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, borrow::Cow, iter};
15 use itertools::{multipeek, MultiPeek};
16 use syntax::codemap::Span;
19 use rewrite::RewriteContext;
20 use shape::{Indent, Shape};
21 use string::{rewrite_string, StringFormat};
22 use utils::{count_newlines, first_line_width, last_line_width};
24 fn is_custom_comment(comment: &str) -> bool {
25 if !comment.starts_with("//") {
27 } else if let Some(c) = comment.chars().nth(2) {
28 !c.is_alphanumeric() && !c.is_whitespace()
34 #[derive(Copy, Clone, PartialEq, Eq)]
35 pub enum CommentStyle<'a> {
45 fn custom_opener(s: &str) -> &str {
46 s.lines().next().map_or("", |first_line| {
49 .map_or(first_line, |space_index| &first_line[0..space_index + 1])
53 impl<'a> CommentStyle<'a> {
54 pub fn is_doc_comment(&self) -> bool {
56 CommentStyle::TripleSlash | CommentStyle::Doc => true,
61 pub fn opener(&self) -> &'a str {
63 CommentStyle::DoubleSlash => "// ",
64 CommentStyle::TripleSlash => "/// ",
65 CommentStyle::Doc => "//! ",
66 CommentStyle::SingleBullet => "/* ",
67 CommentStyle::DoubleBullet => "/** ",
68 CommentStyle::Exclamation => "/*! ",
69 CommentStyle::Custom(opener) => opener,
73 pub fn closer(&self) -> &'a str {
75 CommentStyle::DoubleSlash
76 | CommentStyle::TripleSlash
77 | CommentStyle::Custom(..)
78 | CommentStyle::Doc => "",
79 CommentStyle::DoubleBullet => " **/",
80 CommentStyle::SingleBullet | CommentStyle::Exclamation => " */",
84 pub fn line_start(&self) -> &'a str {
86 CommentStyle::DoubleSlash => "// ",
87 CommentStyle::TripleSlash => "/// ",
88 CommentStyle::Doc => "//! ",
89 CommentStyle::SingleBullet | CommentStyle::Exclamation => " * ",
90 CommentStyle::DoubleBullet => " ** ",
91 CommentStyle::Custom(opener) => opener,
95 pub fn to_str_tuplet(&self) -> (&'a str, &'a str, &'a str) {
96 (self.opener(), self.closer(), self.line_start())
99 pub fn line_with_same_comment_style(&self, line: &str, normalize_comments: bool) -> bool {
101 CommentStyle::DoubleSlash | CommentStyle::TripleSlash | CommentStyle::Doc => {
102 line.trim_left().starts_with(self.line_start().trim_left())
103 || comment_style(line, normalize_comments) == *self
105 CommentStyle::DoubleBullet | CommentStyle::SingleBullet | CommentStyle::Exclamation => {
106 line.trim_left().starts_with(self.closer().trim_left())
107 || line.trim_left().starts_with(self.line_start().trim_left())
108 || comment_style(line, normalize_comments) == *self
110 CommentStyle::Custom(opener) => line.trim_left().starts_with(opener.trim_right()),
115 fn comment_style(orig: &str, normalize_comments: bool) -> CommentStyle {
116 if !normalize_comments {
117 if orig.starts_with("/**") && !orig.starts_with("/**/") {
118 CommentStyle::DoubleBullet
119 } else if orig.starts_with("/*!") {
120 CommentStyle::Exclamation
121 } else if orig.starts_with("/*") {
122 CommentStyle::SingleBullet
123 } else if orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/') {
124 CommentStyle::TripleSlash
125 } else if orig.starts_with("//!") {
127 } else if is_custom_comment(orig) {
128 CommentStyle::Custom(custom_opener(orig))
130 CommentStyle::DoubleSlash
132 } else if (orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/'))
133 || (orig.starts_with("/**") && !orig.starts_with("/**/"))
135 CommentStyle::TripleSlash
136 } else if orig.starts_with("//!") || orig.starts_with("/*!") {
138 } else if is_custom_comment(orig) {
139 CommentStyle::Custom(custom_opener(orig))
141 CommentStyle::DoubleSlash
145 /// Combine `prev_str` and `next_str` into a single `String`. `span` may contain
146 /// comments between two strings. If there are such comments, then that will be
147 /// recovered. If `allow_extend` is true and there is no comment between the two
148 /// strings, then they will be put on a single line as long as doing so does not
149 /// exceed max width.
