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);
391 let mut config = config.clone();
392 config.set().wrap_comments(false);
393 match ::format_code_block(&code_block_buffer, &config) {
394 Some(ref s) => trim_custom_comment_prefix(s),
395 None => trim_custom_comment_prefix(&code_block_buffer),
398 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
399 code_block_buffer.clear();
400 result.push_str(&comment_line_separator);
401 result.push_str(line);
403 code_block_buffer.push_str(&hide_sharp_behind_comment(line));
404 code_block_buffer.push('\n');
407 // There is an code block that is not properly enclosed by backticks.
408 // We will leave them untouched.
409 result.push_str(&comment_line_separator);
410 result.push_str(&join_code_block_with_comment_line_separator(
411 &trim_custom_comment_prefix(&code_block_buffer),
418 inside_code_block = line.starts_with("```");
420 if result == opener {
421 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
422 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
428 } else if is_prev_line_multi_line && !line.is_empty() {
430 } else if is_last && !closer.is_empty() && line.is_empty() {
431 result.push_str(&indent_str);
433 result.push_str(&comment_line_separator);
434 if !has_leading_whitespace && result.ends_with(' ') {
440 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
441 match rewrite_string(line, &fmt, Some(max_chars)) {
443 is_prev_line_multi_line = s.contains('\n');
446 None if is_prev_line_multi_line => {
447 // We failed to put the current `line` next to the previous `line`.
448 // Remove the trailing space, then start rewrite on the next line.
450 result.push_str(&comment_line_separator);
451 fmt.shape = Shape::legacy(max_chars, fmt_indent);
452 match rewrite_string(line, &fmt, Some(max_chars)) {
454 is_prev_line_multi_line = s.contains('\n');
458 is_prev_line_multi_line = false;
459 result.push_str(line);
464 is_prev_line_multi_line = false;
465 result.push_str(line);
469 fmt.shape = if is_prev_line_multi_line {
471 let offset = 1 + last_line_width(&result) - line_start.len();
473 width: max_chars.saturating_sub(offset),
475 offset: fmt.shape.offset + offset,
478 Shape::legacy(max_chars, fmt_indent)
481 if line.is_empty() && result.ends_with(' ') && !is_last {
482 // Remove space if this is an empty comment or a doc comment.
485 result.push_str(line);
486 fmt.shape = Shape::legacy(max_chars, fmt_indent);
487 is_prev_line_multi_line = false;
491 result.push_str(closer);
492 if result.ends_with(opener) && opener.ends_with(' ') {
500 const RUSTFMT_CUSTOM_COMMENT_PREFIX: &str = "//#### ";
502 fn hide_sharp_behind_comment<'a>(s: &'a str) -> Cow<'a, str> {
503 if s.trim_left().starts_with('#') {
504 Cow::from(format!("{}{}", RUSTFMT_CUSTOM_COMMENT_PREFIX, s))
510 fn trim_custom_comment_prefix(s: &str) -> String {
513 let left_trimmed = line.trim_left();
514 if left_trimmed.starts_with(RUSTFMT_CUSTOM_COMMENT_PREFIX) {
515 left_trimmed.trim_left_matches(RUSTFMT_CUSTOM_COMMENT_PREFIX)
524 /// Returns true if the given string MAY include URLs or alike.
525 fn has_url(s: &str) -> bool {
526 // This function may return false positive, but should get its job done in most cases.
527 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
530 /// Given the span, rewrite the missing comment inside it if available.
531 /// Note that the given span must only include comments (or leading/trailing whitespaces).
532 pub fn rewrite_missing_comment(
535 context: &RewriteContext,
536 ) -> Option<String> {
537 let missing_snippet = context.snippet(span);
538 let trimmed_snippet = missing_snippet.trim();
539 if !trimmed_snippet.is_empty() {
540 rewrite_comment(trimmed_snippet, false, shape, context.config)
546 /// Recover the missing comments in the specified span, if available.
