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');
399 inside_code_block = line.starts_with("```rust");
401 if result == opener {
402 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
403 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
409 } else if is_prev_line_multi_line && !line.is_empty() {
411 } else if is_last && !closer.is_empty() && line.is_empty() {
412 result.push_str(&indent_str);
414 result.push_str(&comment_line_separator);
415 if !has_leading_whitespace && result.ends_with(' ') {
421 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
422 match rewrite_string(line, &fmt, Some(max_chars)) {
424 is_prev_line_multi_line = s.contains('\n');
427 None if is_prev_line_multi_line => {
428 // We failed to put the current `line` next to the previous `line`.
429 // Remove the trailing space, then start rewrite on the next line.
431 result.push_str(&comment_line_separator);
432 fmt.shape = Shape::legacy(max_chars, fmt_indent);
433 match rewrite_string(line, &fmt, Some(max_chars)) {
435 is_prev_line_multi_line = s.contains('\n');
439 is_prev_line_multi_line = false;
440 result.push_str(line);
445 is_prev_line_multi_line = false;
446 result.push_str(line);
450 fmt.shape = if is_prev_line_multi_line {
452 let offset = 1 + last_line_width(&result) - line_start.len();
454 width: max_chars.checked_sub(offset).unwrap_or(0),
456 offset: fmt.shape.offset + offset,
459 Shape::legacy(max_chars, fmt_indent)
462 if line.is_empty() && result.ends_with(' ') && !is_last {
463 // Remove space if this is an empty comment or a doc comment.
466 result.push_str(line);
467 fmt.shape = Shape::legacy(max_chars, fmt_indent);
468 is_prev_line_multi_line = false;
472 result.push_str(closer);
473 if result.ends_with(opener) && opener.ends_with(' ') {
481 /// Returns true if the given string MAY include URLs or alike.
482 fn has_url(s: &str) -> bool {
483 // This function may return false positive, but should get its job done in most cases.
484 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
487 /// Given the span, rewrite the missing comment inside it if available.
488 /// Note that the given span must only include comments (or leading/trailing whitespaces).
489 pub fn rewrite_missing_comment(
492 context: &RewriteContext,
493 ) -> Option<String> {
494 let missing_snippet = context.snippet(span);
495 let trimmed_snippet = missing_snippet.trim();
496 if !trimmed_snippet.is_empty() {
497 rewrite_comment(trimmed_snippet, false, shape, context.config)
503 /// Recover the missing comments in the specified span, if available.
504 /// The layout of the comments will be preserved as long as it does not break the code
505 /// and its total width does not exceed the max width.
506 pub fn recover_missing_comment_in_span(
509 context: &RewriteContext,
511 ) -> Option<String> {
512 let missing_comment = rewrite_missing_comment(span, shape, context)?;
513 if missing_comment.is_empty() {
516 let missing_snippet = context.snippet(span);
517 let pos = missing_snippet.chars().position(|c| c == '/').unwrap_or(0);
519 let total_width = missing_comment.len() + used_width + 1;
520 let force_new_line_before_comment =
521 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
522 let sep = if force_new_line_before_comment {
523 shape.indent.to_string_with_newline(context.config)
527 Some(format!("{}{}", sep, missing_comment))
531 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
532 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
533 if is_doc_comment && s.ends_with(" ") {
540 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
541 fn light_rewrite_comment(
545 is_doc_comment: bool,
546 ) -> Option<String> {
547 let lines: Vec<&str> = orig.lines()
549 // This is basically just l.trim(), but in the case that a line starts
550 // with `*` we want to leave one space before it, so it aligns with the
552 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
553 let left_trimmed = if let Some(fnw) = first_non_whitespace {
554 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
562 // Preserve markdown's double-space line break syntax in doc comment.
563 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
566 Some(lines.join(&format!("\n{}", offset.to_string(config))))
569 /// Trims comment characters and possibly a single space from the left of a string.
570 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
571 /// this function returns true.
572 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
573 if line.starts_with("//! ") || line.starts_with("/// ") || line.starts_with("/*! ")
574 || line.starts_with("/** ")
577 } else if let CommentStyle::Custom(opener) = *style {
578 if line.starts_with(opener) {
579 (&line[opener.len()..], true)
581 (&line[opener.trim_right().len()..], false)
583 } else if line.starts_with("/* ") || line.starts_with("// ") || line.starts_with("//!")