150 pub fn combine_strs_with_missing_comments(
151 context: &RewriteContext,
157 ) -> Option<String> {
159 String::with_capacity(prev_str.len() + next_str.len() + shape.indent.width() + 128);
160 result.push_str(prev_str);
161 let mut allow_one_line = !prev_str.contains('\n') && !next_str.contains('\n');
162 let first_sep = if prev_str.is_empty() || next_str.is_empty() {
167 let mut one_line_width =
168 last_line_width(prev_str) + first_line_width(next_str) + first_sep.len();
170 let config = context.config;
171 let indent = shape.indent;
172 let missing_comment = rewrite_missing_comment(span, shape, context)?;
174 if missing_comment.is_empty() {
175 if allow_extend && prev_str.len() + first_sep.len() + next_str.len() <= shape.width {
176 result.push_str(first_sep);
177 } else if !prev_str.is_empty() {
178 result.push_str(&indent.to_string_with_newline(config))
180 result.push_str(next_str);
184 // We have a missing comment between the first expression and the second expression.
186 // Peek the the original source code and find out whether there is a newline between the first
187 // expression and the second expression or the missing comment. We will preserve the original
188 // layout whenever possible.
189 let original_snippet = context.snippet(span);
190 let prefer_same_line = if let Some(pos) = original_snippet.find('/') {
191 !original_snippet[..pos].contains('\n')
193 !original_snippet.contains('\n')
196 one_line_width -= first_sep.len();
197 let first_sep = if prev_str.is_empty() || missing_comment.is_empty() {
200 let one_line_width = last_line_width(prev_str) + first_line_width(&missing_comment) + 1;
201 if prefer_same_line && one_line_width <= shape.width {
204 indent.to_string_with_newline(config)
207 result.push_str(&first_sep);
208 result.push_str(&missing_comment);
210 let second_sep = if missing_comment.is_empty() || next_str.is_empty() {
212 } else if missing_comment.starts_with("//") {
213 indent.to_string_with_newline(config)
215 one_line_width += missing_comment.len() + first_sep.len() + 1;
216 allow_one_line &= !missing_comment.starts_with("//") && !missing_comment.contains('\n');
217 if prefer_same_line && allow_one_line && one_line_width <= shape.width {
220 indent.to_string_with_newline(config)
223 result.push_str(&second_sep);
224 result.push_str(next_str);
229 pub fn rewrite_doc_comment(orig: &str, shape: Shape, config: &Config) -> Option<String> {
230 _rewrite_comment(orig, false, shape, config, true)
233 pub fn rewrite_comment(
238 ) -> Option<String> {
239 _rewrite_comment(orig, block_style, shape, config, false)
247 is_doc_comment: bool,
248 ) -> Option<String> {
249 // If there are lines without a starting sigil, we won't format them correctly
250 // so in that case we won't even re-align (if !config.normalize_comments()) and
251 // we should stop now.
