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::source_map::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};
23 use {ErrorKind, FormattingError};
25 fn is_custom_comment(comment: &str) -> bool {
26 if !comment.starts_with("//") {
28 } else if let Some(c) = comment.chars().nth(2) {
29 !c.is_alphanumeric() && !c.is_whitespace()
35 #[derive(Copy, Clone, PartialEq, Eq)]
36 pub enum CommentStyle<'a> {
46 fn custom_opener(s: &str) -> &str {
47 s.lines().next().map_or("", |first_line| {
50 .map_or(first_line, |space_index| &first_line[0..=space_index])
54 impl<'a> CommentStyle<'a> {
55 /// Returns true if the commenting style covers a line only.
56 pub fn is_line_comment(&self) -> bool {
58 CommentStyle::DoubleSlash
59 | CommentStyle::TripleSlash
61 | CommentStyle::Custom(_) => true,
66 /// Returns true if the commenting style can span over multiple lines.
67 pub fn is_block_comment(&self) -> bool {
69 CommentStyle::SingleBullet | CommentStyle::DoubleBullet | CommentStyle::Exclamation => {
76 /// Returns true if the commenting style is for documentation.
77 pub fn is_doc_comment(&self) -> bool {
79 CommentStyle::TripleSlash | CommentStyle::Doc => true,
84 pub fn opener(&self) -> &'a str {
86 CommentStyle::DoubleSlash => "// ",
87 CommentStyle::TripleSlash => "/// ",
88 CommentStyle::Doc => "//! ",
89 CommentStyle::SingleBullet => "/* ",
90 CommentStyle::DoubleBullet => "/** ",
91 CommentStyle::Exclamation => "/*! ",
92 CommentStyle::Custom(opener) => opener,
96 pub fn closer(&self) -> &'a str {
98 CommentStyle::DoubleSlash
99 | CommentStyle::TripleSlash
100 | CommentStyle::Custom(..)
101 | CommentStyle::Doc => "",
102 CommentStyle::DoubleBullet => " **/",
103 CommentStyle::SingleBullet | CommentStyle::Exclamation => " */",
107 pub fn line_start(&self) -> &'a str {
109 CommentStyle::DoubleSlash => "// ",
110 CommentStyle::TripleSlash => "/// ",
111 CommentStyle::Doc => "//! ",
112 CommentStyle::SingleBullet | CommentStyle::Exclamation => " * ",
113 CommentStyle::DoubleBullet => " ** ",
114 CommentStyle::Custom(opener) => opener,
118 pub fn to_str_tuplet(&self) -> (&'a str, &'a str, &'a str) {
119 (self.opener(), self.closer(), self.line_start())
123 fn comment_style(orig: &str, normalize_comments: bool) -> CommentStyle {
124 if !normalize_comments {
125 if orig.starts_with("/**") && !orig.starts_with("/**/") {
126 CommentStyle::DoubleBullet
127 } else if orig.starts_with("/*!") {
128 CommentStyle::Exclamation
129 } else if orig.starts_with("/*") {
130 CommentStyle::SingleBullet
131 } else if orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/') {
132 CommentStyle::TripleSlash
133 } else if orig.starts_with("//!") {
135 } else if is_custom_comment(orig) {
136 CommentStyle::Custom(custom_opener(orig))
138 CommentStyle::DoubleSlash
140 } else if (orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/'))
141 || (orig.starts_with("/**") && !orig.starts_with("/**/"))
143 CommentStyle::TripleSlash
144 } else if orig.starts_with("//!") || orig.starts_with("/*!") {
146 } else if is_custom_comment(orig) {
147 CommentStyle::Custom(custom_opener(orig))
149 CommentStyle::DoubleSlash
153 /// Combine `prev_str` and `next_str` into a single `String`. `span` may contain
154 /// comments between two strings. If there are such comments, then that will be
155 /// recovered. If `allow_extend` is true and there is no comment between the two
156 /// strings, then they will be put on a single line as long as doing so does not
157 /// exceed max width.
158 pub fn combine_strs_with_missing_comments(
159 context: &RewriteContext,
165 ) -> Option<String> {
167 String::with_capacity(prev_str.len() + next_str.len() + shape.indent.width() + 128);
168 result.push_str(prev_str);
169 let mut allow_one_line = !prev_str.contains('\n') && !next_str.contains('\n');
170 let first_sep = if prev_str.is_empty() || next_str.is_empty() {
175 let mut one_line_width =
176 last_line_width(prev_str) + first_line_width(next_str) + first_sep.len();
178 let config = context.config;
179 let indent = shape.indent;
180 let missing_comment = rewrite_missing_comment(span, shape, context)?;
182 if missing_comment.is_empty() {
183 if allow_extend && prev_str.len() + first_sep.len() + next_str.len() <= shape.width {
184 result.push_str(first_sep);
185 } else if !prev_str.is_empty() {
186 result.push_str(&indent.to_string_with_newline(config))
188 result.push_str(next_str);
192 // We have a missing comment between the first expression and the second expression.
194 // Peek the the original source code and find out whether there is a newline between the first
195 // expression and the second expression or the missing comment. We will preserve the original
196 // layout whenever possible.
