empty() {
158 if allow_extend && prev_str.len() + first_sep.len() + next_str.len() <= shape.width {
159 return Some(format!("{}{}{}", prev_str, first_sep, next_str));
161 let sep = if prev_str.is_empty() {
164 String::from("\n") + &indent_str
166 return Some(format!("{}{}{}", prev_str, sep, next_str));
170 // We have a missing comment between the first expression and the second expression.
172 // Peek the the original source code and find out whether there is a newline between the first
173 // expression and the second expression or the missing comment. We will preserve the original
174 // layout whenever possible.
175 let original_snippet = context.snippet(span);
176 let prefer_same_line = if let Some(pos) = original_snippet.chars().position(|c| c == '/') {
177 !original_snippet[..pos].contains('\n')
179 !original_snippet.contains('\n')
182 one_line_width -= first_sep.len();
183 let first_sep = if prev_str.is_empty() || missing_comment.is_empty() {
186 let one_line_width = last_line_width(prev_str) + first_line_width(&missing_comment) + 1;
187 if prefer_same_line && one_line_width <= shape.width {
190 format!("\n{}", indent_str)
193 let second_sep = if missing_comment.is_empty() || next_str.is_empty() {
195 } else if missing_comment.starts_with("//") {
196 format!("\n{}", indent_str)
198 one_line_width += missing_comment.len() + first_sep.len() + 1;
199 allow_one_line &= !missing_comment.starts_with("//") && !missing_comment.contains('\n');
200 if prefer_same_line && allow_one_line && one_line_width <= shape.width {
203 format!("\n{}", indent_str)
208 prev_str, first_sep, missing_comment, second_sep, next_str,
212 pub fn rewrite_comment(
217 ) -> Option<String> {
218 // If there are lines without a starting sigil, we won't format them correctly
219 // so in that case we won't even re-align (if !config.normalize_comments()) and
220 // we should stop now.
221 let num_bare_lines = orig.lines()
222 .map(|line| line.trim())
223 .filter(|l| !(l.starts_with('*') || l.starts_with("//") || l.starts_with("/*")))
225 if num_bare_lines > 0 && !config.normalize_comments() {
226 return Some(orig.to_owned());
228 if !config.normalize_comments() && !config.wrap_comments() {
229 return light_rewrite_comment(orig, shape.indent, config);
232 identify_comment(orig, block_style, shape, config)
240 ) -> Option<String> {
241 let style = comment_style(orig, false);
242 let first_group = orig.lines()
243 .take_while(|l| style.line_with_same_comment_style(l, false))
246 let rest = orig.lines()
247 .skip(first_group.lines().count())
251 let first_group_str = rewrite_comment_inner(&first_group, block_style, style, shape, config)?;
253 Some(first_group_str)
255 identify_comment(&rest, block_style, shape, config).map(|rest_str| {
259 shape.indent.to_string(config),
266 fn rewrite_comment_inner(
272 ) -> Option<String> {
273 let (opener, closer, line_start) = if block_style {
274 CommentStyle::SingleBullet.to_str_tuplet()
276 comment_style(orig, config.normalize_comments()).to_str_tuplet()
279 let max_chars = shape
281 .checked_sub(closer.len() + opener.len())
283 let indent_str = shape.indent.to_string(config);
284 let fmt_indent = shape.indent + (opener.len() - line_start.len());
285 let mut fmt = StringFormat {
288 line_start: line_start,
290 shape: Shape::legacy(max_chars, fmt_indent),
295 let line_breaks = count_newlines(orig.trim_right());
296 let lines = orig.lines()
298 .map(|(i, mut line)| {
301 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
302 line = line[..(line.len() - 2)].trim_right();
307 .map(|s| left_trim_comment_line(s, &style))
308 .map(|(line, has_leading_whitespace)| {
309 if orig.starts_with("/*") && line_breaks == 0 {
312 has_leading_whitespace || config.normalize_comments(),
315 (line, has_leading_whitespace || config.normalize_comments())
319 let mut result = String::with_capacity(orig.len() * 2);
320 result.push_str(opener);
321 let mut code_block_buffer = String::with_capacity(128);
322 let mut is_prev_line_multi_line = false;
323 let mut inside_code_block = false;
324 let comment_line_separator = format!("\n{}{}", indent_str, line_start);
325 let join_code_block_with_comment_line_separator = |s: &str| {
326 let mut result = String::with_capacity(s.len() + 128);
327 let mut iter = s.lines().peekable();
328 while let Some(line) = iter.next() {
329 result.push_str(line);
330 result.push_str(match iter.peek() {
331 Some(ref next_line) if next_line.is_empty() => comment_line_separator.trim_right(),
332 Some(..) => &comment_line_separator,