252 let num_bare_lines = orig
254 .map(|line| line.trim())
255 .filter(|l| !(l.starts_with('*') || l.starts_with("//") || l.starts_with("/*")))
257 if num_bare_lines > 0 && !config.normalize_comments() {
258 return Some(orig.to_owned());
260 if !config.normalize_comments() && !config.wrap_comments() {
261 return light_rewrite_comment(orig, shape.indent, config, is_doc_comment);
264 identify_comment(orig, block_style, shape, config, is_doc_comment)
272 is_doc_comment: bool,
273 ) -> Option<String> {
274 let style = comment_style(orig, false);
275 let first_group = orig
277 .take_while(|l| style.line_with_same_comment_style(l, false))
282 .skip(first_group.lines().count())
286 let first_group_str = rewrite_comment_inner(
292 is_doc_comment || style.is_doc_comment(),
295 Some(first_group_str)
297 identify_comment(&rest, block_style, shape, config, is_doc_comment).map(|rest_str| {
301 shape.indent.to_string(config),
308 fn rewrite_comment_inner(
314 is_doc_comment: bool,
315 ) -> Option<String> {
316 let (opener, closer, line_start) = if block_style {
317 CommentStyle::SingleBullet.to_str_tuplet()
319 comment_style(orig, config.normalize_comments()).to_str_tuplet()
322 let max_chars = shape
324 .checked_sub(closer.len() + opener.len())
326 let indent_str = shape.indent.to_string_with_newline(config);
327 let fmt_indent = shape.indent + (opener.len() - line_start.len());
328 let mut fmt = StringFormat {
333 shape: Shape::legacy(max_chars, fmt_indent),
338 let line_breaks = count_newlines(orig.trim_right());
342 .map(|(i, mut line)| {
343 line = trim_right_unless_two_whitespaces(line.trim_left(), is_doc_comment);
345 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
346 line = line[..(line.len() - 2)].trim_right();
351 .map(|s| left_trim_comment_line(s, &style))
352 .map(|(line, has_leading_whitespace)| {
353 if orig.starts_with("/*") && line_breaks == 0 {
356 has_leading_whitespace || config.normalize_comments(),
359 (line, has_leading_whitespace || config.normalize_comments())
363 let mut result = String::with_capacity(orig.len() * 2);
364 result.push_str(opener);
365 let mut code_block_buffer = String::with_capacity(128);
366 let mut is_prev_line_multi_line = false;
367 let mut inside_code_block = false;
368 let comment_line_separator = format!("{}{}", indent_str, line_start);
369 let join_code_block_with_comment_line_separator = |s: &str| {
370 let mut result = String::with_capacity(s.len() + 128);
371 let mut iter = s.lines().peekable();
372 while let Some(line) = iter.next() {
373 result.push_str(line);
374 result.push_str(match iter.peek() {
375 Some(next_line) if next_line.is_empty() => comment_line_separator.trim_right(),
376 Some(..) => &comment_line_separator,
383 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
384 let is_last = i == count_newlines(orig);
386 if inside_code_block {
387 if line.starts_with("```") {
388 inside_code_block = false;
389 result.push_str(&comment_line_separator);
390 let code_block = ::format_code_block(&code_block_buffer, config)
391 .unwrap_or_else(|| code_block_buffer.to_owned());
392 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
393 code_block_buffer.clear();
394 result.push_str(&comment_line_separator);
395 result.push_str(line);
397 code_block_buffer.push_str(line);
398 code_block_buffer.push('\n');
401 // There is an code block that is not properly enclosed by backticks.
402 // We will leave them untouched.
403 result.push_str(&comment_line_separator);
404 result.push_str(&join_code_block_with_comment_line_separator(
412 inside_code_block = line.starts_with("```rust");
414 if result == opener {
415 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
416 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
422 } else if is_prev_line_multi_line && !line.is_empty() {
424 } else if is_last && !closer.is_empty() && line.is_empty() {
425 result.push_str(&indent_str);
427 result.push_str(&comment_line_separator);
428 if !has_leading_whitespace && result.ends_with(' ') {
434 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
435 match rewrite_string(line, &fmt, Some(max_chars)) {
437 is_prev_line_multi_line = s.contains('\n');
440 None if is_prev_line_multi_line => {
441 // We failed to put the current `line` next to the previous `line`.
442 // Remove the trailing space, then start rewrite on the next line.
444 result.push_str(&comment_line_separator);
445 fmt.shape = Shape::legacy(max_chars, fmt_indent);
446 match rewrite_string(line, &fmt, Some(max_chars)) {
448 is_prev_line_multi_line = s.contains('\n');
452 is_prev_line_multi_line = false;
453 result.push_str(line);
458 is_prev_line_multi_line = false;
459 result.push_str(line);
463 fmt.shape = if is_prev_line_multi_line {
465 let offset = 1 + last_line_width(&result) - line_start.len();
467 width: max_chars.saturating_sub(offset),
469 offset: fmt.shape.offset + offset,
472 Shape::legacy(max_chars, fmt_indent)
475 if line.is_empty() && result.ends_with(' ') && !is_last {
476 // Remove space if this is an empty comment or a doc comment.
