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, borrow::Cow};
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.chars().position(|c| c == '/') {
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.lines()
253 .map(|line| line.trim())
254 .filter(|l| !(l.starts_with('*') || l.starts_with("//") || l.starts_with("/*")))
256 if num_bare_lines > 0 && !config.normalize_comments() {
257 return Some(orig.to_owned());
259 if !config.normalize_comments() && !config.wrap_comments() {
260 return light_rewrite_comment(orig, shape.indent, config, is_doc_comment);
263 identify_comment(orig, block_style, shape, config, is_doc_comment)
271 is_doc_comment: bool,
272 ) -> Option<String> {
273 let style = comment_style(orig, false);
274 let first_group = orig.lines()
275 .take_while(|l| style.line_with_same_comment_style(l, false))
278 let rest = orig.lines()
279 .skip(first_group.lines().count())
283 let first_group_str = rewrite_comment_inner(
289 is_doc_comment || style.is_doc_comment(),
292 Some(first_group_str)
294 identify_comment(&rest, block_style, shape, config, is_doc_comment).map(|rest_str| {
298 shape.indent.to_string(config),
305 fn rewrite_comment_inner(
311 is_doc_comment: bool,
312 ) -> Option<String> {
313 let (opener, closer, line_start) = if block_style {
314 CommentStyle::SingleBullet.to_str_tuplet()
316 comment_style(orig, config.normalize_comments()).to_str_tuplet()
319 let max_chars = shape
321 .checked_sub(closer.len() + opener.len())
323 let indent_str = shape.indent.to_string_with_newline(config);
324 let fmt_indent = shape.indent + (opener.len() - line_start.len());
325 let mut fmt = StringFormat {
330 shape: Shape::legacy(max_chars, fmt_indent),
335 let line_breaks = count_newlines(orig.trim_right());
336 let lines = orig.lines()
338 .map(|(i, mut line)| {
339 line = trim_right_unless_two_whitespaces(line.trim_left(), is_doc_comment);
341 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
342 line = line[..(line.len() - 2)].trim_right();
347 .map(|s| left_trim_comment_line(s, &style))
348 .map(|(line, has_leading_whitespace)| {
349 if orig.starts_with("/*") && line_breaks == 0 {
352 has_leading_whitespace || config.normalize_comments(),
355 (line, has_leading_whitespace || config.normalize_comments())
359 let mut result = String::with_capacity(orig.len() * 2);
360 result.push_str(opener);
361 let mut code_block_buffer = String::with_capacity(128);
362 let mut is_prev_line_multi_line = false;
363 let mut inside_code_block = false;
364 let comment_line_separator = format!("{}{}", indent_str, line_start);
365 let join_code_block_with_comment_line_separator = |s: &str| {
366 let mut result = String::with_capacity(s.len() + 128);
367 let mut iter = s.lines().peekable();
368 while let Some(line) = iter.next() {
369 result.push_str(line);
370 result.push_str(match iter.peek() {
371 Some(next_line) if next_line.is_empty() => comment_line_separator.trim_right(),
372 Some(..) => &comment_line_separator,
379 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
380 let is_last = i == count_newlines(orig);
382 if inside_code_block {
383 if line.starts_with("```") {
384 inside_code_block = false;
385 result.push_str(&comment_line_separator);
386 let code_block = ::format_code_block(&code_block_buffer, config)
387 .unwrap_or_else(|| code_block_buffer.to_owned());
388 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
389 code_block_buffer.clear();
390 result.push_str(&comment_line_separator);
391 result.push_str(line);
393 code_block_buffer.push_str(line);
394 code_block_buffer.push('\n');
397 // There is an code block that is not properly enclosed by backticks.
398 // We will leave them untouched.
399 result.push_str(&comment_line_separator);
400 result.push_str(&join_code_block_with_comment_line_separator(
408 inside_code_block = line.starts_with("```rust");
410 if result == opener {
411 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
412 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
418 } else if is_prev_line_multi_line && !line.is_empty() {
420 } else if is_last && !closer.is_empty() && line.is_empty() {
421 result.push_str(&indent_str);
423 result.push_str(&comment_line_separator);
424 if !has_leading_whitespace && result.ends_with(' ') {
430 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
431 match rewrite_string(line, &fmt, Some(max_chars)) {
433 is_prev_line_multi_line = s.contains('\n');
436 None if is_prev_line_multi_line => {
437 // We failed to put the current `line` next to the previous `line`.
