1 // Formatting and tools for comments.
3 use std::{self, borrow::Cow, iter};
5 use itertools::{multipeek, MultiPeek};
6 use lazy_static::lazy_static;
10 use crate::config::Config;
11 use crate::rewrite::RewriteContext;
12 use crate::shape::{Indent, Shape};
13 use crate::string::{rewrite_string, StringFormat};
15 count_newlines, first_line_width, last_line_width, trim_left_preserve_layout,
16 trimmed_last_line_width, unicode_str_width,
18 use crate::{ErrorKind, FormattingError};
21 /// A regex matching reference doc links.
26 /// /// [example]: this::is::a::link
28 static ref REFERENCE_LINK_URL: Regex = Regex::new(r"^\[.+\]\s?:").unwrap();
31 fn is_custom_comment(comment: &str) -> bool {
32 if !comment.starts_with("//") {
34 } else if let Some(c) = comment.chars().nth(2) {
35 !c.is_alphanumeric() && !c.is_whitespace()
41 #[derive(Copy, Clone, PartialEq, Eq)]
42 pub(crate) enum CommentStyle<'a> {
52 fn custom_opener(s: &str) -> &str {
53 s.lines().next().map_or("", |first_line| {
56 .map_or(first_line, |space_index| &first_line[0..=space_index])
60 impl<'a> CommentStyle<'a> {
61 /// Returns `true` if the commenting style covers a line only.
62 pub(crate) fn is_line_comment(&self) -> bool {
64 CommentStyle::DoubleSlash
65 | CommentStyle::TripleSlash
67 | CommentStyle::Custom(_) => true,
72 /// Returns `true` if the commenting style can span over multiple lines.
73 pub(crate) fn is_block_comment(&self) -> bool {
75 CommentStyle::SingleBullet | CommentStyle::DoubleBullet | CommentStyle::Exclamation => {
82 /// Returns `true` if the commenting style is for documentation.
83 pub(crate) fn is_doc_comment(&self) -> bool {
84 matches!(*self, CommentStyle::TripleSlash | CommentStyle::Doc)
87 pub(crate) fn opener(&self) -> &'a str {
89 CommentStyle::DoubleSlash => "// ",
90 CommentStyle::TripleSlash => "/// ",
91 CommentStyle::Doc => "//! ",
92 CommentStyle::SingleBullet => "/* ",
93 CommentStyle::DoubleBullet => "/** ",
94 CommentStyle::Exclamation => "/*! ",
95 CommentStyle::Custom(opener) => opener,
99 pub(crate) fn closer(&self) -> &'a str {
101 CommentStyle::DoubleSlash
102 | CommentStyle::TripleSlash
103 | CommentStyle::Custom(..)
104 | CommentStyle::Doc => "",
105 CommentStyle::SingleBullet | CommentStyle::DoubleBullet | CommentStyle::Exclamation => {
111 pub(crate) fn line_start(&self) -> &'a str {
113 CommentStyle::DoubleSlash => "// ",
114 CommentStyle::TripleSlash => "/// ",
115 CommentStyle::Doc => "//! ",
116 CommentStyle::SingleBullet | CommentStyle::DoubleBullet | CommentStyle::Exclamation => {
119 CommentStyle::Custom(opener) => opener,
123 pub(crate) fn to_str_tuplet(&self) -> (&'a str, &'a str, &'a str) {
124 (self.opener(), self.closer(), self.line_start())
128 pub(crate) fn comment_style(orig: &str, normalize_comments: bool) -> CommentStyle<'_> {
129 if !normalize_comments {
130 if orig.starts_with("/**") && !orig.starts_with("/**/") {
131 CommentStyle::DoubleBullet
132 } else if orig.starts_with("/*!") {
133 CommentStyle::Exclamation
134 } else if orig.starts_with("/*") {
135 CommentStyle::SingleBullet
136 } else if orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/') {
137 CommentStyle::TripleSlash
138 } else if orig.starts_with("//!") {
140 } else if is_custom_comment(orig) {
141 CommentStyle::Custom(custom_opener(orig))
143 CommentStyle::DoubleSlash
145 } else if (orig.starts_with("///") && orig.chars().nth(3).map_or(true, |c| c != '/'))
146 || (orig.starts_with("/**") && !orig.starts_with("/**/"))
148 CommentStyle::TripleSlash
149 } else if orig.starts_with("//!") || orig.starts_with("/*!") {
151 } else if is_custom_comment(orig) {
152 CommentStyle::Custom(custom_opener(orig))
154 CommentStyle::DoubleSlash
158 /// Returns true if the last line of the passed string finishes with a block-comment.
159 pub(crate) fn is_last_comment_block(s: &str) -> bool {
160 s.trim_end().ends_with("*/")
163 /// Combine `prev_str` and `next_str` into a single `String`. `span` may contain
164 /// comments between two strings. If there are such comments, then that will be
165 /// recovered. If `allow_extend` is true and there is no comment between the two
166 /// strings, then they will be put on a single line as long as doing so does not
167 /// exceed max width.