584 || line.starts_with("///") || line.starts_with("** ")
585 || line.starts_with("/*!")
586 || (line.starts_with("/**") && !line.starts_with("/**/"))
588 (&line[3..], line.chars().nth(2).unwrap() == ' ')
589 } else if line.starts_with("/*") || line.starts_with("* ") || line.starts_with("//")
590 || line.starts_with("**")
592 (&line[2..], line.chars().nth(1).unwrap() == ' ')
593 } else if line.starts_with('*') {
596 (line, line.starts_with(' '))
600 pub trait FindUncommented {
601 fn find_uncommented(&self, pat: &str) -> Option<usize>;
604 impl FindUncommented for str {
605 fn find_uncommented(&self, pat: &str) -> Option<usize> {
606 let mut needle_iter = pat.chars();
607 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
608 match needle_iter.next() {
610 return Some(i - pat.len());
612 Some(c) => match kind {
613 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
615 needle_iter = pat.chars();
621 // Handle case where the pattern is a suffix of the search string
622 match needle_iter.next() {
624 None => Some(self.len() - pat.len()),
629 // Returns the first byte position after the first comment. The given string
630 // is expected to be prefixed by a comment, including delimiters.
631 // Good: "/* /* inner */ outer */ code();"
632 // Bad: "code(); // hello\n world!"
633 pub fn find_comment_end(s: &str) -> Option<usize> {
634 let mut iter = CharClasses::new(s.char_indices());
635 for (kind, (i, _c)) in &mut iter {
636 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
641 // Handle case where the comment ends at the end of s.
642 if iter.status == CharClassesStatus::Normal {
649 /// Returns true if text contains any comment.
650 pub fn contains_comment(text: &str) -> bool {
651 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
654 /// Remove trailing spaces from the specified snippet. We do not remove spaces
655 /// inside strings or comments.
656 pub fn remove_trailing_white_spaces(text: &str) -> String {
657 let mut buffer = String::with_capacity(text.len());
658 let mut space_buffer = String::with_capacity(128);
659 for (char_kind, c) in CharClasses::new(text.chars()) {
662 if char_kind == FullCodeCharKind::InString {
663 buffer.push_str(&space_buffer);
665 space_buffer.clear();
668 _ if c.is_whitespace() => {
669 space_buffer.push(c);
672 if !space_buffer.is_empty() {
673 buffer.push_str(&space_buffer);
674 space_buffer.clear();
683 pub struct CharClasses<T>
689 status: CharClassesStatus,
693 fn get_char(&self) -> char;
696 impl RichChar for char {
697 fn get_char(&self) -> char {
702 impl RichChar for (usize, char) {
703 fn get_char(&self) -> char {
708 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
709 enum CharClassesStatus {
715 // The u32 is the nesting deepness of the comment
717 // Status when the '/' has been consumed, but not yet the '*', deepness is
718 // the new deepness (after the comment opening).
719 BlockCommentOpening(u32),
720 // Status when the '*' has been consumed, but not yet the '/', deepness is
721 // the new deepness (after the comment closing).
722 BlockCommentClosing(u32),
726 /// Distinguish between functional part of code and comments
727 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
728 pub enum CodeCharKind {
733 /// Distinguish between functional part of code and comments,
734 /// describing opening and closing of comments for ease when chunking
735 /// code from tagged characters
736 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
737 pub enum FullCodeCharKind {
739 /// The first character of a comment, there is only one for a comment (always '/')
741 /// Any character inside a comment including the second character of comment
742 /// marks ("//", "/*")
744 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
750 impl FullCodeCharKind {
751 pub fn is_comment(&self) -> bool {
753 FullCodeCharKind::StartComment
754 | FullCodeCharKind::InComment
755 | FullCodeCharKind::EndComment => true,
760 pub fn is_string(&self) -> bool {
761 *self == FullCodeCharKind::InString
764 fn to_codecharkind(&self) -> CodeCharKind {
765 if self.is_comment() {
766 CodeCharKind::Comment
773 impl<T> CharClasses<T>
778 pub fn new(base: T) -> CharClasses<T> {
780 base: multipeek(base),
781 status: CharClassesStatus::Normal,
786 impl<T> Iterator for CharClasses<T>
791 type Item = (FullCodeCharKind, T::Item);
793 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
794 let item = self.base.next()?;
795 let chr = item.get_char();
796 let mut char_kind = FullCodeCharKind::Normal;
797 self.status = match self.status {
798 CharClassesStatus::LitString => match chr {
799 '"' => CharClassesStatus::Normal,
801 char_kind = FullCodeCharKind::InString;
802 CharClassesStatus::LitStringEscape
805 char_kind = FullCodeCharKind::InString;
806 CharClassesStatus::LitString
809 CharClassesStatus::LitStringEscape => {
810 char_kind = FullCodeCharKind::InString;
811 CharClassesStatus::LitString
813 CharClassesStatus::LitChar => match chr {
814 '\\' => CharClassesStatus::LitCharEscape,
815 '\'' => CharClassesStatus::Normal,
816 _ => CharClassesStatus::LitChar,
818 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
819 CharClassesStatus::Normal => match chr {
821 char_kind = FullCodeCharKind::InString;
822 CharClassesStatus::LitString
825 // HACK: Work around mut borrow.