547 /// The layout of the comments will be preserved as long as it does not break the code
548 /// and its total width does not exceed the max width.
549 pub fn recover_missing_comment_in_span(
552 context: &RewriteContext,
554 ) -> Option<String> {
555 let missing_comment = rewrite_missing_comment(span, shape, context)?;
556 if missing_comment.is_empty() {
559 let missing_snippet = context.snippet(span);
560 let pos = missing_snippet.find('/').unwrap_or(0);
562 let total_width = missing_comment.len() + used_width + 1;
563 let force_new_line_before_comment =
564 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
565 let sep = if force_new_line_before_comment {
566 shape.indent.to_string_with_newline(context.config)
570 Some(format!("{}{}", sep, missing_comment))
574 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
575 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
576 if is_doc_comment && s.ends_with(" ") {
583 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
584 fn light_rewrite_comment(
588 is_doc_comment: bool,
589 ) -> Option<String> {
590 let lines: Vec<&str> = orig
593 // This is basically just l.trim(), but in the case that a line starts
594 // with `*` we want to leave one space before it, so it aligns with the
596 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
597 let left_trimmed = if let Some(fnw) = first_non_whitespace {
598 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
606 // Preserve markdown's double-space line break syntax in doc comment.
607 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
610 Some(lines.join(&format!("\n{}", offset.to_string(config))))
613 /// Trims comment characters and possibly a single space from the left of a string.
614 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
615 /// this function returns true.
616 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
617 if line.starts_with("//! ")
618 || line.starts_with("/// ")
619 || line.starts_with("/*! ")
620 || line.starts_with("/** ")
623 } else if let CommentStyle::Custom(opener) = *style {
624 if line.starts_with(opener) {
625 (&line[opener.len()..], true)
627 (&line[opener.trim_right().len()..], false)
629 } else if line.starts_with("/* ")
630 || line.starts_with("// ")
631 || line.starts_with("//!")
632 || line.starts_with("///")
633 || line.starts_with("** ")
634 || line.starts_with("/*!")
635 || (line.starts_with("/**") && !line.starts_with("/**/"))
637 (&line[3..], line.chars().nth(2).unwrap() == ' ')
638 } else if line.starts_with("/*")
639 || line.starts_with("* ")
640 || line.starts_with("//")
641 || line.starts_with("**")
643 (&line[2..], line.chars().nth(1).unwrap() == ' ')
644 } else if line.starts_with('*') {
647 (line, line.starts_with(' '))
651 pub trait FindUncommented {
652 fn find_uncommented(&self, pat: &str) -> Option<usize>;
655 impl FindUncommented for str {
656 fn find_uncommented(&self, pat: &str) -> Option<usize> {
657 let mut needle_iter = pat.chars();
658 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
659 match needle_iter.next() {
661 return Some(i - pat.len());
663 Some(c) => match kind {
664 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
666 needle_iter = pat.chars();
672 // Handle case where the pattern is a suffix of the search string
673 match needle_iter.next() {
675 None => Some(self.len() - pat.len()),
680 // Returns the first byte position after the first comment. The given string
681 // is expected to be prefixed by a comment, including delimiters.
682 // Good: "/* /* inner */ outer */ code();"
683 // Bad: "code(); // hello\n world!"
684 pub fn find_comment_end(s: &str) -> Option<usize> {
685 let mut iter = CharClasses::new(s.char_indices());
686 for (kind, (i, _c)) in &mut iter {
687 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
692 // Handle case where the comment ends at the end of s.
693 if iter.status == CharClassesStatus::Normal {
700 /// Returns true if text contains any comment.
701 pub fn contains_comment(text: &str) -> bool {
702 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
705 /// Remove trailing spaces from the specified snippet. We do not remove spaces
706 /// inside strings or comments.