197 let original_snippet = context.snippet(span);
198 let prefer_same_line = if let Some(pos) = original_snippet.find('/') {
199 !original_snippet[..pos].contains('\n')
201 !original_snippet.contains('\n')
204 one_line_width -= first_sep.len();
205 let first_sep = if prev_str.is_empty() || missing_comment.is_empty() {
208 let one_line_width = last_line_width(prev_str) + first_line_width(&missing_comment) + 1;
209 if prefer_same_line && one_line_width <= shape.width {
212 indent.to_string_with_newline(config)
215 result.push_str(&first_sep);
216 result.push_str(&missing_comment);
218 let second_sep = if missing_comment.is_empty() || next_str.is_empty() {
220 } else if missing_comment.starts_with("//") {
221 indent.to_string_with_newline(config)
223 one_line_width += missing_comment.len() + first_sep.len() + 1;
224 allow_one_line &= !missing_comment.starts_with("//") && !missing_comment.contains('\n');
225 if prefer_same_line && allow_one_line && one_line_width <= shape.width {
228 indent.to_string_with_newline(config)
231 result.push_str(&second_sep);
232 result.push_str(next_str);
237 pub fn rewrite_doc_comment(orig: &str, shape: Shape, config: &Config) -> Option<String> {
238 identify_comment(orig, false, shape, config, true)
241 pub fn rewrite_comment(
246 ) -> Option<String> {
247 identify_comment(orig, block_style, shape, config, false)
255 is_doc_comment: bool,
256 ) -> Option<String> {
257 let style = comment_style(orig, false);
259 // Computes the len of line taking into account a newline if the line is part of a paragraph.
260 fn compute_len(orig: &str, line: &str) -> usize {
261 if orig.len() > line.len() {
262 if orig.as_bytes()[line.len()] == b'\r' {
272 // Get the first group of line comments having the same commenting style.
274 // Returns a tuple with:
275 // - a boolean indicating if there is a blank line
276 // - a number indicating the size of the first group of comments
277 fn consume_same_line_comments(
282 let mut first_group_ending = 0;
285 for line in orig.lines() {
286 let trimmed_line = line.trim_left();
287 if trimmed_line.is_empty() {
290 } else if trimmed_line.starts_with(line_start)
291 || comment_style(trimmed_line, false) == style
293 first_group_ending += compute_len(&orig[first_group_ending..], line);
298 (hbl, first_group_ending)
301 let (has_bare_lines, first_group_ending) = match style {
302 CommentStyle::DoubleSlash | CommentStyle::TripleSlash | CommentStyle::Doc => {
303 let line_start = style.line_start().trim_left();
304 consume_same_line_comments(style, orig, line_start)
306 CommentStyle::Custom(opener) => {
307 let trimmed_opener = opener.trim_right();
308 consume_same_line_comments(style, orig, trimmed_opener)
310 // for a block comment, search for the closing symbol
311 CommentStyle::DoubleBullet | CommentStyle::SingleBullet | CommentStyle::Exclamation => {
312 let closer = style.closer().trim_left();
313 let mut closing_symbol_offset = 0;
315 for line in orig.lines() {
316 closing_symbol_offset += compute_len(&orig[closing_symbol_offset..], line);
317 let trimmed_line = line.trim_left();
318 if !trimmed_line.starts_with('*')
319 && !trimmed_line.starts_with("//")
320 && !trimmed_line.starts_with("/*")
324 if trimmed_line.ends_with(closer) {
328 (hbl, closing_symbol_offset)
332 let (first_group, rest) = orig.split_at(first_group_ending);
333 let rewritten_first_group =
334 if !config.normalize_comments() && has_bare_lines && style.is_block_comment() {
335 light_rewrite_block_comment_with_bare_lines(first_group, shape, config)?
336 } else if !config.normalize_comments() && !config.wrap_comments() {
337 light_rewrite_comment(first_group, shape.indent, config, is_doc_comment)?
339 rewrite_comment_inner(
345 is_doc_comment || style.is_doc_comment(),
349 Some(rewritten_first_group)
351 identify_comment(rest.trim_left(), block_style, shape, config, is_doc_comment).map(
355 rewritten_first_group,
356 // insert back the blank line
357 if has_bare_lines && style.is_line_comment() {
362 shape.indent.to_string(config),
370 /// Trims a minimum of leading whitespaces so that the content layout is kept and aligns to indent.
371 fn light_rewrite_block_comment_with_bare_lines(
375 ) -> Option<String> {
376 let prefix_whitespace_min = orig
378 // skip the line with the starting sigil since the leading whitespace is removed
379 // otherwise, the minimum would always be zero
381 .filter(|line| !line.is_empty())
384 for c in line.chars() {
387 '\t' => width += config.tab_spaces(),
395 let indent_str = shape.indent.to_string(config);
396 let mut lines = orig.lines();
397 let first_line = lines.next()?;
403 &line[prefix_whitespace_min..]
406 .collect::<Vec<&str>>()
407 .join(&format!("\n{}", indent_str));
408 Some(format!("{}\n{}{}", first_line, indent_str, rest))
411 /// Attributes for code blocks in rustdoc.