479 result.push_str(line);
480 fmt.shape = Shape::legacy(max_chars, fmt_indent);
481 is_prev_line_multi_line = false;
485 result.push_str(closer);
486 if result.ends_with(opener) && opener.ends_with(' ') {
494 /// Returns true if the given string MAY include URLs or alike.
495 fn has_url(s: &str) -> bool {
496 // This function may return false positive, but should get its job done in most cases.
497 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
500 /// Given the span, rewrite the missing comment inside it if available.
501 /// Note that the given span must only include comments (or leading/trailing whitespaces).
502 pub fn rewrite_missing_comment(
505 context: &RewriteContext,
506 ) -> Option<String> {
507 let missing_snippet = context.snippet(span);
508 let trimmed_snippet = missing_snippet.trim();
509 if !trimmed_snippet.is_empty() {
510 rewrite_comment(trimmed_snippet, false, shape, context.config)
516 /// Recover the missing comments in the specified span, if available.
517 /// The layout of the comments will be preserved as long as it does not break the code
518 /// and its total width does not exceed the max width.
519 pub fn recover_missing_comment_in_span(
522 context: &RewriteContext,
524 ) -> Option<String> {
525 let missing_comment = rewrite_missing_comment(span, shape, context)?;
526 if missing_comment.is_empty() {
529 let missing_snippet = context.snippet(span);
530 let pos = missing_snippet.find('/').unwrap_or(0);
532 let total_width = missing_comment.len() + used_width + 1;
533 let force_new_line_before_comment =
534 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
535 let sep = if force_new_line_before_comment {
536 shape.indent.to_string_with_newline(context.config)
540 Some(format!("{}{}", sep, missing_comment))
544 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
545 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
546 if is_doc_comment && s.ends_with(" ") {
553 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
554 fn light_rewrite_comment(
558 is_doc_comment: bool,
559 ) -> Option<String> {
560 let lines: Vec<&str> = orig
563 // This is basically just l.trim(), but in the case that a line starts
564 // with `*` we want to leave one space before it, so it aligns with the
566 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
567 let left_trimmed = if let Some(fnw) = first_non_whitespace {
568 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
576 // Preserve markdown's double-space line break syntax in doc comment.
577 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
580 Some(lines.join(&format!("\n{}", offset.to_string(config))))
583 /// Trims comment characters and possibly a single space from the left of a string.
584 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
585 /// this function returns true.
586 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
587 if line.starts_with("//! ")
588 || line.starts_with("/// ")
589 || line.starts_with("/*! ")
590 || line.starts_with("/** ")
593 } else if let CommentStyle::Custom(opener) = *style {
594 if line.starts_with(opener) {
595 (&line[opener.len()..], true)
597 (&line[opener.trim_right().len()..], false)
599 } else if line.starts_with("/* ")
600 || line.starts_with("// ")
601 || line.starts_with("//!")
602 || line.starts_with("///")
603 || line.starts_with("** ")
604 || line.starts_with("/*!")
605 || (line.starts_with("/**") && !line.starts_with("/**/"))
607 (&line[3..], line.chars().nth(2).unwrap() == ' ')
608 } else if line.starts_with("/*")
609 || line.starts_with("* ")
610 || line.starts_with("//")
611 || line.starts_with("**")
613 (&line[2..], line.chars().nth(1).unwrap() == ' ')
614 } else if line.starts_with('*') {
617 (line, line.starts_with(' '))
621 pub trait FindUncommented {
622 fn find_uncommented(&self, pat: &str) -> Option<usize>;
625 impl FindUncommented for str {
626 fn find_uncommented(&self, pat: &str) -> Option<usize> {
627 let mut needle_iter = pat.chars();
628 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
629 match needle_iter.next() {
631 return Some(i - pat.len());
633 Some(c) => match kind {
634 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
636 needle_iter = pat.chars();
642 // Handle case where the pattern is a suffix of the search string
643 match needle_iter.next() {
645 None => Some(self.len() - pat.len()),
650 // Returns the first byte position after the first comment. The given string
651 // is expected to be prefixed by a comment, including delimiters.