707 pub fn remove_trailing_white_spaces(text: &str) -> String {
708 let mut buffer = String::with_capacity(text.len());
709 let mut space_buffer = String::with_capacity(128);
710 for (char_kind, c) in CharClasses::new(text.chars()) {
713 if char_kind == FullCodeCharKind::InString {
714 buffer.push_str(&space_buffer);
716 space_buffer.clear();
719 _ if c.is_whitespace() => {
720 space_buffer.push(c);
723 if !space_buffer.is_empty() {
724 buffer.push_str(&space_buffer);
725 space_buffer.clear();
734 pub struct CharClasses<T>
740 status: CharClassesStatus,
744 fn get_char(&self) -> char;
747 impl RichChar for char {
748 fn get_char(&self) -> char {
753 impl RichChar for (usize, char) {
754 fn get_char(&self) -> char {
759 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
760 enum CharClassesStatus {
766 // The u32 is the nesting deepness of the comment
768 // Status when the '/' has been consumed, but not yet the '*', deepness is
769 // the new deepness (after the comment opening).
770 BlockCommentOpening(u32),
771 // Status when the '*' has been consumed, but not yet the '/', deepness is
772 // the new deepness (after the comment closing).
773 BlockCommentClosing(u32),
777 /// Distinguish between functional part of code and comments
778 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
779 pub enum CodeCharKind {
784 /// Distinguish between functional part of code and comments,
785 /// describing opening and closing of comments for ease when chunking
786 /// code from tagged characters
787 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
788 pub enum FullCodeCharKind {
790 /// The first character of a comment, there is only one for a comment (always '/')
792 /// Any character inside a comment including the second character of comment
793 /// marks ("//", "/*")
795 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
801 impl FullCodeCharKind {
802 pub fn is_comment(&self) -> bool {
804 FullCodeCharKind::StartComment
805 | FullCodeCharKind::InComment
806 | FullCodeCharKind::EndComment => true,
811 pub fn is_string(&self) -> bool {
812 *self == FullCodeCharKind::InString
815 fn to_codecharkind(&self) -> CodeCharKind {
816 if self.is_comment() {
817 CodeCharKind::Comment
824 impl<T> CharClasses<T>
829 pub fn new(base: T) -> CharClasses<T> {
831 base: multipeek(base),
832 status: CharClassesStatus::Normal,
837 impl<T> Iterator for CharClasses<T>
842 type Item = (FullCodeCharKind, T::Item);
844 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
845 let item = self.base.next()?;
846 let chr = item.get_char();
847 let mut char_kind = FullCodeCharKind::Normal;
848 self.status = match self.status {
849 CharClassesStatus::LitString => match chr {
850 '"' => CharClassesStatus::Normal,
852 char_kind = FullCodeCharKind::InString;
853 CharClassesStatus::LitStringEscape
856 char_kind = FullCodeCharKind::InString;
857 CharClassesStatus::LitString
860 CharClassesStatus::LitStringEscape => {
861 char_kind = FullCodeCharKind::InString;
862 CharClassesStatus::LitString
864 CharClassesStatus::LitChar => match chr {
865 '\\' => CharClassesStatus::LitCharEscape,
866 '\'' => CharClassesStatus::Normal,
867 _ => CharClassesStatus::LitChar,
869 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
870 CharClassesStatus::Normal => match chr {
872 char_kind = FullCodeCharKind::InString;
873 CharClassesStatus::LitString
876 // HACK: Work around mut borrow.