168 pub(crate) fn combine_strs_with_missing_comments(
169 context: &RewriteContext<'_>,
175 ) -> Option<String> {
177 "combine_strs_with_missing_comments `{}` `{}` {:?} {:?}",
185 String::with_capacity(prev_str.len() + next_str.len() + shape.indent.width() + 128);
186 result.push_str(prev_str);
187 let mut allow_one_line = !prev_str.contains('\n') && !next_str.contains('\n');
189 if prev_str.is_empty() || next_str.is_empty() || trimmed_last_line_width(prev_str) == 0 {
194 let mut one_line_width =
195 last_line_width(prev_str) + first_line_width(next_str) + first_sep.len();
197 let config = context.config;
198 let indent = shape.indent;
199 let missing_comment = rewrite_missing_comment(span, shape, context)?;
201 if missing_comment.is_empty() {
202 if allow_extend && one_line_width <= shape.width {
203 result.push_str(first_sep);
204 } else if !prev_str.is_empty() {
205 result.push_str(&indent.to_string_with_newline(config))
207 result.push_str(next_str);
211 // We have a missing comment between the first expression and the second expression.
213 // Peek the the original source code and find out whether there is a newline between the first
214 // expression and the second expression or the missing comment. We will preserve the original
215 // layout whenever possible.
216 let original_snippet = context.snippet(span);
217 let prefer_same_line = if let Some(pos) = original_snippet.find('/') {
218 !original_snippet[..pos].contains('\n')
220 !original_snippet.contains('\n')
223 one_line_width -= first_sep.len();
224 let first_sep = if prev_str.is_empty() || missing_comment.is_empty() {
227 let one_line_width = last_line_width(prev_str) + first_line_width(&missing_comment) + 1;
228 if prefer_same_line && one_line_width <= shape.width {
231 indent.to_string_with_newline(config)
234 result.push_str(&first_sep);
235 result.push_str(&missing_comment);
237 let second_sep = if missing_comment.is_empty() || next_str.is_empty() {
239 } else if missing_comment.starts_with("//") {
240 indent.to_string_with_newline(config)
242 one_line_width += missing_comment.len() + first_sep.len() + 1;
243 allow_one_line &= !missing_comment.starts_with("//") && !missing_comment.contains('\n');
244 if prefer_same_line && allow_one_line && one_line_width <= shape.width {
247 indent.to_string_with_newline(config)
250 result.push_str(&second_sep);
251 result.push_str(next_str);
256 pub(crate) fn rewrite_doc_comment(orig: &str, shape: Shape, config: &Config) -> Option<String> {
257 identify_comment(orig, false, shape, config, true)
260 pub(crate) fn rewrite_comment(
265 ) -> Option<String> {
266 identify_comment(orig, block_style, shape, config, false)
274 is_doc_comment: bool,
275 ) -> Option<String> {
276 let style = comment_style(orig, false);
278 // Computes the byte length of line taking into account a newline if the line is part of a
280 fn compute_len(orig: &str, line: &str) -> usize {
281 if orig.len() > line.len() {
282 if orig.as_bytes()[line.len()] == b'\r' {
292 // Get the first group of line comments having the same commenting style.
294 // Returns a tuple with:
295 // - a boolean indicating if there is a blank line
296 // - a number indicating the size of the first group of comments
297 fn consume_same_line_comments(
298 style: CommentStyle<'_>,
302 let mut first_group_ending = 0;
305 for line in orig.lines() {
306 let trimmed_line = line.trim_start();
307 if trimmed_line.is_empty() {
310 } else if trimmed_line.starts_with(line_start)
311 || comment_style(trimmed_line, false) == style
313 first_group_ending += compute_len(&orig[first_group_ending..], line);
318 (hbl, first_group_ending)
321 let (has_bare_lines, first_group_ending) = match style {
322 CommentStyle::DoubleSlash | CommentStyle::TripleSlash | CommentStyle::Doc => {
323 let line_start = style.line_start().trim_start();
324 consume_same_line_comments(style, orig, line_start)
326 CommentStyle::Custom(opener) => {
327 let trimmed_opener = opener.trim_end();
328 consume_same_line_comments(style, orig, trimmed_opener)
330 // for a block comment, search for the closing symbol
331 CommentStyle::DoubleBullet | CommentStyle::SingleBullet | CommentStyle::Exclamation => {
332 let closer = style.closer().trim_start();
333 let mut count = orig.matches(closer).count();
334 let mut closing_symbol_offset = 0;
336 let mut first = true;
337 for line in orig.lines() {
338 closing_symbol_offset += compute_len(&orig[closing_symbol_offset..], line);
339 let mut trimmed_line = line.trim_start();
340 if !trimmed_line.starts_with('*')
341 && !trimmed_line.starts_with("//")
342 && !trimmed_line.starts_with("/*")
347 // Remove opener from consideration when searching for closer
349 let opener = style.opener().trim_end();
350 trimmed_line = &trimmed_line[opener.len()..];
353 if trimmed_line.ends_with(closer) {
360 (hbl, closing_symbol_offset)
364 let (first_group, rest) = orig.split_at(first_group_ending);
365 let rewritten_first_group =
366 if !config.normalize_comments() && has_bare_lines && style.is_block_comment() {
367 trim_left_preserve_layout(first_group, shape.indent, config)?
368 } else if !config.normalize_comments()
369 && !config.wrap_comments()
370 && !config.format_code_in_doc_comments()
372 light_rewrite_comment(first_group, shape.indent, config, is_doc_comment)
374 rewrite_comment_inner(
380 is_doc_comment || style.is_doc_comment(),
384 Some(rewritten_first_group)
396 rewritten_first_group,
397 // insert back the blank line
398 if has_bare_lines && style.is_line_comment() {
403 shape.indent.to_string(config),
410 /// Enum indicating if the code block contains rust based on attributes
411 enum CodeBlockAttribute {
416 impl CodeBlockAttribute {
417 /// Parse comma separated attributes list. Return rust only if all
418 /// attributes are valid rust attributes
419 /// See <https://doc.rust-lang.org/rustdoc/print.html#attributes>
420 fn new(attributes: &str) -> CodeBlockAttribute {
421 for attribute in attributes.split(',') {
422 match attribute.trim() {
423 "" | "rust" | "should_panic" | "no_run" | "edition2015" | "edition2018"
424 | "edition2021" => (),
425 "ignore" | "compile_fail" | "text" => return CodeBlockAttribute::NotRust,
426 _ => return CodeBlockAttribute::NotRust,
429 CodeBlockAttribute::Rust
433 /// Block that is formatted as an item.