412 /// See https://doc.rust-lang.org/rustdoc/print.html#attributes
413 enum CodeBlockAttribute {
422 impl CodeBlockAttribute {
423 fn new(attribute: &str) -> CodeBlockAttribute {
425 "rust" | "" => CodeBlockAttribute::Rust,
426 "ignore" => CodeBlockAttribute::Ignore,
427 "text" => CodeBlockAttribute::Text,
428 "should_panic" => CodeBlockAttribute::ShouldPanic,
429 "no_run" => CodeBlockAttribute::NoRun,
430 "compile_fail" => CodeBlockAttribute::CompileFail,
431 _ => CodeBlockAttribute::Text,
436 fn rewrite_comment_inner(
442 is_doc_comment: bool,
443 ) -> Option<String> {
444 let (opener, closer, line_start) = if block_style {
445 CommentStyle::SingleBullet.to_str_tuplet()
447 comment_style(orig, config.normalize_comments()).to_str_tuplet()
450 let max_chars = shape
452 .checked_sub(closer.len() + opener.len())
454 let indent_str = shape.indent.to_string_with_newline(config);
455 let fmt_indent = shape.indent + (opener.len() - line_start.len());
456 let mut fmt = StringFormat {
461 shape: Shape::legacy(max_chars, fmt_indent),
466 let line_breaks = count_newlines(orig.trim_right());
470 .map(|(i, mut line)| {
471 line = trim_right_unless_two_whitespaces(line.trim_left(), is_doc_comment);
473 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
474 line = line[..(line.len() - 2)].trim_right();
479 .map(|s| left_trim_comment_line(s, &style))
480 .map(|(line, has_leading_whitespace)| {
481 if orig.starts_with("/*") && line_breaks == 0 {
484 has_leading_whitespace || config.normalize_comments(),
487 (line, has_leading_whitespace || config.normalize_comments())
491 let mut result = String::with_capacity(orig.len() * 2);
492 result.push_str(opener);
493 let mut code_block_buffer = String::with_capacity(128);
494 let mut is_prev_line_multi_line = false;
495 let mut code_block_attr = None;
496 let comment_line_separator = format!("{}{}", indent_str, line_start);
497 let join_code_block_with_comment_line_separator = |s: &str| {
498 let mut result = String::with_capacity(s.len() + 128);
499 let mut iter = s.lines().peekable();
500 while let Some(line) = iter.next() {
501 result.push_str(line);
502 result.push_str(match iter.peek() {
503 Some(next_line) if next_line.is_empty() => comment_line_separator.trim_right(),
504 Some(..) => &comment_line_separator,
511 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
512 let is_last = i == count_newlines(orig);
514 if let Some(ref attr) = code_block_attr {
515 if line.starts_with("```") {
516 let code_block = match attr {
517 CodeBlockAttribute::Ignore | CodeBlockAttribute::Text => {
518 trim_custom_comment_prefix(&code_block_buffer)
520 _ if code_block_buffer.is_empty() => String::new(),
522 let mut config = config.clone();
523 config.set().wrap_comments(false);
524 match ::format_code_block(&code_block_buffer, &config) {
525 Some(ref s) => trim_custom_comment_prefix(s),
526 None => trim_custom_comment_prefix(&code_block_buffer),
530 if !code_block.is_empty() {
531 result.push_str(&comment_line_separator);
532 result.push_str(&join_code_block_with_comment_line_separator(&code_block));
534 code_block_buffer.clear();
535 result.push_str(&comment_line_separator);
536 result.push_str(line);
537 code_block_attr = None;
539 code_block_buffer.push_str(&hide_sharp_behind_comment(line));
540 code_block_buffer.push('\n');
543 // There is a code block that is not properly enclosed by backticks.
544 // We will leave them untouched.
545 result.push_str(&comment_line_separator);
546 result.push_str(&join_code_block_with_comment_line_separator(
547 &trim_custom_comment_prefix(&code_block_buffer),
554 code_block_attr = if line.starts_with("```") {
555 Some(CodeBlockAttribute::new(&line[3..]))
560 if result == opener {
561 let force_leading_whitespace = opener == "/* " && count_newlines(orig) == 0;
562 if !has_leading_whitespace && !force_leading_whitespace && result.ends_with(' ') {
568 } else if is_prev_line_multi_line && !line.is_empty() {
570 } else if is_last && line.is_empty() {
571 // trailing blank lines are unwanted
572 if !closer.is_empty() {
573 result.push_str(&indent_str);
577 result.push_str(&comment_line_separator);
578 if !has_leading_whitespace && result.ends_with(' ') {
584 if config.wrap_comments() && line.len() > fmt.shape.width && !has_url(line) {
585 match rewrite_string(line, &fmt) {
587 is_prev_line_multi_line = s.contains('\n');
590 None if is_prev_line_multi_line => {
591 // We failed to put the current `line` next to the previous `line`.
592 // Remove the trailing space, then start rewrite on the next line.
594 result.push_str(&comment_line_separator);
595 fmt.shape = Shape::legacy(max_chars, fmt_indent);
596 match rewrite_string(line, &fmt) {
598 is_prev_line_multi_line = s.contains('\n');
602 is_prev_line_multi_line = false;
603 result.push_str(line);
608 is_prev_line_multi_line = false;
609 result.push_str(line);
613 fmt.shape = if is_prev_line_multi_line {
615 let offset = 1 + last_line_width(&result) - line_start.len();
617 width: max_chars.saturating_sub(offset),
619 offset: fmt.shape.offset + offset,
622 Shape::legacy(max_chars, fmt_indent)
625 if line.is_empty() && result.ends_with(' ') && !is_last {
626 // Remove space if this is an empty comment or a doc comment.