877 match self.base.peek() {
878 Some(next) if next.get_char() == '\\' => {
879 self.status = CharClassesStatus::LitChar;
880 return Some((char_kind, item));
885 match self.base.peek() {
886 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
887 _ => CharClassesStatus::Normal,
890 '/' => match self.base.peek() {
891 Some(next) if next.get_char() == '*' => {
892 self.status = CharClassesStatus::BlockCommentOpening(1);
893 return Some((FullCodeCharKind::StartComment, item));
895 Some(next) if next.get_char() == '/' => {
896 self.status = CharClassesStatus::LineComment;
897 return Some((FullCodeCharKind::StartComment, item));
899 _ => CharClassesStatus::Normal,
901 _ => CharClassesStatus::Normal,
903 CharClassesStatus::BlockComment(deepness) => {
904 assert_ne!(deepness, 0);
905 self.status = match self.base.peek() {
906 Some(next) if next.get_char() == '/' && chr == '*' => {
907 CharClassesStatus::BlockCommentClosing(deepness - 1)
909 Some(next) if next.get_char() == '*' && chr == '/' => {
910 CharClassesStatus::BlockCommentOpening(deepness + 1)
912 _ => CharClassesStatus::BlockComment(deepness),
914 return Some((FullCodeCharKind::InComment, item));
916 CharClassesStatus::BlockCommentOpening(deepness) => {
917 assert_eq!(chr, '*');
918 self.status = CharClassesStatus::BlockComment(deepness);
919 return Some((FullCodeCharKind::InComment, item));
921 CharClassesStatus::BlockCommentClosing(deepness) => {
922 assert_eq!(chr, '/');
924 self.status = CharClassesStatus::Normal;
925 return Some((FullCodeCharKind::EndComment, item));
927 self.status = CharClassesStatus::BlockComment(deepness);
928 return Some((FullCodeCharKind::InComment, item));
931 CharClassesStatus::LineComment => match chr {
933 self.status = CharClassesStatus::Normal;
934 return Some((FullCodeCharKind::EndComment, item));
937 self.status = CharClassesStatus::LineComment;
938 return Some((FullCodeCharKind::InComment, item));
942 Some((char_kind, item))
946 /// An iterator over the lines of a string, paired with the char kind at the
948 pub struct LineClasses<'a> {
949 base: iter::Peekable<CharClasses<std::str::Chars<'a>>>,
950 kind: FullCodeCharKind,
953 impl<'a> LineClasses<'a> {
954 pub fn new(s: &'a str) -> Self {
956 base: CharClasses::new(s.chars()).peekable(),
957 kind: FullCodeCharKind::Normal,
962 impl<'a> Iterator for LineClasses<'a> {
963 type Item = (FullCodeCharKind, String);
965 fn next(&mut self) -> Option<Self::Item> {
966 if self.base.peek().is_none() {
970 let mut line = String::new();
972 while let Some((kind, c)) = self.base.next() {
981 Some((self.kind, line))
985 /// Iterator over functional and commented parts of a string. Any part of a string is either
986 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
987 /// comments is functional code. Line comments contain their ending newlines.
988 struct UngroupedCommentCodeSlices<'a> {
990 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
993 impl<'a> UngroupedCommentCodeSlices<'a> {
994 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
995 UngroupedCommentCodeSlices {
997 iter: CharClasses::new(code.char_indices()).peekable(),
1002 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
1003 type Item = (CodeCharKind, usize, &'a str);
1005 fn next(&mut self) -> Option<Self::Item> {
1006 let (kind, (start_idx, _)) = self.iter.next()?;
1008 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
1009 // Consume all the Normal code
1010 while let Some(&(char_kind, _)) = self.iter.peek() {
1011 if char_kind.is_comment() {
1014 let _ = self.iter.next();
1017 FullCodeCharKind::StartComment => {
1018 // Consume the whole comment
1019 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
1023 let slice = match self.iter.peek() {
1024 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
1025 None => &self.slice[start_idx..],
1028 if kind.is_comment() {
1029 CodeCharKind::Comment
1031 CodeCharKind::Normal
1039 /// Iterator over an alternating sequence of functional and commented parts of
1040 /// a string. The first item is always a, possibly zero length, subslice of
1041 /// functional text. Line style comments contain their ending newlines.