438 // Remove the trailing space, then start rewrite on the next line.
440 result.push_str(&comment_line_separator);
441 fmt.shape = Shape::legacy(max_chars, fmt_indent);
442 match rewrite_string(line, &fmt, Some(max_chars)) {
444 is_prev_line_multi_line = s.contains('\n');
448 is_prev_line_multi_line = false;
449 result.push_str(line);
454 is_prev_line_multi_line = false;
455 result.push_str(line);
459 fmt.shape = if is_prev_line_multi_line {
461 let offset = 1 + last_line_width(&result) - line_start.len();
463 width: max_chars.checked_sub(offset).unwrap_or(0),
465 offset: fmt.shape.offset + offset,
468 Shape::legacy(max_chars, fmt_indent)
471 if line.is_empty() && result.ends_with(' ') && !is_last {
472 // Remove space if this is an empty comment or a doc comment.
475 result.push_str(line);
476 fmt.shape = Shape::legacy(max_chars, fmt_indent);
477 is_prev_line_multi_line = false;
481 result.push_str(closer);
482 if result.ends_with(opener) && opener.ends_with(' ') {
490 /// Returns true if the given string MAY include URLs or alike.
491 fn has_url(s: &str) -> bool {
492 // This function may return false positive, but should get its job done in most cases.
493 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
496 /// Given the span, rewrite the missing comment inside it if available.
497 /// Note that the given span must only include comments (or leading/trailing whitespaces).
498 pub fn rewrite_missing_comment(
501 context: &RewriteContext,
502 ) -> Option<String> {
503 let missing_snippet = context.snippet(span);
504 let trimmed_snippet = missing_snippet.trim();
505 if !trimmed_snippet.is_empty() {
506 rewrite_comment(trimmed_snippet, false, shape, context.config)
512 /// Recover the missing comments in the specified span, if available.
513 /// The layout of the comments will be preserved as long as it does not break the code
514 /// and its total width does not exceed the max width.
515 pub fn recover_missing_comment_in_span(
518 context: &RewriteContext,
520 ) -> Option<String> {
521 let missing_comment = rewrite_missing_comment(span, shape, context)?;
522 if missing_comment.is_empty() {
525 let missing_snippet = context.snippet(span);
526 let pos = missing_snippet.chars().position(|c| c == '/').unwrap_or(0);
528 let total_width = missing_comment.len() + used_width + 1;
529 let force_new_line_before_comment =
530 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
531 let sep = if force_new_line_before_comment {
532 shape.indent.to_string_with_newline(context.config)
536 Some(format!("{}{}", sep, missing_comment))
540 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
541 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
542 if is_doc_comment && s.ends_with(" ") {
549 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
550 fn light_rewrite_comment(
554 is_doc_comment: bool,
555 ) -> Option<String> {
556 let lines: Vec<&str> = orig.lines()
558 // This is basically just l.trim(), but in the case that a line starts
559 // with `*` we want to leave one space before it, so it aligns with the
561 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
562 let left_trimmed = if let Some(fnw) = first_non_whitespace {
563 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
571 // Preserve markdown's double-space line break syntax in doc comment.
572 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
575 Some(lines.join(&format!("\n{}", offset.to_string(config))))
578 /// Trims comment characters and possibly a single space from the left of a string.
579 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
580 /// this function returns true.
581 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
582 if line.starts_with("//! ") || line.starts_with("/// ") || line.starts_with("/*! ")
583 || line.starts_with("/** ")
586 } else if let CommentStyle::Custom(opener) = *style {
587 if line.starts_with(opener) {
588 (&line[opener.len()..], true)
590 (&line[opener.trim_right().len()..], false)
592 } else if line.starts_with("/* ") || line.starts_with("// ") || line.starts_with("//!")
593 || line.starts_with("///") || line.starts_with("** ")
594 || line.starts_with("/*!")