652 // Good: "/* /* inner */ outer */ code();"
653 // Bad: "code(); // hello\n world!"
654 pub fn find_comment_end(s: &str) -> Option<usize> {
655 let mut iter = CharClasses::new(s.char_indices());
656 for (kind, (i, _c)) in &mut iter {
657 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
662 // Handle case where the comment ends at the end of s.
663 if iter.status == CharClassesStatus::Normal {
670 /// Returns true if text contains any comment.
671 pub fn contains_comment(text: &str) -> bool {
672 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
675 /// Remove trailing spaces from the specified snippet. We do not remove spaces
676 /// inside strings or comments.
677 pub fn remove_trailing_white_spaces(text: &str) -> String {
678 let mut buffer = String::with_capacity(text.len());
679 let mut space_buffer = String::with_capacity(128);
680 for (char_kind, c) in CharClasses::new(text.chars()) {
683 if char_kind == FullCodeCharKind::InString {
684 buffer.push_str(&space_buffer);
686 space_buffer.clear();
689 _ if c.is_whitespace() => {
690 space_buffer.push(c);
693 if !space_buffer.is_empty() {
694 buffer.push_str(&space_buffer);
695 space_buffer.clear();
704 pub struct CharClasses<T>
710 status: CharClassesStatus,
714 fn get_char(&self) -> char;
717 impl RichChar for char {
718 fn get_char(&self) -> char {
723 impl RichChar for (usize, char) {
724 fn get_char(&self) -> char {
729 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
730 enum CharClassesStatus {
736 // The u32 is the nesting deepness of the comment
738 // Status when the '/' has been consumed, but not yet the '*', deepness is
739 // the new deepness (after the comment opening).
740 BlockCommentOpening(u32),
741 // Status when the '*' has been consumed, but not yet the '/', deepness is
742 // the new deepness (after the comment closing).
743 BlockCommentClosing(u32),
747 /// Distinguish between functional part of code and comments
748 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
749 pub enum CodeCharKind {
754 /// Distinguish between functional part of code and comments,
755 /// describing opening and closing of comments for ease when chunking
756 /// code from tagged characters
757 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
758 pub enum FullCodeCharKind {
760 /// The first character of a comment, there is only one for a comment (always '/')
762 /// Any character inside a comment including the second character of comment
763 /// marks ("//", "/*")
765 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
771 impl FullCodeCharKind {
772 pub fn is_comment(&self) -> bool {
774 FullCodeCharKind::StartComment
775 | FullCodeCharKind::InComment
776 | FullCodeCharKind::EndComment => true,
781 pub fn is_string(&self) -> bool {
782 *self == FullCodeCharKind::InString
785 fn to_codecharkind(&self) -> CodeCharKind {
786 if self.is_comment() {
787 CodeCharKind::Comment
794 impl<T> CharClasses<T>
799 pub fn new(base: T) -> CharClasses<T> {
801 base: multipeek(base),
802 status: CharClassesStatus::Normal,
807 impl<T> Iterator for CharClasses<T>
812 type Item = (FullCodeCharKind, T::Item);
814 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
815 let item = self.base.next()?;
816 let chr = item.get_char();
817 let mut char_kind = FullCodeCharKind::Normal;
818 self.status = match self.status {
819 CharClassesStatus::LitString => match chr {
820 '"' => CharClassesStatus::Normal,
822 char_kind = FullCodeCharKind::InString;
823 CharClassesStatus::LitStringEscape
826 char_kind = FullCodeCharKind::InString;
827 CharClassesStatus::LitString
830 CharClassesStatus::LitStringEscape => {
831 char_kind = FullCodeCharKind::InString;
832 CharClassesStatus::LitString
834 CharClassesStatus::LitChar => match chr {
835 '\\' => CharClassesStatus::LitCharEscape,
836 '\'' => CharClassesStatus::Normal,
837 _ => CharClassesStatus::LitChar,
839 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
840 CharClassesStatus::Normal => match chr {
842 char_kind = FullCodeCharKind::InString;
843 CharClassesStatus::LitString
846 // HACK: Work around mut borrow.