629 result.push_str(line);
630 fmt.shape = Shape::legacy(max_chars, fmt_indent);
631 is_prev_line_multi_line = false;
635 result.push_str(closer);
636 if result.ends_with(opener) && opener.ends_with(' ') {
644 const RUSTFMT_CUSTOM_COMMENT_PREFIX: &str = "//#### ";
646 fn hide_sharp_behind_comment(s: &str) -> Cow<str> {
647 if s.trim_left().starts_with("# ") {
648 Cow::from(format!("{}{}", RUSTFMT_CUSTOM_COMMENT_PREFIX, s))
654 fn trim_custom_comment_prefix(s: &str) -> String {
657 let left_trimmed = line.trim_left();
658 if left_trimmed.starts_with(RUSTFMT_CUSTOM_COMMENT_PREFIX) {
659 left_trimmed.trim_left_matches(RUSTFMT_CUSTOM_COMMENT_PREFIX)
668 /// Returns true if the given string MAY include URLs or alike.
669 fn has_url(s: &str) -> bool {
670 // This function may return false positive, but should get its job done in most cases.
671 s.contains("https://") || s.contains("http://") || s.contains("ftp://") || s.contains("file://")
674 /// Given the span, rewrite the missing comment inside it if available.
675 /// Note that the given span must only include comments (or leading/trailing whitespaces).
676 pub fn rewrite_missing_comment(
679 context: &RewriteContext,
680 ) -> Option<String> {
681 let missing_snippet = context.snippet(span);
682 let trimmed_snippet = missing_snippet.trim();
683 if !trimmed_snippet.is_empty() {
684 rewrite_comment(trimmed_snippet, false, shape, context.config)
690 /// Recover the missing comments in the specified span, if available.
691 /// The layout of the comments will be preserved as long as it does not break the code
692 /// and its total width does not exceed the max width.
693 pub fn recover_missing_comment_in_span(
696 context: &RewriteContext,
698 ) -> Option<String> {
699 let missing_comment = rewrite_missing_comment(span, shape, context)?;
700 if missing_comment.is_empty() {
703 let missing_snippet = context.snippet(span);
704 let pos = missing_snippet.find('/').unwrap_or(0);
706 let total_width = missing_comment.len() + used_width + 1;
707 let force_new_line_before_comment =
708 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
709 let sep = if force_new_line_before_comment {
710 shape.indent.to_string_with_newline(context.config)
714 Some(format!("{}{}", sep, missing_comment))
718 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
719 fn trim_right_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
720 if is_doc_comment && s.ends_with(" ") {
727 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
728 fn light_rewrite_comment(
732 is_doc_comment: bool,
733 ) -> Option<String> {
734 let lines: Vec<&str> = orig
737 // This is basically just l.trim(), but in the case that a line starts
738 // with `*` we want to leave one space before it, so it aligns with the
740 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
741 let left_trimmed = if let Some(fnw) = first_non_whitespace {
742 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
750 // Preserve markdown's double-space line break syntax in doc comment.
751 trim_right_unless_two_whitespaces(left_trimmed, is_doc_comment)
754 Some(lines.join(&format!("\n{}", offset.to_string(config))))
757 /// Trims comment characters and possibly a single space from the left of a string.
758 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
759 /// this function returns true.
760 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle) -> (&'a str, bool) {
761 if line.starts_with("//! ")
762 || line.starts_with("/// ")
763 || line.starts_with("/*! ")
764 || line.starts_with("/** ")
767 } else if let CommentStyle::Custom(opener) = *style {
768 if line.starts_with(opener) {
769 (&line[opener.len()..], true)
771 (&line[opener.trim_right().len()..], false)
773 } else if line.starts_with("/* ")
774 || line.starts_with("// ")
775 || line.starts_with("//!")
776 || line.starts_with("///")
777 || line.starts_with("** ")
778 || line.starts_with("/*!")
779 || (line.starts_with("/**") && !line.starts_with("/**/"))
781 (&line[3..], line.chars().nth(2).unwrap() == ' ')
782 } else if line.starts_with("/*")
783 || line.starts_with("* ")
784 || line.starts_with("//")
785 || line.starts_with("**")
787 (&line[2..], line.chars().nth(1).unwrap() == ' ')
788 } else if line.starts_with('*') {
791 (line, line.starts_with(' '))
795 pub trait FindUncommented {
796 fn find_uncommented(&self, pat: &str) -> Option<usize>;
799 impl FindUncommented for str {
800 fn find_uncommented(&self, pat: &str) -> Option<usize> {
801 let mut needle_iter = pat.chars();
802 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
803 match needle_iter.next() {
805 return Some(i - pat.len());
807 Some(c) => match kind {
808 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
810 needle_iter = pat.chars();
816 // Handle case where the pattern is a suffix of the search string
817 match needle_iter.next() {
819 None => Some(self.len() - pat.len()),
824 // Returns the first byte position after the first comment. The given string
825 // is expected to be prefixed by a comment, including delimiters.
826 // Good: "/* /* inner */ outer */ code();"
827 // Bad: "code(); // hello\n world!"
828 pub fn find_comment_end(s: &str) -> Option<usize> {
829 let mut iter = CharClasses::new(s.char_indices());
830 for (kind, (i, _c)) in &mut iter {
831 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
836 // Handle case where the comment ends at the end of s.
837 if iter.status == CharClassesStatus::Normal {
844 /// Returns true if text contains any comment.