595 || (line.starts_with("/**") && !line.starts_with("/**/"))
597 (&line[3..], line.chars().nth(2).unwrap() == ' ')
598 } else if line.starts_with("/*") || line.starts_with("* ") || line.starts_with("//")
599 || line.starts_with("**")
601 (&line[2..], line.chars().nth(1).unwrap() == ' ')
602 } else if line.starts_with('*') {
605 (line, line.starts_with(' '))
609 pub trait FindUncommented {
610 fn find_uncommented(&self, pat: &str) -> Option<usize>;
613 impl FindUncommented for str {
614 fn find_uncommented(&self, pat: &str) -> Option<usize> {
615 let mut needle_iter = pat.chars();
616 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
617 match needle_iter.next() {
619 return Some(i - pat.len());
621 Some(c) => match kind {
622 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
624 needle_iter = pat.chars();
630 // Handle case where the pattern is a suffix of the search string
631 match needle_iter.next() {
633 None => Some(self.len() - pat.len()),
638 // Returns the first byte position after the first comment. The given string
639 // is expected to be prefixed by a comment, including delimiters.
640 // Good: "/* /* inner */ outer */ code();"
641 // Bad: "code(); // hello\n world!"
642 pub fn find_comment_end(s: &str) -> Option<usize> {
643 let mut iter = CharClasses::new(s.char_indices());
644 for (kind, (i, _c)) in &mut iter {
645 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
650 // Handle case where the comment ends at the end of s.
651 if iter.status == CharClassesStatus::Normal {
658 /// Returns true if text contains any comment.
659 pub fn contains_comment(text: &str) -> bool {
660 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
663 /// Remove trailing spaces from the specified snippet. We do not remove spaces
664 /// inside strings or comments.
665 pub fn remove_trailing_white_spaces(text: &str) -> String {
666 let mut buffer = String::with_capacity(text.len());
667 let mut space_buffer = String::with_capacity(128);
668 for (char_kind, c) in CharClasses::new(text.chars()) {
671 if char_kind == FullCodeCharKind::InString {
672 buffer.push_str(&space_buffer);
674 space_buffer.clear();
677 _ if c.is_whitespace() => {
678 space_buffer.push(c);
681 if !space_buffer.is_empty() {
682 buffer.push_str(&space_buffer);
683 space_buffer.clear();
692 pub struct CharClasses<T>
698 status: CharClassesStatus,
702 fn get_char(&self) -> char;
705 impl RichChar for char {
706 fn get_char(&self) -> char {
711 impl RichChar for (usize, char) {
712 fn get_char(&self) -> char {
717 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
718 enum CharClassesStatus {
724 // The u32 is the nesting deepness of the comment
726 // Status when the '/' has been consumed, but not yet the '*', deepness is
727 // the new deepness (after the comment opening).
728 BlockCommentOpening(u32),
729 // Status when the '*' has been consumed, but not yet the '/', deepness is
730 // the new deepness (after the comment closing).
731 BlockCommentClosing(u32),
735 /// Distinguish between functional part of code and comments
736 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
737 pub enum CodeCharKind {
742 /// Distinguish between functional part of code and comments,
743 /// describing opening and closing of comments for ease when chunking
744 /// code from tagged characters
745 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
746 pub enum FullCodeCharKind {
748 /// The first character of a comment, there is only one for a comment (always '/')
750 /// Any character inside a comment including the second character of comment
751 /// marks ("//", "/*")
753 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
759 impl FullCodeCharKind {
760 pub fn is_comment(&self) -> bool {
762 FullCodeCharKind::StartComment
763 | FullCodeCharKind::InComment
764 | FullCodeCharKind::EndComment => true,
769 pub fn is_string(&self) -> bool {
770 *self == FullCodeCharKind::InString
773 fn to_codecharkind(&self) -> CodeCharKind {
774 if self.is_comment() {
775 CodeCharKind::Comment
782 impl<T> CharClasses<T>
787 pub fn new(base: T) -> CharClasses<T> {
789 base: multipeek(base),
790 status: CharClassesStatus::Normal,
795 impl<T> Iterator for CharClasses<T>
800 type Item = (FullCodeCharKind, T::Item);
802 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
803 let item = self.base.next()?;
804 let chr = item.get_char();
805 let mut char_kind = FullCodeCharKind::Normal;
806 self.status = match self.status {
807 CharClassesStatus::LitString => match chr {
808 '"' => CharClassesStatus::Normal,
810 char_kind = FullCodeCharKind::InString;
811 CharClassesStatus::LitStringEscape
814 char_kind = FullCodeCharKind::InString;
815 CharClassesStatus::LitString
818 CharClassesStatus::LitStringEscape => {
819 char_kind = FullCodeCharKind::InString;
820 CharClassesStatus::LitString
822 CharClassesStatus::LitChar => match chr {
823 '\\' => CharClassesStatus::LitCharEscape,
824 '\'' => CharClassesStatus::Normal,
825 _ => CharClassesStatus::LitChar,
827 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
828 CharClassesStatus::Normal => match chr {
830 char_kind = FullCodeCharKind::InString;
831 CharClassesStatus::LitString
834 // HACK: Work around mut borrow.