847 match self.base.peek() {
848 Some(next) if next.get_char() == '\\' => {
849 self.status = CharClassesStatus::LitChar;
850 return Some((char_kind, item));
855 match self.base.peek() {
856 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
857 _ => CharClassesStatus::Normal,
860 '/' => match self.base.peek() {
861 Some(next) if next.get_char() == '*' => {
862 self.status = CharClassesStatus::BlockCommentOpening(1);
863 return Some((FullCodeCharKind::StartComment, item));
865 Some(next) if next.get_char() == '/' => {
866 self.status = CharClassesStatus::LineComment;
867 return Some((FullCodeCharKind::StartComment, item));
869 _ => CharClassesStatus::Normal,
871 _ => CharClassesStatus::Normal,
873 CharClassesStatus::BlockComment(deepness) => {
874 assert_ne!(deepness, 0);
875 self.status = match self.base.peek() {
876 Some(next) if next.get_char() == '/' && chr == '*' => {
877 CharClassesStatus::BlockCommentClosing(deepness - 1)
879 Some(next) if next.get_char() == '*' && chr == '/' => {
880 CharClassesStatus::BlockCommentOpening(deepness + 1)
882 _ => CharClassesStatus::BlockComment(deepness),
884 return Some((FullCodeCharKind::InComment, item));
886 CharClassesStatus::BlockCommentOpening(deepness) => {
887 assert_eq!(chr, '*');
888 self.status = CharClassesStatus::BlockComment(deepness);
889 return Some((FullCodeCharKind::InComment, item));
891 CharClassesStatus::BlockCommentClosing(deepness) => {
892 assert_eq!(chr, '/');
894 self.status = CharClassesStatus::Normal;
895 return Some((FullCodeCharKind::EndComment, item));
897 self.status = CharClassesStatus::BlockComment(deepness);
898 return Some((FullCodeCharKind::InComment, item));
901 CharClassesStatus::LineComment => match chr {
903 self.status = CharClassesStatus::Normal;
904 return Some((FullCodeCharKind::EndComment, item));
907 self.status = CharClassesStatus::LineComment;
908 return Some((FullCodeCharKind::InComment, item));
912 Some((char_kind, item))
916 /// An iterator over the lines of a string, paired with the char kind at the
918 pub struct LineClasses<'a> {
919 base: iter::Peekable<CharClasses<std::str::Chars<'a>>>,
920 kind: FullCodeCharKind,
923 impl<'a> LineClasses<'a> {
924 pub fn new(s: &'a str) -> Self {
926 base: CharClasses::new(s.chars()).peekable(),
927 kind: FullCodeCharKind::Normal,
932 impl<'a> Iterator for LineClasses<'a> {
933 type Item = (FullCodeCharKind, String);
935 fn next(&mut self) -> Option<Self::Item> {
936 if self.base.peek().is_none() {
940 let mut line = String::new();
942 while let Some((kind, c)) = self.base.next() {
951 Some((self.kind, line))
955 /// Iterator over functional and commented parts of a string. Any part of a string is either
956 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
957 /// comments is functional code. Line comments contain their ending newlines.
958 struct UngroupedCommentCodeSlices<'a> {
960 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
963 impl<'a> UngroupedCommentCodeSlices<'a> {
964 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
965 UngroupedCommentCodeSlices {
967 iter: CharClasses::new(code.char_indices()).peekable(),
972 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
973 type Item = (CodeCharKind, usize, &'a str);
975 fn next(&mut self) -> Option<Self::Item> {
976 let (kind, (start_idx, _)) = self.iter.next()?;
978 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
979 // Consume all the Normal code
980 while let Some(&(char_kind, _)) = self.iter.peek() {
981 if char_kind.is_comment() {
984 let _ = self.iter.next();
987 FullCodeCharKind::StartComment => {
988 // Consume the whole comment
989 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
993 let slice = match self.iter.peek() {
994 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
995 None => &self.slice[start_idx..],
998 if kind.is_comment() {
999 CodeCharKind::Comment
1001 CodeCharKind::Normal
1009 /// Iterator over an alternating sequence of functional and commented parts of
1010 /// a string. The first item is always a, possibly zero length, subslice of
1011 /// functional text. Line style comments contain their ending newlines.