1042 pub struct CommentCodeSlices<'a> {
1044 last_slice_kind: CodeCharKind,
1045 last_slice_end: usize,
1048 impl<'a> CommentCodeSlices<'a> {
1049 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
1052 last_slice_kind: CodeCharKind::Comment,
1058 impl<'a> Iterator for CommentCodeSlices<'a> {
1059 type Item = (CodeCharKind, usize, &'a str);
1061 fn next(&mut self) -> Option<Self::Item> {
1062 if self.last_slice_end == self.slice.len() {
1066 let mut sub_slice_end = self.last_slice_end;
1067 let mut first_whitespace = None;
1068 let subslice = &self.slice[self.last_slice_end..];
1069 let mut iter = CharClasses::new(subslice.char_indices());
1071 for (kind, (i, c)) in &mut iter {
1072 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
1073 && &subslice[..2] == "//"
1074 && [' ', '\t'].contains(&c);
1076 if is_comment_connector && first_whitespace.is_none() {
1077 first_whitespace = Some(i);
1080 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
1081 let last_index = match first_whitespace {
1085 sub_slice_end = self.last_slice_end + last_index;
1089 if !is_comment_connector {
1090 first_whitespace = None;
1094 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1095 // This was the last subslice.
1096 sub_slice_end = match first_whitespace {
1097 Some(i) => self.last_slice_end + i,
1098 None => self.slice.len(),
1102 let kind = match self.last_slice_kind {
1103 CodeCharKind::Comment => CodeCharKind::Normal,
1104 CodeCharKind::Normal => CodeCharKind::Comment,
1108 self.last_slice_end,
1109 &self.slice[self.last_slice_end..sub_slice_end],
1111 self.last_slice_end = sub_slice_end;
1112 self.last_slice_kind = kind;
1118 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1119 /// (if it fits in the width/offset, else return None), else return `new`
1120 pub fn recover_comment_removed(
1123 context: &RewriteContext,
1124 ) -> Option<String> {
1125 let snippet = context.snippet(span);
1126 if snippet != new && changed_comment_content(snippet, &new) {
1127 // We missed some comments. Keep the original text.
1128 Some(snippet.to_owned())
1134 /// Return true if the two strings of code have the same payload of comments.
1135 /// The payload of comments is everything in the string except:
1136 /// - actual code (not comments)
1137 /// - comment start/end marks
1139 /// - '*' at the beginning of lines in block comments
1140 fn changed_comment_content(orig: &str, new: &str) -> bool {
1141 // Cannot write this as a fn since we cannot return types containing closures
1142 let code_comment_content = |code| {
1143 let slices = UngroupedCommentCodeSlices::new(code);
1145 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1146 .flat_map(|(_, _, s)| CommentReducer::new(s))
1148 let res = code_comment_content(orig).ne(code_comment_content(new));
1150 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1154 code_comment_content(orig).collect::<String>(),
1155 code_comment_content(new).collect::<String>()
1160 /// Iterator over the 'payload' characters of a comment.
1161 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1162 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1164 struct CommentReducer<'a> {
1166 at_start_line: bool,
1167 iter: std::str::Chars<'a>,
1170 impl<'a> CommentReducer<'a> {
1171 fn new(comment: &'a str) -> CommentReducer<'a> {
1172 let is_block = comment.starts_with("/*");
1173 let comment = remove_comment_header(comment);
1176 at_start_line: false, // There are no supplementary '*' on the first line
1177 iter: comment.chars(),
1182 impl<'a> Iterator for CommentReducer<'a> {
1185 fn next(&mut self) -> Option<Self::Item> {
1187 let mut c = self.iter.next()?;
1188 if self.is_block && self.at_start_line {
1189 while c.is_whitespace() {
1190 c = self.iter.next()?;
1192 // Ignore leading '*'
1194 c = self.iter.next()?;
1196 } else if c == '\n' {
1197 self.at_start_line = true;
1199 if !c.is_whitespace() {
1206 fn remove_comment_header(comment: &str) -> &str {
1207 if comment.starts_with("///") || comment.starts_with("//!") {
1209 } else if comment.starts_with("//") {
1211 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1212 || comment.starts_with("/*!")