835 match self.base.peek() {
836 Some(next) if next.get_char() == '\\' => {
837 self.status = CharClassesStatus::LitChar;
838 return Some((char_kind, item));
843 match self.base.peek() {
844 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
845 _ => CharClassesStatus::Normal,
848 '/' => match self.base.peek() {
849 Some(next) if next.get_char() == '*' => {
850 self.status = CharClassesStatus::BlockCommentOpening(1);
851 return Some((FullCodeCharKind::StartComment, item));
853 Some(next) if next.get_char() == '/' => {
854 self.status = CharClassesStatus::LineComment;
855 return Some((FullCodeCharKind::StartComment, item));
857 _ => CharClassesStatus::Normal,
859 _ => CharClassesStatus::Normal,
861 CharClassesStatus::BlockComment(deepness) => {
862 assert_ne!(deepness, 0);
863 self.status = match self.base.peek() {
864 Some(next) if next.get_char() == '/' && chr == '*' => {
865 CharClassesStatus::BlockCommentClosing(deepness - 1)
867 Some(next) if next.get_char() == '*' && chr == '/' => {
868 CharClassesStatus::BlockCommentOpening(deepness + 1)
870 _ => CharClassesStatus::BlockComment(deepness),
872 return Some((FullCodeCharKind::InComment, item));
874 CharClassesStatus::BlockCommentOpening(deepness) => {
875 assert_eq!(chr, '*');
876 self.status = CharClassesStatus::BlockComment(deepness);
877 return Some((FullCodeCharKind::InComment, item));
879 CharClassesStatus::BlockCommentClosing(deepness) => {
880 assert_eq!(chr, '/');
882 self.status = CharClassesStatus::Normal;
883 return Some((FullCodeCharKind::EndComment, item));
885 self.status = CharClassesStatus::BlockComment(deepness);
886 return Some((FullCodeCharKind::InComment, item));
889 CharClassesStatus::LineComment => match chr {
891 self.status = CharClassesStatus::Normal;
892 return Some((FullCodeCharKind::EndComment, item));
895 self.status = CharClassesStatus::LineComment;
896 return Some((FullCodeCharKind::InComment, item));
900 Some((char_kind, item))
904 /// Iterator over functional and commented parts of a string. Any part of a string is either
905 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
906 /// comments is functional code. Line comments contain their ending newlines.
907 struct UngroupedCommentCodeSlices<'a> {
909 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
912 impl<'a> UngroupedCommentCodeSlices<'a> {
913 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
914 UngroupedCommentCodeSlices {
916 iter: CharClasses::new(code.char_indices()).peekable(),
921 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
922 type Item = (CodeCharKind, usize, &'a str);
924 fn next(&mut self) -> Option<Self::Item> {
925 let (kind, (start_idx, _)) = self.iter.next()?;
927 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
928 // Consume all the Normal code
929 while let Some(&(char_kind, _)) = self.iter.peek() {
930 if char_kind.is_comment() {
933 let _ = self.iter.next();
936 FullCodeCharKind::StartComment => {
937 // Consume the whole comment
938 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
942 let slice = match self.iter.peek() {
943 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
944 None => &self.slice[start_idx..],
947 if kind.is_comment() {
948 CodeCharKind::Comment
958 /// Iterator over an alternating sequence of functional and commented parts of
959 /// a string. The first item is always a, possibly zero length, subslice of
960 /// functional text. Line style comments contain their ending newlines.