845 pub fn contains_comment(text: &str) -> bool {
846 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
849 /// Remove trailing spaces from the specified snippet. We do not remove spaces
850 /// inside strings or comments.
851 pub fn remove_trailing_white_spaces(text: &str) -> String {
852 let mut buffer = String::with_capacity(text.len());
853 let mut space_buffer = String::with_capacity(128);
854 for (char_kind, c) in CharClasses::new(text.chars()) {
857 if char_kind == FullCodeCharKind::InString {
858 buffer.push_str(&space_buffer);
860 space_buffer.clear();
863 _ if c.is_whitespace() => {
864 space_buffer.push(c);
867 if !space_buffer.is_empty() {
868 buffer.push_str(&space_buffer);
869 space_buffer.clear();
878 pub struct CharClasses<T>
884 status: CharClassesStatus,
888 fn get_char(&self) -> char;
891 impl RichChar for char {
892 fn get_char(&self) -> char {
897 impl RichChar for (usize, char) {
898 fn get_char(&self) -> char {
903 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
904 enum CharClassesStatus {
909 RawStringPrefix(u32),
910 RawStringSuffix(u32),
913 // The u32 is the nesting deepness of the comment
915 // Status when the '/' has been consumed, but not yet the '*', deepness is
916 // the new deepness (after the comment opening).
917 BlockCommentOpening(u32),
918 // Status when the '*' has been consumed, but not yet the '/', deepness is
919 // the new deepness (after the comment closing).
920 BlockCommentClosing(u32),
924 /// Distinguish between functional part of code and comments
925 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
926 pub enum CodeCharKind {
931 /// Distinguish between functional part of code and comments,
932 /// describing opening and closing of comments for ease when chunking
933 /// code from tagged characters
934 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
935 pub enum FullCodeCharKind {
937 /// The first character of a comment, there is only one for a comment (always '/')
939 /// Any character inside a comment including the second character of comment
940 /// marks ("//", "/*")
942 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
944 /// Start of a mutlitine string
946 /// End of a mutlitine string
952 impl FullCodeCharKind {
953 pub fn is_comment(self) -> bool {
955 FullCodeCharKind::StartComment
956 | FullCodeCharKind::InComment
957 | FullCodeCharKind::EndComment => true,
962 pub fn is_string(self) -> bool {
963 self == FullCodeCharKind::InString || self == FullCodeCharKind::StartString
966 fn to_codecharkind(self) -> CodeCharKind {
967 if self.is_comment() {
968 CodeCharKind::Comment
975 impl<T> CharClasses<T>
980 pub fn new(base: T) -> CharClasses<T> {
982 base: multipeek(base),
983 status: CharClassesStatus::Normal,
988 fn is_raw_string_suffix<T>(iter: &mut MultiPeek<T>, count: u32) -> bool
995 Some(c) if c.get_char() == '#' => continue,
1002 impl<T> Iterator for CharClasses<T>
1007 type Item = (FullCodeCharKind, T::Item);
1009 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
1010 let item = self.base.next()?;
1011 let chr = item.get_char();
1012 let mut char_kind = FullCodeCharKind::Normal;
1013 self.status = match self.status {
1014 CharClassesStatus::LitRawString(sharps) => {
1015 char_kind = FullCodeCharKind::InString;
1019 char_kind = FullCodeCharKind::Normal;
1020 CharClassesStatus::Normal
1021 } else if is_raw_string_suffix(&mut self.base, sharps) {
1022 CharClassesStatus::RawStringSuffix(sharps)
1024 CharClassesStatus::LitRawString(sharps)
1027 _ => CharClassesStatus::LitRawString(sharps),
1030 CharClassesStatus::RawStringPrefix(sharps) => {
1031 char_kind = FullCodeCharKind::InString;
1033 '#' => CharClassesStatus::RawStringPrefix(sharps + 1),
1034 '"' => CharClassesStatus::LitRawString(sharps),
1035 _ => CharClassesStatus::Normal, // Unreachable.
1038 CharClassesStatus::RawStringSuffix(sharps) => {
1042 CharClassesStatus::Normal
1044 char_kind = FullCodeCharKind::InString;
1045 CharClassesStatus::RawStringSuffix(sharps - 1)
1048 _ => CharClassesStatus::Normal, // Unreachable
1051 CharClassesStatus::LitString => {
1052 char_kind = FullCodeCharKind::InString;
1054 '"' => CharClassesStatus::Normal,
1055 '\\' => CharClassesStatus::LitStringEscape,
1056 _ => CharClassesStatus::LitString,
1059 CharClassesStatus::LitStringEscape => {
1060 char_kind = FullCodeCharKind::InString;
1061 CharClassesStatus::LitString
1063 CharClassesStatus::LitChar => match chr {
1064 '\\' => CharClassesStatus::LitCharEscape,
1065 '\'' => CharClassesStatus::Normal,
1066 _ => CharClassesStatus::LitChar,
1068 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
1069 CharClassesStatus::Normal => match chr {
1070 'r' => match self.base.peek().map(|c| c.get_char()) {
1071 Some('#') | Some('"') => {
1072 char_kind = FullCodeCharKind::InString;
1073 CharClassesStatus::RawStringPrefix(0)
1075 _ => CharClassesStatus::Normal,
1078 char_kind = FullCodeCharKind::InString;
1079 CharClassesStatus::LitString
1082 // HACK: Work around mut borrow.