339 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
340 let is_last = i == count_newlines(orig);
342 if inside_code_block {
343 if line.starts_with("```") {
344 inside_code_block = false;
345 result.push_str(&comment_line_separator);
346 let code_block = ::format_code_block(&code_block_buffer, config)
347 .unwrap_or_else(|| code_block_buffer.to_owned());
348 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
349 code_block_buffer.clear();
350 result.push_str(&comment_line_separator);
351 result.push_str(line);
353 code_block_buffer.push_str(line);
354 code_block_buffer.push('\n');
359 inside_code_block = line.starts_with("```");
361 if result == opener {
362 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
363 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
369 } else if is_prev_line_multi_line && !line.is_empty() {
371 } else if is_last && !closer.is_empty() && line.is_empty() {
373 result.push_str(&indent_str);
375 result.push_str(&comment_line_separator);
376 if !has_leading_whitespace && result.ends_with(' ') {
382 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
383 match rewrite_string(line, &fmt, Some(max_chars)) {
385 is_prev_line_multi_line = s.contains('\n');
388 None if is_prev_line_multi_line => {
389 // We failed to put the current `line` next to the previous `line`.
390 // Remove the trailing space, then start rewrite on the next line.
392 result.push_str(&comment_line_separator);
393 fmt.shape = Shape::legacy(max_chars, fmt_indent);
394 match rewrite_string(line, &fmt, Some(max_chars)) {
396 is_prev_line_multi_line = s.contains('\n');
400 is_prev_line_multi_line = false;
401 result.push_str(line);
406 is_prev_line_multi_line = false;
407 result.push_str(line);
411 fmt.shape = if is_prev_line_multi_line {
413 let offset = 1 + last_line_width(&result) - line_start.len();
415 width: max_chars.checked_sub(offset).unwrap_or(0),
417 offset: fmt.shape.offset + offset,
420 Shape::legacy(max_chars, fmt_indent)
423 if line.is_empty() && result.ends_with(' ') && !is_last {
424 // Remove space if this is an empty comment or a doc comment.
427 result.push_str(line);
428 fmt.shape = Shape::legacy(max_chars, fmt_indent);
429 is_prev_line_multi_line = false;
433 result.push_str(closer);
434 if result == opener && result.ends_with(' ') {
442 /// Returns true if the given string MAY include URLs or alike.
443 fn has_url(s: &str) -> bool {
444 // This function may return false positive, but should get its job done in most cases.
445 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
448 /// Given the span, rewrite the missing comment inside it if available.
449 /// Note that the given span must only include comments (or leading/trailing whitespaces).
450 pub fn rewrite_missing_comment(
453 context: &RewriteContext,
454 ) -> Option<String> {
455 let missing_snippet = context.snippet(span);
456 let trimmed_snippet = missing_snippet.trim();
457 if !trimmed_snippet.is_empty() {
458 rewrite_comment(trimmed_snippet, false, shape, context.config)
464 /// Recover the missing comments in the specified span, if available.
465 /// The layout of the comments will be preserved as long as it does not break the code
466 /// and its total width does not exceed the max width.
467 pub fn recover_missing_comment_in_span(
470 context: &RewriteContext,
472 ) -> Option<String> {
473 let missing_comment = rewrite_missing_comment(span, shape, context)?;
474 if missing_comment.is_empty() {
477 let missing_snippet = context.snippet(span);
478 let pos = missing_snippet.chars().position(|c| c == '/').unwrap_or(0);
480 let total_width = missing_comment.len() + used_width + 1;
481 let force_new_line_before_comment =
482 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
483 let sep = if force_new_line_before_comment {
484 format!("\n{}", shape.indent.to_string(context.config))
488 Some(format!("{}{}", sep, missing_comment))
492 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
493 fn light_rewrite_comment(orig: &str, offset: Indent, config: &Config) -> Option<String> {
494 let lines: Vec<&str> = orig.lines()
496 // This is basically just l.trim(), but in the case that a line starts
497 // with `*` we want to leave one space before it, so it aligns with the
499 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
500 if let Some(fnw) = first_non_whitespace {
501 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
511 Some(lines.join(&format!("\n{}", offset.to_string(config))))
514 /// Trims comment characters and possibly a single space from the left of a string.