1214 &comment[3..comment.len() - 2]
1217 comment.starts_with("/*"),
1218 format!("string '{}' is not a comment", comment)
1220 &comment[2..comment.len() - 2]
1227 use shape::{Indent, Shape};
1231 let mut iter = CharClasses::new("//\n\n".chars());
1233 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1234 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1235 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1236 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1237 assert_eq!(None, iter.next());
1241 fn comment_code_slices() {
1242 let input = "code(); /* test */ 1 + 1";
1243 let mut iter = CommentCodeSlices::new(input);
1245 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1247 (CodeCharKind::Comment, 8, "/* test */"),
1248 iter.next().unwrap()
1250 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1251 assert_eq!(None, iter.next());
1255 fn comment_code_slices_two() {
1256 let input = "// comment\n test();";
1257 let mut iter = CommentCodeSlices::new(input);
1259 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1261 (CodeCharKind::Comment, 0, "// comment\n"),
1262 iter.next().unwrap()
1265 (CodeCharKind::Normal, 11, " test();"),
1266 iter.next().unwrap()
1268 assert_eq!(None, iter.next());
1272 fn comment_code_slices_three() {
1273 let input = "1 // comment\n // comment2\n\n";
1274 let mut iter = CommentCodeSlices::new(input);
1276 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1278 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1279 iter.next().unwrap()
1281 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1282 assert_eq!(None, iter.next());
1287 fn format_comments() {
1288 let mut config: ::config::Config = Default::default();
1289 config.set().wrap_comments(true);
1290 config.set().normalize_comments(true);
1292 let comment = rewrite_comment(" //test",
1294 Shape::legacy(100, Indent::new(0, 100)),
1296 assert_eq!("/* test */", comment);
1298 let comment = rewrite_comment("// comment on a",
1300 Shape::legacy(10, Indent::empty()),
1302 assert_eq!("// comment\n// on a", comment);
1304 let comment = rewrite_comment("// A multi line comment\n // between args.",
1306 Shape::legacy(60, Indent::new(0, 12)),
1308 assert_eq!("// A multi line comment\n // between args.", comment);
1310 let input = "// comment";
1313 let comment = rewrite_comment(input,
1315 Shape::legacy(9, Indent::new(0, 69)),
1317 assert_eq!(expected, comment);
1319 let comment = rewrite_comment("/* trimmed */",
1321 Shape::legacy(100, Indent::new(0, 100)),
1323 assert_eq!("/* trimmed */", comment);
1326 // This is probably intended to be a non-test fn, but it is not used. I'm
1327 // keeping it around unless it helps us test stuff.
1328 fn uncommented(text: &str) -> String {
1329 CharClasses::new(text.chars())
1330 .filter_map(|(s, c)| match s {
1331 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1338 fn test_uncommented() {
1339 assert_eq!(&uncommented("abc/*...*/"), "abc");
1341 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1344 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1348 fn test_contains_comment() {
1349 assert_eq!(contains_comment("abc"), false);
1350 assert_eq!(contains_comment("abc // qsdf"), true);
1351 assert_eq!(contains_comment("abc /* kqsdf"), true);
1352 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1356 fn test_find_uncommented() {
1357 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1358 assert_eq!(expected, haystack.find_uncommented(needle));
1361 check("/*/ */test", "test", Some(6));
1362 check("//test\ntest", "test", Some(7));
1363 check("/* comment only */", "whatever", None);
1365 "/* comment */ some text /* more commentary */ result",
1369 check("sup // sup", "p", Some(2));
1370 check("sup", "x", None);
1371 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1372 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1373 check("hel/*lohello*/lo", "hello", None);
1374 check("acb", "ab", None);
1375 check(",/*A*/ ", ",", Some(0));
1376 check("abc", "abc", Some(0));
1377 check("/* abc */", "abc", None);
1378 check("/**/abc/* */", "abc", Some(4));
1379 check("\"/* abc */\"", "abc", Some(4));
1380 check("\"/* abc", "abc", Some(4));
1384 fn test_remove_trailing_white_spaces() {
1385 let s = format!(" r#\"\n test\n \"#");
1386 assert_eq!(remove_trailing_white_spaces(&s), s);