826 match self.base.peek() {
827 Some(next) if next.get_char() == '\\' => {
828 self.status = CharClassesStatus::LitChar;
829 return Some((char_kind, item));
834 match self.base.peek() {
835 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
836 _ => CharClassesStatus::Normal,
839 '/' => match self.base.peek() {
840 Some(next) if next.get_char() == '*' => {
841 self.status = CharClassesStatus::BlockCommentOpening(1);
842 return Some((FullCodeCharKind::StartComment, item));
844 Some(next) if next.get_char() == '/' => {
845 self.status = CharClassesStatus::LineComment;
846 return Some((FullCodeCharKind::StartComment, item));
848 _ => CharClassesStatus::Normal,
850 _ => CharClassesStatus::Normal,
852 CharClassesStatus::BlockComment(deepness) => {
853 assert_ne!(deepness, 0);
854 self.status = match self.base.peek() {
855 Some(next) if next.get_char() == '/' && chr == '*' => {
856 CharClassesStatus::BlockCommentClosing(deepness - 1)
858 Some(next) if next.get_char() == '*' && chr == '/' => {
859 CharClassesStatus::BlockCommentOpening(deepness + 1)
861 _ => CharClassesStatus::BlockComment(deepness),
863 return Some((FullCodeCharKind::InComment, item));
865 CharClassesStatus::BlockCommentOpening(deepness) => {
866 assert_eq!(chr, '*');
867 self.status = CharClassesStatus::BlockComment(deepness);
868 return Some((FullCodeCharKind::InComment, item));
870 CharClassesStatus::BlockCommentClosing(deepness) => {
871 assert_eq!(chr, '/');
873 self.status = CharClassesStatus::Normal;
874 return Some((FullCodeCharKind::EndComment, item));
876 self.status = CharClassesStatus::BlockComment(deepness);
877 return Some((FullCodeCharKind::InComment, item));
880 CharClassesStatus::LineComment => match chr {
882 self.status = CharClassesStatus::Normal;
883 return Some((FullCodeCharKind::EndComment, item));
886 self.status = CharClassesStatus::LineComment;
887 return Some((FullCodeCharKind::InComment, item));
891 Some((char_kind, item))
895 /// Iterator over functional and commented parts of a string. Any part of a string is either
896 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
897 /// comments is functional code. Line comments contain their ending newlines.
898 struct UngroupedCommentCodeSlices<'a> {
900 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
903 impl<'a> UngroupedCommentCodeSlices<'a> {
904 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
905 UngroupedCommentCodeSlices {
907 iter: CharClasses::new(code.char_indices()).peekable(),
912 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
913 type Item = (CodeCharKind, usize, &'a str);
915 fn next(&mut self) -> Option<Self::Item> {
916 let (kind, (start_idx, _)) = self.iter.next()?;
918 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
919 // Consume all the Normal code
920 while let Some(&(char_kind, _)) = self.iter.peek() {
921 if char_kind.is_comment() {
924 let _ = self.iter.next();
927 FullCodeCharKind::StartComment => {
928 // Consume the whole comment
929 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
933 let slice = match self.iter.peek() {
934 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
935 None => &self.slice[start_idx..],
938 if kind.is_comment() {
939 CodeCharKind::Comment
949 /// Iterator over an alternating sequence of functional and commented parts of
950 /// a string. The first item is always a, possibly zero length, subslice of
951 /// functional text. Line style comments contain their ending newlines.