515 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
516 /// this function returns true.
517 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
518 if line.starts_with("//! ") || line.starts_with("/// ") || line.starts_with("/*! ")
519 || line.starts_with("/** ")
522 } else if let CommentStyle::Custom(opener) = *style {
523 if line.starts_with(opener) {
524 (&line[opener.len()..], true)
526 (&line[opener.trim_right().len()..], false)
528 } else if line.starts_with("/* ") || line.starts_with("// ") || line.starts_with("//!")
529 || line.starts_with("///") || line.starts_with("** ")
530 || line.starts_with("/*!")
531 || (line.starts_with("/**") && !line.starts_with("/**/"))
533 (&line[3..], line.chars().nth(2).unwrap() == ' ')
534 } else if line.starts_with("/*") || line.starts_with("* ") || line.starts_with("//")
535 || line.starts_with("**")
537 (&line[2..], line.chars().nth(1).unwrap() == ' ')
538 } else if line.starts_with('*') {
541 (line, line.starts_with(' '))
545 pub trait FindUncommented {
546 fn find_uncommented(&self, pat: &str) -> Option<usize>;
549 impl FindUncommented for str {
550 fn find_uncommented(&self, pat: &str) -> Option<usize> {
551 let mut needle_iter = pat.chars();
552 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
553 match needle_iter.next() {
555 return Some(i - pat.len());
557 Some(c) => match kind {
558 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
560 needle_iter = pat.chars();
566 // Handle case where the pattern is a suffix of the search string
567 match needle_iter.next() {
569 None => Some(self.len() - pat.len()),
574 // Returns the first byte position after the first comment. The given string
575 // is expected to be prefixed by a comment, including delimiters.
576 // Good: "/* /* inner */ outer */ code();"
577 // Bad: "code(); // hello\n world!"
578 pub fn find_comment_end(s: &str) -> Option<usize> {
579 let mut iter = CharClasses::new(s.char_indices());
580 for (kind, (i, _c)) in &mut iter {
581 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
586 // Handle case where the comment ends at the end of s.
587 if iter.status == CharClassesStatus::Normal {
594 /// Returns true if text contains any comment.
595 pub fn contains_comment(text: &str) -> bool {
596 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
599 /// Remove trailing spaces from the specified snippet. We do not remove spaces
600 /// inside strings or comments.
601 pub fn remove_trailing_white_spaces(text: &str) -> String {
602 let mut buffer = String::with_capacity(text.len());
603 let mut space_buffer = String::with_capacity(128);
604 for (char_kind, c) in CharClasses::new(text.chars()) {
607 if char_kind == FullCodeCharKind::InString {
608 buffer.push_str(&space_buffer);
610 space_buffer.clear();
613 _ if c.is_whitespace() => {
614 space_buffer.push(c);
617 if !space_buffer.is_empty() {
618 buffer.push_str(&space_buffer);
619 space_buffer.clear();
628 pub struct CharClasses<T>
633 base: iter::Peekable<T>,
634 status: CharClassesStatus,
638 fn get_char(&self) -> char;
641 impl RichChar for char {
642 fn get_char(&self) -> char {
647 impl RichChar for (usize, char) {
648 fn get_char(&self) -> char {
653 impl RichChar for (char, usize) {
654 fn get_char(&self) -> char {
659 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
660 enum CharClassesStatus {
666 // The u32 is the nesting deepness of the comment
668 // Status when the '/' has been consumed, but not yet the '*', deepness is
669 // the new deepness (after the comment opening).
670 BlockCommentOpening(u32),
671 // Status when the '*' has been consumed, but not yet the '/', deepness is
672 // the new deepness (after the comment closing).