961 pub struct CommentCodeSlices<'a> {
963 last_slice_kind: CodeCharKind,
964 last_slice_end: usize,
967 impl<'a> CommentCodeSlices<'a> {
968 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
971 last_slice_kind: CodeCharKind::Comment,
977 impl<'a> Iterator for CommentCodeSlices<'a> {
978 type Item = (CodeCharKind, usize, &'a str);
980 fn next(&mut self) -> Option<Self::Item> {
981 if self.last_slice_end == self.slice.len() {
985 let mut sub_slice_end = self.last_slice_end;
986 let mut first_whitespace = None;
987 let subslice = &self.slice[self.last_slice_end..];
988 let mut iter = CharClasses::new(subslice.char_indices());
990 for (kind, (i, c)) in &mut iter {
991 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
992 && &subslice[..2] == "//"
993 && [' ', '\t'].contains(&c);
995 if is_comment_connector && first_whitespace.is_none() {
996 first_whitespace = Some(i);
999 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
1000 let last_index = match first_whitespace {
1004 sub_slice_end = self.last_slice_end + last_index;
1008 if !is_comment_connector {
1009 first_whitespace = None;
1013 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1014 // This was the last subslice.
1015 sub_slice_end = match first_whitespace {
1016 Some(i) => self.last_slice_end + i,
1017 None => self.slice.len(),
1021 let kind = match self.last_slice_kind {
1022 CodeCharKind::Comment => CodeCharKind::Normal,
1023 CodeCharKind::Normal => CodeCharKind::Comment,
1027 self.last_slice_end,
1028 &self.slice[self.last_slice_end..sub_slice_end],
1030 self.last_slice_end = sub_slice_end;
1031 self.last_slice_kind = kind;
1037 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1038 /// (if it fits in the width/offset, else return None), else return `new`
1039 pub fn recover_comment_removed(
1042 context: &RewriteContext,
1043 ) -> Option<String> {
1044 let snippet = context.snippet(span);
1045 if snippet != new && changed_comment_content(snippet, &new) {
1046 // We missed some comments. Keep the original text.
1047 Some(snippet.to_owned())
1053 /// Return true if the two strings of code have the same payload of comments.
1054 /// The payload of comments is everything in the string except:
1055 /// - actual code (not comments)
1056 /// - comment start/end marks
1058 /// - '*' at the beginning of lines in block comments
1059 fn changed_comment_content(orig: &str, new: &str) -> bool {
1060 // Cannot write this as a fn since we cannot return types containing closures
1061 let code_comment_content = |code| {
1062 let slices = UngroupedCommentCodeSlices::new(code);
1064 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1065 .flat_map(|(_, _, s)| CommentReducer::new(s))
1067 let res = code_comment_content(orig).ne(code_comment_content(new));
1069 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1073 code_comment_content(orig).collect::<String>(),
1074 code_comment_content(new).collect::<String>()
1079 /// Iterator over the 'payload' characters of a comment.
1080 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1081 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1083 struct CommentReducer<'a> {
1085 at_start_line: bool,
1086 iter: std::str::Chars<'a>,
1089 impl<'a> CommentReducer<'a> {
1090 fn new(comment: &'a str) -> CommentReducer<'a> {
1091 let is_block = comment.starts_with("/*");
1092 let comment = remove_comment_header(comment);
1095 at_start_line: false, // There are no supplementary '*' on the first line
1096 iter: comment.chars(),
1101 impl<'a> Iterator for CommentReducer<'a> {
1103 fn next(&mut self) -> Option<Self::Item> {
1105 let mut c = self.iter.next()?;
1106 if self.is_block && self.at_start_line {
1107 while c.is_whitespace() {
1108 c = self.iter.next()?;
1110 // Ignore leading '*'
1112 c = self.iter.next()?;
1114 } else if c == '\n' {
1115 self.at_start_line = true;
1117 if !c.is_whitespace() {
1124 fn remove_comment_header(comment: &str) -> &str {
1125 if comment.starts_with("///") || comment.starts_with("//!") {
1127 } else if comment.starts_with("//") {
1129 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1130 || comment.starts_with("/*!")