435 /// An item starts with either a star `*` a dash `-` or a greater-than `>`.
436 /// Different level of indentation are handled by shrinking the shape accordingly.
437 struct ItemizedBlock {
438 /// the lines that are identified as part of an itemized block
440 /// the number of characters (typically whitespaces) up to the item sigil
442 /// the string that marks the start of an item
444 /// sequence of characters (typically whitespaces) to prefix new lines that are part of the item
449 /// Returns `true` if the line is formatted as an item
450 fn is_itemized_line(line: &str) -> bool {
451 let trimmed = line.trim_start();
452 trimmed.starts_with("* ") || trimmed.starts_with("- ") || trimmed.starts_with("> ")
455 /// Creates a new ItemizedBlock described with the given line.
456 /// The `is_itemized_line` needs to be called first.
457 fn new(line: &str) -> ItemizedBlock {
458 let space_to_sigil = line.chars().take_while(|c| c.is_whitespace()).count();
459 // +2 = '* ', which will add the appropriate amount of whitespace to keep itemized
460 // content formatted correctly.
461 let mut indent = space_to_sigil + 2;
462 let mut line_start = " ".repeat(indent);
464 // Markdown blockquote start with a "> "
465 if line.trim_start().starts_with(">") {
466 // remove the original +2 indent because there might be multiple nested block quotes
467 // and it's easier to reason about the final indent by just taking the length
468 // of th new line_start. We update the indent because it effects the max width
469 // of each formatted line.
470 line_start = itemized_block_quote_start(line, line_start, 2);
471 indent = line_start.len();
474 lines: vec![line[indent..].to_string()],
476 opener: line[..indent].to_string(),
481 /// Returns a `StringFormat` used for formatting the content of an item.
482 fn create_string_format<'a>(&'a self, fmt: &'a StringFormat<'_>) -> StringFormat<'a> {
488 shape: Shape::legacy(fmt.shape.width.saturating_sub(self.indent), Indent::empty()),
494 /// Returns `true` if the line is part of the current itemized block.
495 /// If it is, then it is added to the internal lines list.
496 fn add_line(&mut self, line: &str) -> bool {
497 if !ItemizedBlock::is_itemized_line(line)
498 && self.indent <= line.chars().take_while(|c| c.is_whitespace()).count()
500 self.lines.push(line.to_string());
506 /// Returns the block as a string, with each line trimmed at the start.
507 fn trimmed_block_as_string(&self) -> String {
510 .map(|line| format!("{} ", line.trim_start()))
514 /// Returns the block as a string under its original form.
515 fn original_block_as_string(&self) -> String {
516 self.lines.join("\n")
520 /// Determine the line_start when formatting markdown block quotes.
521 /// The original line_start likely contains indentation (whitespaces), which we'd like to
522 /// replace with '> ' characters.
523 fn itemized_block_quote_start(line: &str, mut line_start: String, remove_indent: usize) -> String {
524 let quote_level = line
526 .take_while(|c| !c.is_alphanumeric())
527 .fold(0, |acc, c| if c == '>' { acc + 1 } else { acc });
529 for _ in 0..remove_indent {
533 for _ in 0..quote_level {
534 line_start.push_str("> ")
539 struct CommentRewrite<'a> {
541 code_block_buffer: String,
542 is_prev_line_multi_line: bool,
543 code_block_attr: Option<CodeBlockAttribute>,
544 item_block: Option<ItemizedBlock>,
545 comment_line_separator: String,
549 fmt: StringFormat<'a>,
554 style: CommentStyle<'a>,
557 impl<'a> CommentRewrite<'a> {
563 ) -> CommentRewrite<'a> {
564 let ((opener, closer, line_start), style) = if block_style {
566 CommentStyle::SingleBullet.to_str_tuplet(),
567 CommentStyle::SingleBullet,
570 let style = comment_style(orig, config.normalize_comments());
571 (style.to_str_tuplet(), style)
574 let max_width = shape
576 .checked_sub(closer.len() + opener.len())
578 let indent_str = shape.indent.to_string_with_newline(config).to_string();
580 let mut cr = CommentRewrite {
581 result: String::with_capacity(orig.len() * 2),
582 code_block_buffer: String::with_capacity(128),
583 is_prev_line_multi_line: false,
584 code_block_attr: None,
586 comment_line_separator: format!("{}{}", indent_str, line_start),
589 fmt_indent: shape.indent,
596 shape: Shape::legacy(max_width, shape.indent),
601 opener: opener.to_owned(),
602 closer: closer.to_owned(),
603 line_start: line_start.to_owned(),
606 cr.result.push_str(opener);
610 fn join_block(s: &str, sep: &str) -> String {
611 let mut result = String::with_capacity(s.len() + 128);
612 let mut iter = s.lines().peekable();
613 while let Some(line) = iter.next() {
614 result.push_str(line);
615 result.push_str(match iter.peek() {
616 Some(next_line) if next_line.is_empty() => sep.trim_end(),
624 /// Check if any characters were written to the result buffer after the start of the comment.