952 pub struct CommentCodeSlices<'a> {
954 last_slice_kind: CodeCharKind,
955 last_slice_end: usize,
958 impl<'a> CommentCodeSlices<'a> {
959 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
962 last_slice_kind: CodeCharKind::Comment,
968 impl<'a> Iterator for CommentCodeSlices<'a> {
969 type Item = (CodeCharKind, usize, &'a str);
971 fn next(&mut self) -> Option<Self::Item> {
972 if self.last_slice_end == self.slice.len() {
976 let mut sub_slice_end = self.last_slice_end;
977 let mut first_whitespace = None;
978 let subslice = &self.slice[self.last_slice_end..];
979 let mut iter = CharClasses::new(subslice.char_indices());
981 for (kind, (i, c)) in &mut iter {
982 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
983 && &subslice[..2] == "//"
984 && [' ', '\t'].contains(&c);
986 if is_comment_connector && first_whitespace.is_none() {
987 first_whitespace = Some(i);
990 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
991 let last_index = match first_whitespace {
995 sub_slice_end = self.last_slice_end + last_index;
999 if !is_comment_connector {
1000 first_whitespace = None;
1004 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1005 // This was the last subslice.
1006 sub_slice_end = match first_whitespace {
1007 Some(i) => self.last_slice_end + i,
1008 None => self.slice.len(),
1012 let kind = match self.last_slice_kind {
1013 CodeCharKind::Comment => CodeCharKind::Normal,
1014 CodeCharKind::Normal => CodeCharKind::Comment,
1018 self.last_slice_end,
1019 &self.slice[self.last_slice_end..sub_slice_end],
1021 self.last_slice_end = sub_slice_end;
1022 self.last_slice_kind = kind;
1028 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1029 /// (if it fits in the width/offset, else return None), else return `new`
1030 pub fn recover_comment_removed(
1033 context: &RewriteContext,
1034 ) -> Option<String> {
1035 let snippet = context.snippet(span);
1036 if snippet != new && changed_comment_content(snippet, &new) {
1037 // We missed some comments. Keep the original text.
1038 Some(snippet.to_owned())
1044 /// Return true if the two strings of code have the same payload of comments.
1045 /// The payload of comments is everything in the string except:
1046 /// - actual code (not comments)
1047 /// - comment start/end marks
1049 /// - '*' at the beginning of lines in block comments
1050 fn changed_comment_content(orig: &str, new: &str) -> bool {
1051 // Cannot write this as a fn since we cannot return types containing closures
1052 let code_comment_content = |code| {
1053 let slices = UngroupedCommentCodeSlices::new(code);
1055 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1056 .flat_map(|(_, _, s)| CommentReducer::new(s))
1058 let res = code_comment_content(orig).ne(code_comment_content(new));
1060 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1064 code_comment_content(orig).collect::<String>(),
1065 code_comment_content(new).collect::<String>()
1070 /// Iterator over the 'payload' characters of a comment.
1071 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1072 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1074 struct CommentReducer<'a> {
1076 at_start_line: bool,
1077 iter: std::str::Chars<'a>,
1080 impl<'a> CommentReducer<'a> {
1081 fn new(comment: &'a str) -> CommentReducer<'a> {
1082 let is_block = comment.starts_with("/*");
1083 let comment = remove_comment_header(comment);
1086 at_start_line: false, // There are no supplementary '*' on the first line
1087 iter: comment.chars(),
1092 impl<'a> Iterator for CommentReducer<'a> {
1094 fn next(&mut self) -> Option<Self::Item> {
1096 let mut c = self.iter.next()?;
1097 if self.is_block && self.at_start_line {
1098 while c.is_whitespace() {
1099 c = self.iter.next()?;
1101 // Ignore leading '*'
1103 c = self.iter.next()?;
1105 } else if c == '\n' {
1106 self.at_start_line = true;
1108 if !c.is_whitespace() {
1115 fn remove_comment_header(comment: &str) -> &str {
1116 if comment.starts_with("///") || comment.starts_with("//!") {
1118 } else if comment.starts_with("//") {
1120 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1121 || comment.starts_with("/*!")