1083 match self.base.peek() {
1084 Some(next) if next.get_char() == '\\' => {
1085 self.status = CharClassesStatus::LitChar;
1086 return Some((char_kind, item));
1091 match self.base.peek() {
1092 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
1093 _ => CharClassesStatus::Normal,
1096 '/' => match self.base.peek() {
1097 Some(next) if next.get_char() == '*' => {
1098 self.status = CharClassesStatus::BlockCommentOpening(1);
1099 return Some((FullCodeCharKind::StartComment, item));
1101 Some(next) if next.get_char() == '/' => {
1102 self.status = CharClassesStatus::LineComment;
1103 return Some((FullCodeCharKind::StartComment, item));
1105 _ => CharClassesStatus::Normal,
1107 _ => CharClassesStatus::Normal,
1109 CharClassesStatus::BlockComment(deepness) => {
1110 assert_ne!(deepness, 0);
1111 self.status = match self.base.peek() {
1112 Some(next) if next.get_char() == '/' && chr == '*' => {
1113 CharClassesStatus::BlockCommentClosing(deepness - 1)
1115 Some(next) if next.get_char() == '*' && chr == '/' => {
1116 CharClassesStatus::BlockCommentOpening(deepness + 1)
1118 _ => CharClassesStatus::BlockComment(deepness),
1120 return Some((FullCodeCharKind::InComment, item));
1122 CharClassesStatus::BlockCommentOpening(deepness) => {
1123 assert_eq!(chr, '*');
1124 self.status = CharClassesStatus::BlockComment(deepness);
1125 return Some((FullCodeCharKind::InComment, item));
1127 CharClassesStatus::BlockCommentClosing(deepness) => {
1128 assert_eq!(chr, '/');
1130 self.status = CharClassesStatus::Normal;
1131 return Some((FullCodeCharKind::EndComment, item));
1133 self.status = CharClassesStatus::BlockComment(deepness);
1134 return Some((FullCodeCharKind::InComment, item));
1137 CharClassesStatus::LineComment => match chr {
1139 self.status = CharClassesStatus::Normal;
1140 return Some((FullCodeCharKind::EndComment, item));
1143 self.status = CharClassesStatus::LineComment;
1144 return Some((FullCodeCharKind::InComment, item));
1148 Some((char_kind, item))
1152 /// An iterator over the lines of a string, paired with the char kind at the
1153 /// end of the line.
1154 pub struct LineClasses<'a> {
1155 base: iter::Peekable<CharClasses<std::str::Chars<'a>>>,
1156 kind: FullCodeCharKind,
1159 impl<'a> LineClasses<'a> {
1160 pub fn new(s: &'a str) -> Self {
1162 base: CharClasses::new(s.chars()).peekable(),
1163 kind: FullCodeCharKind::Normal,
1168 impl<'a> Iterator for LineClasses<'a> {
1169 type Item = (FullCodeCharKind, String);
1171 fn next(&mut self) -> Option<Self::Item> {
1174 let mut line = String::new();
1176 let start_class = match self.base.peek() {
1177 Some((kind, _)) => *kind,
1178 None => FullCodeCharKind::Normal,
1181 while let Some((kind, c)) = self.base.next() {
1183 self.kind = match (start_class, kind) {
1184 (FullCodeCharKind::Normal, FullCodeCharKind::InString) => {
1185 FullCodeCharKind::StartString
1187 (FullCodeCharKind::InString, FullCodeCharKind::Normal) => {
1188 FullCodeCharKind::EndString
1198 Some((self.kind, line))
1202 /// Iterator over functional and commented parts of a string. Any part of a string is either
1203 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
1204 /// comments is functional code. Line comments contain their ending newlines.
1205 struct UngroupedCommentCodeSlices<'a> {
1207 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
1210 impl<'a> UngroupedCommentCodeSlices<'a> {
1211 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
1212 UngroupedCommentCodeSlices {
1214 iter: CharClasses::new(code.char_indices()).peekable(),
1219 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
1220 type Item = (CodeCharKind, usize, &'a str);
1222 fn next(&mut self) -> Option<Self::Item> {
1223 let (kind, (start_idx, _)) = self.iter.next()?;
1225 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
1226 // Consume all the Normal code
1227 while let Some(&(char_kind, _)) = self.iter.peek() {
1228 if char_kind.is_comment() {
1231 let _ = self.iter.next();
1234 FullCodeCharKind::StartComment => {
1235 // Consume the whole comment
1236 while let Some((FullCodeCharKind::InComment, (_, _))) = self.iter.next() {}
1240 let slice = match self.iter.peek() {
1241 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
1242 None => &self.slice[start_idx..],
1245 if kind.is_comment() {
1246 CodeCharKind::Comment
1248 CodeCharKind::Normal
1256 /// Iterator over an alternating sequence of functional and commented parts of
1257 /// a string. The first item is always a, possibly zero length, subslice of
1258 /// functional text. Line style comments contain their ending newlines.