1012 pub struct CommentCodeSlices<'a> {
1014 last_slice_kind: CodeCharKind,
1015 last_slice_end: usize,
1018 impl<'a> CommentCodeSlices<'a> {
1019 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
1022 last_slice_kind: CodeCharKind::Comment,
1028 impl<'a> Iterator for CommentCodeSlices<'a> {
1029 type Item = (CodeCharKind, usize, &'a str);
1031 fn next(&mut self) -> Option<Self::Item> {
1032 if self.last_slice_end == self.slice.len() {
1036 let mut sub_slice_end = self.last_slice_end;
1037 let mut first_whitespace = None;
1038 let subslice = &self.slice[self.last_slice_end..];
1039 let mut iter = CharClasses::new(subslice.char_indices());
1041 for (kind, (i, c)) in &mut iter {
1042 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
1043 && &subslice[..2] == "//"
1044 && [' ', '\t'].contains(&c);
1046 if is_comment_connector && first_whitespace.is_none() {
1047 first_whitespace = Some(i);
1050 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
1051 let last_index = match first_whitespace {
1055 sub_slice_end = self.last_slice_end + last_index;
1059 if !is_comment_connector {
1060 first_whitespace = None;
1064 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1065 // This was the last subslice.
1066 sub_slice_end = match first_whitespace {
1067 Some(i) => self.last_slice_end + i,
1068 None => self.slice.len(),
1072 let kind = match self.last_slice_kind {
1073 CodeCharKind::Comment => CodeCharKind::Normal,
1074 CodeCharKind::Normal => CodeCharKind::Comment,
1078 self.last_slice_end,
1079 &self.slice[self.last_slice_end..sub_slice_end],
1081 self.last_slice_end = sub_slice_end;
1082 self.last_slice_kind = kind;
1088 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1089 /// (if it fits in the width/offset, else return None), else return `new`
1090 pub fn recover_comment_removed(
1093 context: &RewriteContext,
1094 ) -> Option<String> {
1095 let snippet = context.snippet(span);
1096 if snippet != new && changed_comment_content(snippet, &new) {
1097 // We missed some comments. Keep the original text.
1098 Some(snippet.to_owned())
1104 /// Return true if the two strings of code have the same payload of comments.
1105 /// The payload of comments is everything in the string except:
1106 /// - actual code (not comments)
1107 /// - comment start/end marks
1109 /// - '*' at the beginning of lines in block comments
1110 fn changed_comment_content(orig: &str, new: &str) -> bool {
1111 // Cannot write this as a fn since we cannot return types containing closures
1112 let code_comment_content = |code| {
1113 let slices = UngroupedCommentCodeSlices::new(code);
1115 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1116 .flat_map(|(_, _, s)| CommentReducer::new(s))
1118 let res = code_comment_content(orig).ne(code_comment_content(new));
1120 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1124 code_comment_content(orig).collect::<String>(),
1125 code_comment_content(new).collect::<String>()
1130 /// Iterator over the 'payload' characters of a comment.
1131 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1132 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1134 struct CommentReducer<'a> {
1136 at_start_line: bool,
1137 iter: std::str::Chars<'a>,
1140 impl<'a> CommentReducer<'a> {
1141 fn new(comment: &'a str) -> CommentReducer<'a> {
1142 let is_block = comment.starts_with("/*");
1143 let comment = remove_comment_header(comment);
1146 at_start_line: false, // There are no supplementary '*' on the first line
1147 iter: comment.chars(),
1152 impl<'a> Iterator for CommentReducer<'a> {
1155 fn next(&mut self) -> Option<Self::Item> {
1157 let mut c = self.iter.next()?;
1158 if self.is_block && self.at_start_line {
1159 while c.is_whitespace() {
1160 c = self.iter.next()?;
1162 // Ignore leading '*'
1164 c = self.iter.next()?;
1166 } else if c == '\n' {
1167 self.at_start_line = true;
1169 if !c.is_whitespace() {
1176 fn remove_comment_header(comment: &str) -> &str {
1177 if comment.starts_with("///") || comment.starts_with("//!") {
1179 } else if comment.starts_with("//") {
1181 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1182 || comment.starts_with("/*!")