625 /// when calling [`CommentRewrite::new()`] the result buffer is initiazlied with the opening
626 /// characters for the comment.
627 fn buffer_contains_comment(&self) -> bool {
628 // if self.result.len() < self.opener.len() then an empty comment is in the buffer
629 // if self.result.len() > self.opener.len() then a non empty comment is in the buffer
630 self.result.len() != self.opener.len()
633 fn finish(mut self) -> String {
634 if !self.code_block_buffer.is_empty() {
635 // There is a code block that is not properly enclosed by backticks.
636 // We will leave them untouched.
637 self.result.push_str(&self.comment_line_separator);
638 self.result.push_str(&Self::join_block(
639 &trim_custom_comment_prefix(&self.code_block_buffer),
640 &self.comment_line_separator,
644 if let Some(ref ib) = self.item_block {
645 // the last few lines are part of an itemized block
646 self.fmt.shape = Shape::legacy(self.max_width, self.fmt_indent);
647 let item_fmt = ib.create_string_format(&self.fmt);
649 // only push a comment_line_separator for ItemizedBlocks if the comment is not empty
650 if self.buffer_contains_comment() {
651 self.result.push_str(&self.comment_line_separator);
654 self.result.push_str(&ib.opener);
655 match rewrite_string(
656 &ib.trimmed_block_as_string(),
658 self.max_width.saturating_sub(ib.indent),
660 Some(s) => self.result.push_str(&Self::join_block(
662 &format!("{}{}", self.comment_line_separator, ib.line_start),
664 None => self.result.push_str(&Self::join_block(
665 &ib.original_block_as_string(),
666 &self.comment_line_separator,
671 self.result.push_str(&self.closer);
672 if self.result.ends_with(&self.opener) && self.opener.ends_with(' ') {
685 has_leading_whitespace: bool,
686 is_doc_comment: bool,
688 let num_newlines = count_newlines(orig);
689 let is_last = i == num_newlines;
690 let needs_new_comment_line = if self.style.is_block_comment() {
691 num_newlines > 0 || self.buffer_contains_comment()
693 self.buffer_contains_comment()
696 if let Some(ref mut ib) = self.item_block {
697 if ib.add_line(line) {
700 self.is_prev_line_multi_line = false;
701 self.fmt.shape = Shape::legacy(self.max_width, self.fmt_indent);
702 let item_fmt = ib.create_string_format(&self.fmt);
704 // only push a comment_line_separator if we need to start a new comment line
705 if needs_new_comment_line {
706 self.result.push_str(&self.comment_line_separator);
709 self.result.push_str(&ib.opener);
710 match rewrite_string(
711 &ib.trimmed_block_as_string(),
713 self.max_width.saturating_sub(ib.indent),
715 Some(s) => self.result.push_str(&Self::join_block(
717 &format!("{}{}", self.comment_line_separator, ib.line_start),
719 None => self.result.push_str(&Self::join_block(
720 &ib.original_block_as_string(),
721 &self.comment_line_separator,
724 } else if self.code_block_attr.is_some() {
725 if line.starts_with("```") {
726 let code_block = match self.code_block_attr.as_ref().unwrap() {
727 CodeBlockAttribute::Rust
728 if self.fmt.config.format_code_in_doc_comments()
729 && !self.code_block_buffer.is_empty() =>
731 let mut config = self.fmt.config.clone();
732 config.set().wrap_comments(false);
734 crate::format_code_block(&self.code_block_buffer, &config, false)
736 trim_custom_comment_prefix(&s.snippet)
738 trim_custom_comment_prefix(&self.code_block_buffer)
741 _ => trim_custom_comment_prefix(&self.code_block_buffer),
743 if !code_block.is_empty() {
744 self.result.push_str(&self.comment_line_separator);
746 .push_str(&Self::join_block(&code_block, &self.comment_line_separator));
748 self.code_block_buffer.clear();
749 self.result.push_str(&self.comment_line_separator);
750 self.result.push_str(line);
751 self.code_block_attr = None;
753 self.code_block_buffer
754 .push_str(&hide_sharp_behind_comment(line));
755 self.code_block_buffer.push('\n');
760 self.code_block_attr = None;
761 self.item_block = None;
762 if let Some(stripped) = line.strip_prefix("```") {
763 self.code_block_attr = Some(CodeBlockAttribute::new(stripped))
764 } else if self.fmt.config.wrap_comments() && ItemizedBlock::is_itemized_line(line) {
765 let ib = ItemizedBlock::new(line);
766 self.item_block = Some(ib);
770 if self.result == self.opener {
771 let force_leading_whitespace = &self.opener == "/* " && count_newlines(orig) == 0;
772 if !has_leading_whitespace && !force_leading_whitespace && self.result.ends_with(' ') {
778 } else if self.is_prev_line_multi_line && !line.is_empty() {
779 self.result.push(' ')
780 } else if is_last && line.is_empty() {
781 // trailing blank lines are unwanted
782 if !self.closer.is_empty() {
783 self.result.push_str(&self.indent_str);
787 self.result.push_str(&self.comment_line_separator);
788 if !has_leading_whitespace && self.result.ends_with(' ') {
793 let is_markdown_header_doc_comment = is_doc_comment && line.starts_with("#");
795 // We only want to wrap the comment if:
796 // 1) wrap_comments = true is configured
797 // 2) The comment is not the start of a markdown header doc comment
798 // 3) The comment width exceeds the shape's width
799 // 4) No URLS were found in the commnet
800 let should_wrap_comment = self.fmt.config.wrap_comments()
801 && !is_markdown_header_doc_comment
802 && unicode_str_width(line) > self.fmt.shape.width
805 if should_wrap_comment {
806 match rewrite_string(line, &self.fmt, self.max_width) {
808 self.is_prev_line_multi_line = s.contains('\n');
809 self.result.push_str(s);
811 None if self.is_prev_line_multi_line => {
812 // We failed to put the current `line` next to the previous `line`.