673 BlockCommentClosing(u32),
677 /// Distinguish between functional part of code and comments
678 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
679 pub enum CodeCharKind {
684 /// Distinguish between functional part of code and comments,
685 /// describing opening and closing of comments for ease when chunking
686 /// code from tagged characters
687 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
688 pub enum FullCodeCharKind {
690 /// The first character of a comment, there is only one for a comment (always '/')
692 /// Any character inside a comment including the second character of comment
693 /// marks ("//", "/*")
695 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
701 impl FullCodeCharKind {
702 pub fn is_comment(&self) -> bool {
704 FullCodeCharKind::StartComment
705 | FullCodeCharKind::InComment
706 | FullCodeCharKind::EndComment => true,
711 pub fn is_string(&self) -> bool {
712 *self == FullCodeCharKind::InString
715 fn to_codecharkind(&self) -> CodeCharKind {
716 if self.is_comment() {
717 CodeCharKind::Comment
724 impl<T> CharClasses<T>
729 pub fn new(base: T) -> CharClasses<T> {
731 base: base.peekable(),
732 status: CharClassesStatus::Normal,
737 impl<T> Iterator for CharClasses<T>
742 type Item = (FullCodeCharKind, T::Item);
744 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
745 let item = self.base.next()?;
746 let chr = item.get_char();
747 let mut char_kind = FullCodeCharKind::Normal;
748 self.status = match self.status {
749 CharClassesStatus::LitString => match chr {
750 '"' => CharClassesStatus::Normal,
752 char_kind = FullCodeCharKind::InString;
753 CharClassesStatus::LitStringEscape
756 char_kind = FullCodeCharKind::InString;
757 CharClassesStatus::LitString
760 CharClassesStatus::LitStringEscape => {
761 char_kind = FullCodeCharKind::InString;
762 CharClassesStatus::LitString
764 CharClassesStatus::LitChar => match chr {
765 '\\' => CharClassesStatus::LitCharEscape,
766 '\'' => CharClassesStatus::Normal,
767 _ => CharClassesStatus::LitChar,
769 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
770 CharClassesStatus::Normal => match chr {
772 char_kind = FullCodeCharKind::InString;
773 CharClassesStatus::LitString
775 '\'' => CharClassesStatus::LitChar,
776 '/' => match self.base.peek() {
777 Some(next) if next.get_char() == '*' => {
778 self.status = CharClassesStatus::BlockCommentOpening(1);
779 return Some((FullCodeCharKind::StartComment, item));
781 Some(next) if next.get_char() == '/' => {
782 self.status = CharClassesStatus::LineComment;
783 return Some((FullCodeCharKind::StartComment, item));
785 _ => CharClassesStatus::Normal,
787 _ => CharClassesStatus::Normal,
789 CharClassesStatus::BlockComment(deepness) => {
790 assert_ne!(deepness, 0);
791 self.status = match self.base.peek() {
792 Some(next) if next.get_char() == '/' && chr == '*' => {
793 CharClassesStatus::BlockCommentClosing(deepness - 1)
795 Some(next) if next.get_char() == '*' && chr == '/' => {
796 CharClassesStatus::BlockCommentOpening(deepness + 1)
798 _ => CharClassesStatus::BlockComment(deepness),
800 return Some((FullCodeCharKind::InComment, item));
802 CharClassesStatus::BlockCommentOpening(deepness) => {
803 assert_eq!(chr, '*');
804 self.status = CharClassesStatus::BlockComment(deepness);
805 return Some((FullCodeCharKind::InComment, item));
807 CharClassesStatus::BlockCommentClosing(deepness) => {
808 assert_eq!(chr, '/');
810 self.status = CharClassesStatus::Normal;
811 return Some((FullCodeCharKind::EndComment, item));
813 self.status = CharClassesStatus::BlockComment(deepness);
814 return Some((FullCodeCharKind::InComment, item));
817 CharClassesStatus::LineComment => match chr {
819 self.status = CharClassesStatus::Normal;
820 return Some((FullCodeCharKind::EndComment, item));
823 self.status = CharClassesStatus::LineComment;
824 return Some((FullCodeCharKind::InComment, item));
828 Some((char_kind, item))
832 /// Iterator over functional and commented parts of a string. Any part of a string is either
833 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
834 /// comments is functional code. Line comments contain their ending newlines.
835 struct UngroupedCommentCodeSlices<'a> {
837 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
840 impl<'a> UngroupedCommentCodeSlices<'a> {
841 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
842 UngroupedCommentCodeSlices {
844 iter: CharClasses::new(code.char_indices()).peekable(),
849 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
850 type Item = (CodeCharKind, usize, &'a str);
852 fn next(&mut self) -> Option<Self::Item> {
853 let (kind, (start_idx, _)) = self.iter.next()?;
855 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
856 // Consume all the Normal code
857 while let Some(&(char_kind, _)) = self.iter.peek() {
858 if char_kind.is_comment() {
861 let _ = self.iter.next();
864 FullCodeCharKind::StartComment => {
865 // Consume the whole comment
866 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
870 let slice = match self.iter.peek() {
871 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
872 None => &self.slice[start_idx..],
875 if kind.is_comment() {
876 CodeCharKind::Comment
886 /// Iterator over an alternating sequence of functional and commented parts of
887 /// a string. The first item is always a, possibly zero length, subslice of
888 /// functional text. Line style comments contain their ending newlines.