1132 &comment[3..comment.len() - 2]
1135 comment.starts_with("/*"),
1136 format!("string '{}' is not a comment", comment)
1138 &comment[2..comment.len() - 2]
1144 use super::{contains_comment, rewrite_comment, CharClasses, CodeCharKind, CommentCodeSlices,
1145 FindUncommented, FullCodeCharKind};
1146 use shape::{Indent, Shape};
1150 let mut iter = CharClasses::new("//\n\n".chars());
1152 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1153 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1154 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1155 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1156 assert_eq!(None, iter.next());
1160 fn comment_code_slices() {
1161 let input = "code(); /* test */ 1 + 1";
1162 let mut iter = CommentCodeSlices::new(input);
1164 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1166 (CodeCharKind::Comment, 8, "/* test */"),
1167 iter.next().unwrap()
1169 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1170 assert_eq!(None, iter.next());
1174 fn comment_code_slices_two() {
1175 let input = "// comment\n test();";
1176 let mut iter = CommentCodeSlices::new(input);
1178 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1180 (CodeCharKind::Comment, 0, "// comment\n"),
1181 iter.next().unwrap()
1184 (CodeCharKind::Normal, 11, " test();"),
1185 iter.next().unwrap()
1187 assert_eq!(None, iter.next());
1191 fn comment_code_slices_three() {
1192 let input = "1 // comment\n // comment2\n\n";
1193 let mut iter = CommentCodeSlices::new(input);
1195 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1197 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1198 iter.next().unwrap()
1200 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1201 assert_eq!(None, iter.next());
1205 #[cfg_attr(rustfmt, rustfmt_skip)]
1206 fn format_comments() {
1207 let mut config: ::config::Config = Default::default();
1208 config.set().wrap_comments(true);
1209 config.set().normalize_comments(true);
1211 let comment = rewrite_comment(" //test",
1213 Shape::legacy(100, Indent::new(0, 100)),
1215 assert_eq!("/* test */", comment);
1217 let comment = rewrite_comment("// comment on a",
1219 Shape::legacy(10, Indent::empty()),
1221 assert_eq!("// comment\n// on a", comment);
1223 let comment = rewrite_comment("// A multi line comment\n // between args.",
1225 Shape::legacy(60, Indent::new(0, 12)),
1227 assert_eq!("// A multi line comment\n // between args.", comment);
1229 let input = "// comment";
1232 let comment = rewrite_comment(input,
1234 Shape::legacy(9, Indent::new(0, 69)),
1236 assert_eq!(expected, comment);
1238 let comment = rewrite_comment("/* trimmed */",
1240 Shape::legacy(100, Indent::new(0, 100)),
1242 assert_eq!("/* trimmed */", comment);
1245 // This is probably intended to be a non-test fn, but it is not used. I'm
1246 // keeping it around unless it helps us test stuff.
1247 fn uncommented(text: &str) -> String {
1248 CharClasses::new(text.chars())
1249 .filter_map(|(s, c)| match s {
1250 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1257 fn test_uncommented() {
1258 assert_eq!(&uncommented("abc/*...*/"), "abc");
1260 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1263 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1267 fn test_contains_comment() {
1268 assert_eq!(contains_comment("abc"), false);
1269 assert_eq!(contains_comment("abc // qsdf"), true);
1270 assert_eq!(contains_comment("abc /* kqsdf"), true);
1271 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1275 fn test_find_uncommented() {
1276 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1277 assert_eq!(expected, haystack.find_uncommented(needle));
1280 check("/*/ */test", "test", Some(6));
1281 check("//test\ntest", "test", Some(7));
1282 check("/* comment only */", "whatever", None);
1284 "/* comment */ some text /* more commentary */ result",
1288 check("sup // sup", "p", Some(2));
1289 check("sup", "x", None);
1290 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1291 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1292 check("hel/*lohello*/lo", "hello", None);
1293 check("acb", "ab", None);
1294 check(",/*A*/ ", ",", Some(0));
1295 check("abc", "abc", Some(0));
1296 check("/* abc */", "abc", None);
1297 check("/**/abc/* */", "abc", Some(4));
1298 check("\"/* abc */\"", "abc", Some(4));
1299 check("\"/* abc", "abc", Some(4));