813 // Remove the trailing space, then start rewrite on the next line.
815 self.result.push_str(&self.comment_line_separator);
816 self.fmt.shape = Shape::legacy(self.max_width, self.fmt_indent);
817 match rewrite_string(line, &self.fmt, self.max_width) {
819 self.is_prev_line_multi_line = s.contains('\n');
820 self.result.push_str(s);
823 self.is_prev_line_multi_line = false;
824 self.result.push_str(line);
829 self.is_prev_line_multi_line = false;
830 self.result.push_str(line);
834 self.fmt.shape = if self.is_prev_line_multi_line {
836 let offset = 1 + last_line_width(&self.result) - self.line_start.len();
838 width: self.max_width.saturating_sub(offset),
839 indent: self.fmt_indent,
840 offset: self.fmt.shape.offset + offset,
843 Shape::legacy(self.max_width, self.fmt_indent)
846 if line.is_empty() && self.result.ends_with(' ') && !is_last {
847 // Remove space if this is an empty comment or a doc comment.
850 self.result.push_str(line);
851 self.fmt.shape = Shape::legacy(self.max_width, self.fmt_indent);
852 self.is_prev_line_multi_line = false;
859 fn rewrite_comment_inner(
862 style: CommentStyle<'_>,
865 is_doc_comment: bool,
866 ) -> Option<String> {
867 let mut rewriter = CommentRewrite::new(orig, block_style, shape, config);
869 let line_breaks = count_newlines(orig.trim_end());
873 .map(|(i, mut line)| {
874 line = trim_end_unless_two_whitespaces(line.trim_start(), is_doc_comment);
876 if i == line_breaks && line.ends_with("*/") && !line.starts_with("//") {
877 line = line[..(line.len() - 2)].trim_end();
882 .map(|s| left_trim_comment_line(s, &style))
883 .map(|(line, has_leading_whitespace)| {
884 if orig.starts_with("/*") && line_breaks == 0 {
887 has_leading_whitespace || config.normalize_comments(),
890 (line, has_leading_whitespace || config.normalize_comments())
894 for (i, (line, has_leading_whitespace)) in lines.enumerate() {
895 if rewriter.handle_line(orig, i, line, has_leading_whitespace, is_doc_comment) {
900 Some(rewriter.finish())
903 const RUSTFMT_CUSTOM_COMMENT_PREFIX: &str = "//#### ";
905 fn hide_sharp_behind_comment(s: &str) -> Cow<'_, str> {
906 let s_trimmed = s.trim();
907 if s_trimmed.starts_with("# ") || s_trimmed == "#" {
908 Cow::from(format!("{}{}", RUSTFMT_CUSTOM_COMMENT_PREFIX, s))
914 fn trim_custom_comment_prefix(s: &str) -> String {
917 let left_trimmed = line.trim_start();
918 if left_trimmed.starts_with(RUSTFMT_CUSTOM_COMMENT_PREFIX) {
919 left_trimmed.trim_start_matches(RUSTFMT_CUSTOM_COMMENT_PREFIX)
928 /// Returns `true` if the given string MAY include URLs or alike.
929 fn has_url(s: &str) -> bool {
930 // This function may return false positive, but should get its job done in most cases.
931 s.contains("https://")
932 || s.contains("http://")
933 || s.contains("ftp://")
934 || s.contains("file://")
935 || REFERENCE_LINK_URL.is_match(s)
938 /// Given the span, rewrite the missing comment inside it if available.
939 /// Note that the given span must only include comments (or leading/trailing whitespaces).
940 pub(crate) fn rewrite_missing_comment(
943 context: &RewriteContext<'_>,
944 ) -> Option<String> {
945 let missing_snippet = context.snippet(span);
946 let trimmed_snippet = missing_snippet.trim();
947 // check the span starts with a comment
948 let pos = trimmed_snippet.find('/');
949 if !trimmed_snippet.is_empty() && pos.is_some() {
950 rewrite_comment(trimmed_snippet, false, shape, context.config)
956 /// Recover the missing comments in the specified span, if available.
957 /// The layout of the comments will be preserved as long as it does not break the code
958 /// and its total width does not exceed the max width.
959 pub(crate) fn recover_missing_comment_in_span(
962 context: &RewriteContext<'_>,
964 ) -> Option<String> {
965 let missing_comment = rewrite_missing_comment(span, shape, context)?;
966 if missing_comment.is_empty() {
969 let missing_snippet = context.snippet(span);
970 let pos = missing_snippet.find('/')?;
972 let total_width = missing_comment.len() + used_width + 1;
973 let force_new_line_before_comment =
974 missing_snippet[..pos].contains('\n') || total_width > context.config.max_width();
975 let sep = if force_new_line_before_comment {
976 shape.indent.to_string_with_newline(context.config)
980 Some(format!("{}{}", sep, missing_comment))
984 /// Trim trailing whitespaces unless they consist of two or more whitespaces.