1123 &comment[3..comment.len() - 2]
1126 comment.starts_with("/*"),
1127 format!("string '{}' is not a comment", comment)
1129 &comment[2..comment.len() - 2]
1135 use super::{contains_comment, rewrite_comment, CharClasses, CodeCharKind, CommentCodeSlices,
1136 FindUncommented, FullCodeCharKind};
1137 use shape::{Indent, Shape};
1141 let mut iter = CharClasses::new("//\n\n".chars());
1143 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1144 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1145 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1146 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1147 assert_eq!(None, iter.next());
1151 fn comment_code_slices() {
1152 let input = "code(); /* test */ 1 + 1";
1153 let mut iter = CommentCodeSlices::new(input);
1155 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1157 (CodeCharKind::Comment, 8, "/* test */"),
1158 iter.next().unwrap()
1160 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1161 assert_eq!(None, iter.next());
1165 fn comment_code_slices_two() {
1166 let input = "// comment\n test();";
1167 let mut iter = CommentCodeSlices::new(input);
1169 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1171 (CodeCharKind::Comment, 0, "// comment\n"),
1172 iter.next().unwrap()
1175 (CodeCharKind::Normal, 11, " test();"),
1176 iter.next().unwrap()
1178 assert_eq!(None, iter.next());
1182 fn comment_code_slices_three() {
1183 let input = "1 // comment\n // comment2\n\n";
1184 let mut iter = CommentCodeSlices::new(input);
1186 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1188 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1189 iter.next().unwrap()
1191 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1192 assert_eq!(None, iter.next());
1196 #[cfg_attr(rustfmt, rustfmt_skip)]
1197 fn format_comments() {
1198 let mut config: ::config::Config = Default::default();
1199 config.set().wrap_comments(true);
1200 config.set().normalize_comments(true);
1202 let comment = rewrite_comment(" //test",
1204 Shape::legacy(100, Indent::new(0, 100)),
1206 assert_eq!("/* test */", comment);
1208 let comment = rewrite_comment("// comment on a",
1210 Shape::legacy(10, Indent::empty()),
1212 assert_eq!("// comment\n// on a", comment);
1214 let comment = rewrite_comment("// A multi line comment\n // between args.",
1216 Shape::legacy(60, Indent::new(0, 12)),
1218 assert_eq!("// A multi line comment\n // between args.", comment);
1220 let input = "// comment";
1223 let comment = rewrite_comment(input,
1225 Shape::legacy(9, Indent::new(0, 69)),
1227 assert_eq!(expected, comment);
1229 let comment = rewrite_comment("/* trimmed */",
1231 Shape::legacy(100, Indent::new(0, 100)),
1233 assert_eq!("/* trimmed */", comment);
1236 // This is probably intended to be a non-test fn, but it is not used. I'm
1237 // keeping it around unless it helps us test stuff.
1238 fn uncommented(text: &str) -> String {
1239 CharClasses::new(text.chars())
1240 .filter_map(|(s, c)| match s {
1241 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1248 fn test_uncommented() {
1249 assert_eq!(&uncommented("abc/*...*/"), "abc");
1251 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1254 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1258 fn test_contains_comment() {
1259 assert_eq!(contains_comment("abc"), false);
1260 assert_eq!(contains_comment("abc // qsdf"), true);
1261 assert_eq!(contains_comment("abc /* kqsdf"), true);
1262 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1266 fn test_find_uncommented() {
1267 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1268 assert_eq!(expected, haystack.find_uncommented(needle));
1271 check("/*/ */test", "test", Some(6));
1272 check("//test\ntest", "test", Some(7));
1273 check("/* comment only */", "whatever", None);
1275 "/* comment */ some text /* more commentary */ result",
1279 check("sup // sup", "p", Some(2));
1280 check("sup", "x", None);
1281 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1282 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1283 check("hel/*lohello*/lo", "hello", None);
1284 check("acb", "ab", None);
1285 check(",/*A*/ ", ",", Some(0));
1286 check("abc", "abc", Some(0));
1287 check("/* abc */", "abc", None);
1288 check("/**/abc/* */", "abc", Some(4));
1289 check("\"/* abc */\"", "abc", Some(4));
1290 check("\"/* abc", "abc", Some(4));