1259 pub struct CommentCodeSlices<'a> {
1261 last_slice_kind: CodeCharKind,
1262 last_slice_end: usize,
1265 impl<'a> CommentCodeSlices<'a> {
1266 pub fn new(slice: &'a str) -> CommentCodeSlices<'a> {
1269 last_slice_kind: CodeCharKind::Comment,
1275 impl<'a> Iterator for CommentCodeSlices<'a> {
1276 type Item = (CodeCharKind, usize, &'a str);
1278 fn next(&mut self) -> Option<Self::Item> {
1279 if self.last_slice_end == self.slice.len() {
1283 let mut sub_slice_end = self.last_slice_end;
1284 let mut first_whitespace = None;
1285 let subslice = &self.slice[self.last_slice_end..];
1286 let mut iter = CharClasses::new(subslice.char_indices());
1288 for (kind, (i, c)) in &mut iter {
1289 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
1290 && &subslice[..2] == "//"
1291 && [' ', '\t'].contains(&c);
1293 if is_comment_connector && first_whitespace.is_none() {
1294 first_whitespace = Some(i);
1297 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
1298 let last_index = match first_whitespace {
1302 sub_slice_end = self.last_slice_end + last_index;
1306 if !is_comment_connector {
1307 first_whitespace = None;
1311 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1312 // This was the last subslice.
1313 sub_slice_end = match first_whitespace {
1314 Some(i) => self.last_slice_end + i,
1315 None => self.slice.len(),
1319 let kind = match self.last_slice_kind {
1320 CodeCharKind::Comment => CodeCharKind::Normal,
1321 CodeCharKind::Normal => CodeCharKind::Comment,
1325 self.last_slice_end,
1326 &self.slice[self.last_slice_end..sub_slice_end],
1328 self.last_slice_end = sub_slice_end;
1329 self.last_slice_kind = kind;
1335 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1336 /// (if it fits in the width/offset, else return None), else return `new`
1337 pub fn recover_comment_removed(
1340 context: &RewriteContext,
1341 ) -> Option<String> {
1342 let snippet = context.snippet(span);
1343 if snippet != new && changed_comment_content(snippet, &new) {
1344 // We missed some comments. Warn and keep the original text.
1345 if context.config.error_on_unformatted() {
1346 context.report.append(
1347 context.source_map.span_to_filename(span).into(),
1348 vec![FormattingError::from_span(
1350 &context.source_map,
1351 ErrorKind::LostComment,
1355 Some(snippet.to_owned())
1361 pub fn filter_normal_code(code: &str) -> String {
1362 let mut buffer = String::with_capacity(code.len());
1363 LineClasses::new(code).for_each(|(kind, line)| match kind {
1364 FullCodeCharKind::Normal
1365 | FullCodeCharKind::StartString
1366 | FullCodeCharKind::InString
1367 | FullCodeCharKind::EndString => {
1368 buffer.push_str(&line);
1373 if !code.ends_with('\n') && buffer.ends_with('\n') {
1379 /// Return true if the two strings of code have the same payload of comments.
1380 /// The payload of comments is everything in the string except:
1381 /// - actual code (not comments)
1382 /// - comment start/end marks
1384 /// - '*' at the beginning of lines in block comments
1385 fn changed_comment_content(orig: &str, new: &str) -> bool {
1386 // Cannot write this as a fn since we cannot return types containing closures
1387 let code_comment_content = |code| {
1388 let slices = UngroupedCommentCodeSlices::new(code);
1390 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1391 .flat_map(|(_, _, s)| CommentReducer::new(s))
1393 let res = code_comment_content(orig).ne(code_comment_content(new));
1395 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1399 code_comment_content(orig).collect::<String>(),
1400 code_comment_content(new).collect::<String>()
1405 /// Iterator over the 'payload' characters of a comment.
1406 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1407 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1409 struct CommentReducer<'a> {
1411 at_start_line: bool,
1412 iter: std::str::Chars<'a>,
1415 impl<'a> CommentReducer<'a> {
1416 fn new(comment: &'a str) -> CommentReducer<'a> {
1417 let is_block = comment.starts_with("/*");
1418 let comment = remove_comment_header(comment);
1421 at_start_line: false, // There are no supplementary '*' on the first line
1422 iter: comment.chars(),
1427 impl<'a> Iterator for CommentReducer<'a> {
1430 fn next(&mut self) -> Option<Self::Item> {
1432 let mut c = self.iter.next()?;
1433 if self.is_block && self.at_start_line {
1434 while c.is_whitespace() {
1435 c = self.iter.next()?;
1437 // Ignore leading '*'
1439 c = self.iter.next()?;
1441 } else if c == '\n' {
1442 self.at_start_line = true;
1444 if !c.is_whitespace() {
1451 fn remove_comment_header(comment: &str) -> &str {
1452 if comment.starts_with("///") || comment.starts_with("//!") {
1454 } else if comment.starts_with("//") {
1456 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1457 || comment.starts_with("/*!")