985 fn trim_end_unless_two_whitespaces(s: &str, is_doc_comment: bool) -> &str {
986 if is_doc_comment && s.ends_with(" ") {
993 /// Trims whitespace and aligns to indent, but otherwise does not change comments.
994 fn light_rewrite_comment(
998 is_doc_comment: bool,
1000 let lines: Vec<&str> = orig
1003 // This is basically just l.trim(), but in the case that a line starts
1004 // with `*` we want to leave one space before it, so it aligns with the
1006 let first_non_whitespace = l.find(|c| !char::is_whitespace(c));
1007 let left_trimmed = if let Some(fnw) = first_non_whitespace {
1008 if l.as_bytes()[fnw] == b'*' && fnw > 0 {
1016 // Preserve markdown's double-space line break syntax in doc comment.
1017 trim_end_unless_two_whitespaces(left_trimmed, is_doc_comment)
1020 lines.join(&format!("\n{}", offset.to_string(config)))
1023 /// Trims comment characters and possibly a single space from the left of a string.
1024 /// Does not trim all whitespace. If a single space is trimmed from the left of the string,
1025 /// this function returns true.
1026 fn left_trim_comment_line<'a>(line: &'a str, style: &CommentStyle<'_>) -> (&'a str, bool) {
1027 if line.starts_with("//! ")
1028 || line.starts_with("/// ")
1029 || line.starts_with("/*! ")
1030 || line.starts_with("/** ")
1033 } else if let CommentStyle::Custom(opener) = *style {
1034 if let Some(stripped) = line.strip_prefix(opener) {
1037 (&line[opener.trim_end().len()..], false)
1039 } else if line.starts_with("/* ")
1040 || line.starts_with("// ")
1041 || line.starts_with("//!")
1042 || line.starts_with("///")
1043 || line.starts_with("** ")
1044 || line.starts_with("/*!")
1045 || (line.starts_with("/**") && !line.starts_with("/**/"))
1047 (&line[3..], line.chars().nth(2).unwrap() == ' ')
1048 } else if line.starts_with("/*")
1049 || line.starts_with("* ")
1050 || line.starts_with("//")
1051 || line.starts_with("**")
1053 (&line[2..], line.chars().nth(1).unwrap() == ' ')
1054 } else if let Some(stripped) = line.strip_prefix('*') {
1057 (line, line.starts_with(' '))
1061 pub(crate) trait FindUncommented {
1062 fn find_uncommented(&self, pat: &str) -> Option<usize>;
1063 fn find_last_uncommented(&self, pat: &str) -> Option<usize>;
1066 impl FindUncommented for str {
1067 fn find_uncommented(&self, pat: &str) -> Option<usize> {
1068 let mut needle_iter = pat.chars();
1069 for (kind, (i, b)) in CharClasses::new(self.char_indices()) {
1070 match needle_iter.next() {
1072 return Some(i - pat.len());
1074 Some(c) => match kind {
1075 FullCodeCharKind::Normal | FullCodeCharKind::InString if b == c => {}
1077 needle_iter = pat.chars();
1083 // Handle case where the pattern is a suffix of the search string
1084 match needle_iter.next() {
1086 None => Some(self.len() - pat.len()),
1090 fn find_last_uncommented(&self, pat: &str) -> Option<usize> {
1091 if let Some(left) = self.find_uncommented(pat) {
1092 let mut result = left;
1093 // add 1 to use find_last_uncommented for &str after pat
1094 while let Some(next) = self[(result + 1)..].find_last_uncommented(pat) {
1104 // Returns the first byte position after the first comment. The given string
1105 // is expected to be prefixed by a comment, including delimiters.
1106 // Good: `/* /* inner */ outer */ code();`
1107 // Bad: `code(); // hello\n world!`
1108 pub(crate) fn find_comment_end(s: &str) -> Option<usize> {
1109 let mut iter = CharClasses::new(s.char_indices());
1110 for (kind, (i, _c)) in &mut iter {
1111 if kind == FullCodeCharKind::Normal || kind == FullCodeCharKind::InString {
1116 // Handle case where the comment ends at the end of `s`.
1117 if iter.status == CharClassesStatus::Normal {
1124 /// Returns `true` if text contains any comment.
1125 pub(crate) fn contains_comment(text: &str) -> bool {
1126 CharClasses::new(text.chars()).any(|(kind, _)| kind.is_comment())
1129 pub(crate) struct CharClasses<T>
1135 status: CharClassesStatus,
1138 pub(crate) trait RichChar {
1139 fn get_char(&self) -> char;
1142 impl RichChar for char {
1143 fn get_char(&self) -> char {
1148 impl RichChar for (usize, char) {
1149 fn get_char(&self) -> char {
1154 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
1155 enum CharClassesStatus {
1157 /// Character is within a string
1160 /// Character is within a raw string
1162 RawStringPrefix(u32),
1163 RawStringSuffix(u32),
1166 /// Character inside a block comment, with the integer indicating the nesting deepness of the
1169 /// Character inside a block-commented string, with the integer indicating the nesting deepness
1171 StringInBlockComment(u32),
1172 /// Status when the '/' has been consumed, but not yet the '*', deepness is
1173 /// the new deepness (after the comment opening).
1174 BlockCommentOpening(u32),
1175 /// Status when the '*' has been consumed, but not yet the '/', deepness is
1176 /// the new deepness (after the comment closing).