889 pub struct CommentCodeSlices<'a> {
891 last_slice_kind: CodeCharKind,
892 last_slice_end: usize,
895 impl<'a> CommentCodeSlices<'a> {
896 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
899 last_slice_kind: CodeCharKind::Comment,
905 impl<'a> Iterator for CommentCodeSlices<'a> {
906 type Item = (CodeCharKind, usize, &'a str);
908 fn next(&mut self) -> Option<Self::Item> {
909 if self.last_slice_end == self.slice.len() {
913 let mut sub_slice_end = self.last_slice_end;
914 let mut first_whitespace = None;
915 let subslice = &self.slice[self.last_slice_end..];
916 let mut iter = CharClasses::new(subslice.char_indices());
918 for (kind, (i, c)) in &mut iter {
919 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
920 && &subslice[..2] == "//"
921 && [' ', '\t'].contains(&c);
923 if is_comment_connector && first_whitespace.is_none() {
924 first_whitespace = Some(i);
927 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
928 let last_index = match first_whitespace {
932 sub_slice_end = self.last_slice_end + last_index;
936 if !is_comment_connector {
937 first_whitespace = None;
941 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
942 // This was the last subslice.
943 sub_slice_end = match first_whitespace {
944 Some(i) => self.last_slice_end + i,
945 None => self.slice.len(),
949 let kind = match self.last_slice_kind {
950 CodeCharKind::Comment => CodeCharKind::Normal,
951 CodeCharKind::Normal => CodeCharKind::Comment,
956 &self.slice[self.last_slice_end..sub_slice_end],
958 self.last_slice_end = sub_slice_end;
959 self.last_slice_kind = kind;
965 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
966 /// (if it fits in the width/offset, else return None), else return `new`
967 pub fn recover_comment_removed(
970 context: &RewriteContext,
971 ) -> Option<String> {
972 let snippet = context.snippet(span);
973 if snippet != new && changed_comment_content(snippet, &new) {
974 // We missed some comments. Keep the original text.
975 Some(snippet.to_owned())
981 /// Return true if the two strings of code have the same payload of comments.
982 /// The payload of comments is everything in the string except:
983 /// - actual code (not comments)
984 /// - comment start/end marks
986 /// - '*' at the beginning of lines in block comments
987 fn changed_comment_content(orig: &str, new: &str) -> bool {
988 // Cannot write this as a fn since we cannot return types containing closures
989 let code_comment_content = |code| {
990 let slices = UngroupedCommentCodeSlices::new(code);
992 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
993 .flat_map(|(_, _, s)| CommentReducer::new(s))
995 let res = code_comment_content(orig).ne(code_comment_content(new));
997 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1001 code_comment_content(orig).collect::<String>(),
1002 code_comment_content(new).collect::<String>()
1007 /// Iterator over the 'payload' characters of a comment.
1008 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1009 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1011 struct CommentReducer<'a> {
1013 at_start_line: bool,
1014 iter: std::str::Chars<'a>,
1017 impl<'a> CommentReducer<'a> {
1018 fn new(comment: &'a str) -> CommentReducer<'a> {
1019 let is_block = comment.starts_with("/*");
1020 let comment = remove_comment_header(comment);
1023 at_start_line: false, // There are no supplementary '*' on the first line
1024 iter: comment.chars(),
1029 impl<'a> Iterator for CommentReducer<'a> {
1031 fn next(&mut self) -> Option<Self::Item> {
1033 let mut c = self.iter.next()?;
1034 if self.is_block && self.at_start_line {
1035 while c.is_whitespace() {
1036 c = self.iter.next()?;
1038 // Ignore leading '*'
1040 c = self.iter.next()?;
1042 } else if c == '\n' {
1043 self.at_start_line = true;
1045 if !c.is_whitespace() {
1052 fn remove_comment_header(comment: &str) -> &str {
1053 if comment.starts_with("///") || comment.starts_with("//!") {
1055 } else if comment.starts_with("//") {
1057 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1058 || comment.starts_with("/*!")