1459 &comment[3..comment.len() - 2]
1462 comment.starts_with("/*"),
1463 format!("string '{}' is not a comment", comment)
1465 &comment[2..comment.len() - 2]
1472 use shape::{Indent, Shape};
1476 let mut iter = CharClasses::new("//\n\n".chars());
1478 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1479 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1480 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1481 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1482 assert_eq!(None, iter.next());
1486 fn comment_code_slices() {
1487 let input = "code(); /* test */ 1 + 1";
1488 let mut iter = CommentCodeSlices::new(input);
1490 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1492 (CodeCharKind::Comment, 8, "/* test */"),
1493 iter.next().unwrap()
1495 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1496 assert_eq!(None, iter.next());
1500 fn comment_code_slices_two() {
1501 let input = "// comment\n test();";
1502 let mut iter = CommentCodeSlices::new(input);
1504 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1506 (CodeCharKind::Comment, 0, "// comment\n"),
1507 iter.next().unwrap()
1510 (CodeCharKind::Normal, 11, " test();"),
1511 iter.next().unwrap()
1513 assert_eq!(None, iter.next());
1517 fn comment_code_slices_three() {
1518 let input = "1 // comment\n // comment2\n\n";
1519 let mut iter = CommentCodeSlices::new(input);
1521 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1523 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1524 iter.next().unwrap()
1526 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1527 assert_eq!(None, iter.next());
1532 fn format_comments() {
1533 let mut wrap_normalize_config: ::config::Config = Default::default();
1534 wrap_normalize_config.set().wrap_comments(true);
1535 wrap_normalize_config.set().normalize_comments(true);
1537 let mut wrap_config: ::config::Config = Default::default();
1538 wrap_config.set().wrap_comments(true);
1540 let comment = rewrite_comment(" //test",
1542 Shape::legacy(100, Indent::new(0, 100)),
1543 &wrap_normalize_config).unwrap();
1544 assert_eq!("/* test */", comment);
1546 let comment = rewrite_comment("// comment on a",
1548 Shape::legacy(10, Indent::empty()),
1549 &wrap_normalize_config).unwrap();
1550 assert_eq!("// comment\n// on a", comment);
1552 let comment = rewrite_comment("// A multi line comment\n // between args.",
1554 Shape::legacy(60, Indent::new(0, 12)),
1555 &wrap_normalize_config).unwrap();
1556 assert_eq!("// A multi line comment\n // between args.", comment);
1558 let input = "// comment";
1561 let comment = rewrite_comment(input,
1563 Shape::legacy(9, Indent::new(0, 69)),
1564 &wrap_normalize_config).unwrap();
1565 assert_eq!(expected, comment);
1567 let comment = rewrite_comment("/* trimmed */",
1569 Shape::legacy(100, Indent::new(0, 100)),
1570 &wrap_normalize_config).unwrap();
1571 assert_eq!("/* trimmed */", comment);
1573 // check that different comment style are properly recognised
1574 let comment = rewrite_comment(r#"/// test1
1580 Shape::legacy(100, Indent::new(0, 0)),
1581 &wrap_normalize_config).unwrap();
1582 assert_eq!("/// test1\n/// test2\n// test3", comment);
1584 // check that the blank line marks the end of a commented paragraph
1585 let comment = rewrite_comment(r#"// test1
1589 Shape::legacy(100, Indent::new(0, 0)),
1590 &wrap_normalize_config).unwrap();
1591 assert_eq!("// test1\n\n// test2", comment);
1593 // check that the blank line marks the end of a custom-commented paragraph
1594 let comment = rewrite_comment(r#"//@ test1
1598 Shape::legacy(100, Indent::new(0, 0)),
1599 &wrap_normalize_config).unwrap();
1600 assert_eq!("//@ test1\n\n//@ test2", comment);
1602 // check that bare lines are just indented but left unchanged otherwise
1603 let comment = rewrite_comment(r#"// test1
1610 Shape::legacy(100, Indent::new(0, 0)),
1611 &wrap_config).unwrap();
1612 assert_eq!("// test1\n/*\n a bare line!\n\n another bare line!\n*/", comment);
1615 // This is probably intended to be a non-test fn, but it is not used. I'm
1616 // keeping it around unless it helps us test stuff.
1617 fn uncommented(text: &str) -> String {
1618 CharClasses::new(text.chars())
1619 .filter_map(|(s, c)| match s {
1620 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1627 fn test_uncommented() {
1628 assert_eq!(&uncommented("abc/*...*/"), "abc");
1630 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1633 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1637 fn test_contains_comment() {
1638 assert_eq!(contains_comment("abc"), false);
1639 assert_eq!(contains_comment("abc // qsdf"), true);
1640 assert_eq!(contains_comment("abc /* kqsdf"), true);
1641 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1645 fn test_find_uncommented() {
1646 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1647 assert_eq!(expected, haystack.find_uncommented(needle));
1650 check("/*/ */test", "test", Some(6));
1651 check("//test\ntest", "test", Some(7));
1652 check("/* comment only */", "whatever", None);
1654 "/* comment */ some text /* more commentary */ result",
1658 check("sup // sup", "p", Some(2));
1659 check("sup", "x", None);
1660 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1661 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1662 check("hel/*lohello*/lo", "hello", None);
1663 check("acb", "ab", None);
1664 check(",/*A*/ ", ",", Some(0));
1665 check("abc", "abc", Some(0));
1666 check("/* abc */", "abc", None);
1667 check("/**/abc/* */", "abc", Some(4));
1668 check("\"/* abc */\"", "abc", Some(4));
1669 check("\"/* abc", "abc", Some(4));
1673 fn test_remove_trailing_white_spaces() {
1674 let s = " r#\"\n test\n \"#";
1675 assert_eq!(remove_trailing_white_spaces(&s), s);
1679 fn test_filter_normal_code() {
1682 println!("hello, world");
1685 assert_eq!(s, filter_normal_code(s));
1686 let s_with_comment = r#"
1689 println!("hello, world");
1692 assert_eq!(s, filter_normal_code(s_with_comment));