1177 BlockCommentClosing(u32),
1178 /// Character is within a line comment
1182 /// Distinguish between functional part of code and comments
1183 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
1184 pub(crate) enum CodeCharKind {
1189 /// Distinguish between functional part of code and comments,
1190 /// describing opening and closing of comments for ease when chunking
1191 /// code from tagged characters
1192 #[derive(PartialEq, Eq, Debug, Clone, Copy)]
1193 pub(crate) enum FullCodeCharKind {
1195 /// The first character of a comment, there is only one for a comment (always '/')
1197 /// Any character inside a comment including the second character of comment
1198 /// marks ("//", "/*")
1200 /// Last character of a comment, '\n' for a line comment, '/' for a block comment.
1202 /// Start of a mutlitine string inside a comment
1203 StartStringCommented,
1204 /// End of a mutlitine string inside a comment
1206 /// Inside a commented string
1208 /// Start of a mutlitine string
1210 /// End of a mutlitine string
1212 /// Inside a string.
1216 impl FullCodeCharKind {
1217 pub(crate) fn is_comment(self) -> bool {
1219 FullCodeCharKind::StartComment
1220 | FullCodeCharKind::InComment
1221 | FullCodeCharKind::EndComment
1222 | FullCodeCharKind::StartStringCommented
1223 | FullCodeCharKind::InStringCommented
1224 | FullCodeCharKind::EndStringCommented => true,
1229 /// Returns true if the character is inside a comment
1230 pub(crate) fn inside_comment(self) -> bool {
1232 FullCodeCharKind::InComment
1233 | FullCodeCharKind::StartStringCommented
1234 | FullCodeCharKind::InStringCommented
1235 | FullCodeCharKind::EndStringCommented => true,
1240 pub(crate) fn is_string(self) -> bool {
1241 self == FullCodeCharKind::InString || self == FullCodeCharKind::StartString
1244 /// Returns true if the character is within a commented string
1245 pub(crate) fn is_commented_string(self) -> bool {
1246 self == FullCodeCharKind::InStringCommented
1247 || self == FullCodeCharKind::StartStringCommented
1250 fn to_codecharkind(self) -> CodeCharKind {
1251 if self.is_comment() {
1252 CodeCharKind::Comment
1254 CodeCharKind::Normal
1259 impl<T> CharClasses<T>
1264 pub(crate) fn new(base: T) -> CharClasses<T> {
1266 base: multipeek(base),
1267 status: CharClassesStatus::Normal,
1272 fn is_raw_string_suffix<T>(iter: &mut MultiPeek<T>, count: u32) -> bool
1279 Some(c) if c.get_char() == '#' => continue,
1286 impl<T> Iterator for CharClasses<T>
1291 type Item = (FullCodeCharKind, T::Item);
1293 fn next(&mut self) -> Option<(FullCodeCharKind, T::Item)> {
1294 let item = self.base.next()?;
1295 let chr = item.get_char();
1296 let mut char_kind = FullCodeCharKind::Normal;
1297 self.status = match self.status {
1298 CharClassesStatus::LitRawString(sharps) => {
1299 char_kind = FullCodeCharKind::InString;
1303 char_kind = FullCodeCharKind::Normal;
1304 CharClassesStatus::Normal
1305 } else if is_raw_string_suffix(&mut self.base, sharps) {
1306 CharClassesStatus::RawStringSuffix(sharps)
1308 CharClassesStatus::LitRawString(sharps)
1311 _ => CharClassesStatus::LitRawString(sharps),
1314 CharClassesStatus::RawStringPrefix(sharps) => {
1315 char_kind = FullCodeCharKind::InString;
1317 '#' => CharClassesStatus::RawStringPrefix(sharps + 1),
1318 '"' => CharClassesStatus::LitRawString(sharps),
1319 _ => CharClassesStatus::Normal, // Unreachable.
1322 CharClassesStatus::RawStringSuffix(sharps) => {
1326 CharClassesStatus::Normal
1328 char_kind = FullCodeCharKind::InString;
1329 CharClassesStatus::RawStringSuffix(sharps - 1)
1332 _ => CharClassesStatus::Normal, // Unreachable
1335 CharClassesStatus::LitString => {
1336 char_kind = FullCodeCharKind::InString;
1338 '"' => CharClassesStatus::Normal,
1339 '\\' => CharClassesStatus::LitStringEscape,
1340 _ => CharClassesStatus::LitString,
1343 CharClassesStatus::LitStringEscape => {
1344 char_kind = FullCodeCharKind::InString;
1345 CharClassesStatus::LitString
1347 CharClassesStatus::LitChar => match chr {
1348 '\\' => CharClassesStatus::LitCharEscape,
1349 '\'' => CharClassesStatus::Normal,
1350 _ => CharClassesStatus::LitChar,
1352 CharClassesStatus::LitCharEscape => CharClassesStatus::LitChar,
1353 CharClassesStatus::Normal => match chr {
1354 'r' => match self.base.peek().map(RichChar::get_char) {
1355 Some('#') | Some('"') => {
1356 char_kind = FullCodeCharKind::InString;
1357 CharClassesStatus::RawStringPrefix(0)
1359 _ => CharClassesStatus::Normal,
1362 char_kind = FullCodeCharKind::InString;
1363 CharClassesStatus::LitString
1366 // HACK: Work around mut borrow.