1184 &comment[3..comment.len() - 2]
1187 comment.starts_with("/*"),
1188 format!("string '{}' is not a comment", comment)
1190 &comment[2..comment.len() - 2]
1197 use shape::{Indent, Shape};
1201 let mut iter = CharClasses::new("//\n\n".chars());
1203 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1204 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1205 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1206 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1207 assert_eq!(None, iter.next());
1211 fn comment_code_slices() {
1212 let input = "code(); /* test */ 1 + 1";
1213 let mut iter = CommentCodeSlices::new(input);
1215 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1217 (CodeCharKind::Comment, 8, "/* test */"),
1218 iter.next().unwrap()
1220 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1221 assert_eq!(None, iter.next());
1225 fn comment_code_slices_two() {
1226 let input = "// comment\n test();";
1227 let mut iter = CommentCodeSlices::new(input);
1229 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1231 (CodeCharKind::Comment, 0, "// comment\n"),
1232 iter.next().unwrap()
1235 (CodeCharKind::Normal, 11, " test();"),
1236 iter.next().unwrap()
1238 assert_eq!(None, iter.next());
1242 fn comment_code_slices_three() {
1243 let input = "1 // comment\n // comment2\n\n";
1244 let mut iter = CommentCodeSlices::new(input);
1246 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1248 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1249 iter.next().unwrap()
1251 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1252 assert_eq!(None, iter.next());
1257 fn format_comments() {
1258 let mut config: ::config::Config = Default::default();
1259 config.set().wrap_comments(true);
1260 config.set().normalize_comments(true);
1262 let comment = rewrite_comment(" //test",
1264 Shape::legacy(100, Indent::new(0, 100)),
1266 assert_eq!("/* test */", comment);
1268 let comment = rewrite_comment("// comment on a",
1270 Shape::legacy(10, Indent::empty()),
1272 assert_eq!("// comment\n// on a", comment);
1274 let comment = rewrite_comment("// A multi line comment\n // between args.",
1276 Shape::legacy(60, Indent::new(0, 12)),
1278 assert_eq!("// A multi line comment\n // between args.", comment);
1280 let input = "// comment";
1283 let comment = rewrite_comment(input,
1285 Shape::legacy(9, Indent::new(0, 69)),
1287 assert_eq!(expected, comment);
1289 let comment = rewrite_comment("/* trimmed */",
1291 Shape::legacy(100, Indent::new(0, 100)),
1293 assert_eq!("/* trimmed */", comment);
1296 // This is probably intended to be a non-test fn, but it is not used. I'm
1297 // keeping it around unless it helps us test stuff.
1298 fn uncommented(text: &str) -> String {
1299 CharClasses::new(text.chars())
1300 .filter_map(|(s, c)| match s {
1301 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1308 fn test_uncommented() {
1309 assert_eq!(&uncommented("abc/*...*/"), "abc");
1311 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1314 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1318 fn test_contains_comment() {
1319 assert_eq!(contains_comment("abc"), false);
1320 assert_eq!(contains_comment("abc // qsdf"), true);
1321 assert_eq!(contains_comment("abc /* kqsdf"), true);
1322 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1326 fn test_find_uncommented() {
1327 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1328 assert_eq!(expected, haystack.find_uncommented(needle));
1331 check("/*/ */test", "test", Some(6));
1332 check("//test\ntest", "test", Some(7));
1333 check("/* comment only */", "whatever", None);
1335 "/* comment */ some text /* more commentary */ result",
1339 check("sup // sup", "p", Some(2));
1340 check("sup", "x", None);
1341 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1342 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1343 check("hel/*lohello*/lo", "hello", None);
1344 check("acb", "ab", None);
1345 check(",/*A*/ ", ",", Some(0));
1346 check("abc", "abc", Some(0));
1347 check("/* abc */", "abc", None);
1348 check("/**/abc/* */", "abc", Some(4));
1349 check("\"/* abc */\"", "abc", Some(4));
1350 check("\"/* abc", "abc", Some(4));
1354 fn test_remove_trailing_white_spaces() {
1355 let s = format!(" r#\"\n test\n \"#");
1356 assert_eq!(remove_trailing_white_spaces(&s), s);