1060 &comment[3..comment.len() - 2]
1063 comment.starts_with("/*"),
1064 format!("string '{}' is not a comment", comment)
1066 &comment[2..comment.len() - 2]
1072 use super::{contains_comment, rewrite_comment, CharClasses, CodeCharKind, CommentCodeSlices,
1073 FindUncommented, FullCodeCharKind};
1074 use shape::{Indent, Shape};
1078 let mut iter = CharClasses::new("//\n\n".chars());
1080 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1081 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1082 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1083 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1084 assert_eq!(None, iter.next());
1088 fn comment_code_slices() {
1089 let input = "code(); /* test */ 1 + 1";
1090 let mut iter = CommentCodeSlices::new(input);
1092 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1094 (CodeCharKind::Comment, 8, "/* test */"),
1095 iter.next().unwrap()
1097 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1098 assert_eq!(None, iter.next());
1102 fn comment_code_slices_two() {
1103 let input = "// comment\n test();";
1104 let mut iter = CommentCodeSlices::new(input);
1106 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1108 (CodeCharKind::Comment, 0, "// comment\n"),
1109 iter.next().unwrap()
1112 (CodeCharKind::Normal, 11, " test();"),
1113 iter.next().unwrap()
1115 assert_eq!(None, iter.next());
1119 fn comment_code_slices_three() {
1120 let input = "1 // comment\n // comment2\n\n";
1121 let mut iter = CommentCodeSlices::new(input);
1123 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1125 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1126 iter.next().unwrap()
1128 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1129 assert_eq!(None, iter.next());
1133 #[cfg_attr(rustfmt, rustfmt_skip)]
1134 fn format_comments() {
1135 let mut config: ::config::Config = Default::default();
1136 config.set().wrap_comments(true);
1137 config.set().normalize_comments(true);
1139 let comment = rewrite_comment(" //test",
1141 Shape::legacy(100, Indent::new(0, 100)),
1143 assert_eq!("/* test */", comment);
1145 let comment = rewrite_comment("// comment on a",
1147 Shape::legacy(10, Indent::empty()),
1149 assert_eq!("// comment\n// on a", comment);
1151 let comment = rewrite_comment("// A multi line comment\n // between args.",
1153 Shape::legacy(60, Indent::new(0, 12)),
1155 assert_eq!("// A multi line comment\n // between args.", comment);
1157 let input = "// comment";
1160 let comment = rewrite_comment(input,
1162 Shape::legacy(9, Indent::new(0, 69)),
1164 assert_eq!(expected, comment);
1166 let comment = rewrite_comment("/* trimmed */",
1168 Shape::legacy(100, Indent::new(0, 100)),
1170 assert_eq!("/* trimmed */", comment);
1173 // This is probably intended to be a non-test fn, but it is not used. I'm
1174 // keeping it around unless it helps us test stuff.
1175 fn uncommented(text: &str) -> String {
1176 CharClasses::new(text.chars())
1177 .filter_map(|(s, c)| match s {
1178 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1185 fn test_uncommented() {
1186 assert_eq!(&uncommented("abc/*...*/"), "abc");
1188 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1191 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1195 fn test_contains_comment() {
1196 assert_eq!(contains_comment("abc"), false);
1197 assert_eq!(contains_comment("abc // qsdf"), true);
1198 assert_eq!(contains_comment("abc /* kqsdf"), true);
1199 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1203 fn test_find_uncommented() {
1204 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1205 assert_eq!(expected, haystack.find_uncommented(needle));
1208 check("/*/ */test", "test", Some(6));
1209 check("//test\ntest", "test", Some(7));
1210 check("/* comment only */", "whatever", None);
1212 "/* comment */ some text /* more commentary */ result",
1216 check("sup // sup", "p", Some(2));
1217 check("sup", "x", None);
1218 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1219 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1220 check("hel/*lohello*/lo", "hello", None);
1221 check("acb", "ab", None);
1222 check(",/*A*/ ", ",", Some(0));
1223 check("abc", "abc", Some(0));
1224 check("/* abc */", "abc", None);
1225 check("/**/abc/* */", "abc", Some(4));
1226 check("\"/* abc */\"", "abc", Some(4));
1227 check("\"/* abc", "abc", Some(4));