1367 match self.base.peek() {
1368 Some(next) if next.get_char() == '\\' => {
1369 self.status = CharClassesStatus::LitChar;
1370 return Some((char_kind, item));
1375 match self.base.peek() {
1376 Some(next) if next.get_char() == '\'' => CharClassesStatus::LitChar,
1377 _ => CharClassesStatus::Normal,
1380 '/' => match self.base.peek() {
1381 Some(next) if next.get_char() == '*' => {
1382 self.status = CharClassesStatus::BlockCommentOpening(1);
1383 return Some((FullCodeCharKind::StartComment, item));
1385 Some(next) if next.get_char() == '/' => {
1386 self.status = CharClassesStatus::LineComment;
1387 return Some((FullCodeCharKind::StartComment, item));
1389 _ => CharClassesStatus::Normal,
1391 _ => CharClassesStatus::Normal,
1393 CharClassesStatus::StringInBlockComment(deepness) => {
1394 char_kind = FullCodeCharKind::InStringCommented;
1396 CharClassesStatus::BlockComment(deepness)
1397 } else if chr == '*' && self.base.peek().map(RichChar::get_char) == Some('/') {
1398 char_kind = FullCodeCharKind::InComment;
1399 CharClassesStatus::BlockCommentClosing(deepness - 1)
1401 CharClassesStatus::StringInBlockComment(deepness)
1404 CharClassesStatus::BlockComment(deepness) => {
1405 assert_ne!(deepness, 0);
1406 char_kind = FullCodeCharKind::InComment;
1407 match self.base.peek() {
1408 Some(next) if next.get_char() == '/' && chr == '*' => {
1409 CharClassesStatus::BlockCommentClosing(deepness - 1)
1411 Some(next) if next.get_char() == '*' && chr == '/' => {
1412 CharClassesStatus::BlockCommentOpening(deepness + 1)
1414 _ if chr == '"' => CharClassesStatus::StringInBlockComment(deepness),
1418 CharClassesStatus::BlockCommentOpening(deepness) => {
1419 assert_eq!(chr, '*');
1420 self.status = CharClassesStatus::BlockComment(deepness);
1421 return Some((FullCodeCharKind::InComment, item));
1423 CharClassesStatus::BlockCommentClosing(deepness) => {
1424 assert_eq!(chr, '/');
1426 self.status = CharClassesStatus::Normal;
1427 return Some((FullCodeCharKind::EndComment, item));
1429 self.status = CharClassesStatus::BlockComment(deepness);
1430 return Some((FullCodeCharKind::InComment, item));
1433 CharClassesStatus::LineComment => match chr {
1435 self.status = CharClassesStatus::Normal;
1436 return Some((FullCodeCharKind::EndComment, item));
1439 self.status = CharClassesStatus::LineComment;
1440 return Some((FullCodeCharKind::InComment, item));
1444 Some((char_kind, item))
1448 /// An iterator over the lines of a string, paired with the char kind at the
1449 /// end of the line.
1450 pub(crate) struct LineClasses<'a> {
1451 base: iter::Peekable<CharClasses<std::str::Chars<'a>>>,
1452 kind: FullCodeCharKind,
1455 impl<'a> LineClasses<'a> {
1456 pub(crate) fn new(s: &'a str) -> Self {
1458 base: CharClasses::new(s.chars()).peekable(),
1459 kind: FullCodeCharKind::Normal,
1464 impl<'a> Iterator for LineClasses<'a> {
1465 type Item = (FullCodeCharKind, String);
1467 fn next(&mut self) -> Option<Self::Item> {
1470 let mut line = String::new();
1472 let start_kind = match self.base.peek() {
1473 Some((kind, _)) => *kind,
1474 None => unreachable!(),
1477 for (kind, c) in self.base.by_ref() {
1478 // needed to set the kind of the ending character on the last line
1481 self.kind = match (start_kind, kind) {
1482 (FullCodeCharKind::Normal, FullCodeCharKind::InString) => {
1483 FullCodeCharKind::StartString
1485 (FullCodeCharKind::InString, FullCodeCharKind::Normal) => {
1486 FullCodeCharKind::EndString
1488 (FullCodeCharKind::InComment, FullCodeCharKind::InStringCommented) => {
1489 FullCodeCharKind::StartStringCommented
1491 (FullCodeCharKind::InStringCommented, FullCodeCharKind::InComment) => {
1492 FullCodeCharKind::EndStringCommented
1501 // Workaround for CRLF newline.
1502 if line.ends_with('\r') {
1506 Some((self.kind, line))
1510 /// Iterator over functional and commented parts of a string. Any part of a string is either
1511 /// functional code, either *one* block comment, either *one* line comment. Whitespace between
1512 /// comments is functional code. Line comments contain their ending newlines.
1513 struct UngroupedCommentCodeSlices<'a> {
1515 iter: iter::Peekable<CharClasses<std::str::CharIndices<'a>>>,
1518 impl<'a> UngroupedCommentCodeSlices<'a> {
1519 fn new(code: &'a str) -> UngroupedCommentCodeSlices<'a> {
1520 UngroupedCommentCodeSlices {
1522 iter: CharClasses::new(code.char_indices()).peekable(),
1527 impl<'a> Iterator for UngroupedCommentCodeSlices<'a> {
1528 type Item = (CodeCharKind, usize, &'a str);
1530 fn next(&mut self) -> Option<Self::Item> {
1531 let (kind, (start_idx, _)) = self.iter.next()?;
1533 FullCodeCharKind::Normal | FullCodeCharKind::InString => {
1534 // Consume all the Normal code
1535 while let Some(&(char_kind, _)) = self.iter.peek() {
1536 if char_kind.is_comment() {
1539 let _ = self.iter.next();
1542 FullCodeCharKind::StartComment => {
1543 // Consume the whole comment
1545 match self.iter.next() {
1546 Some((kind, ..)) if kind.inside_comment() => continue,
1553 let slice = match self.iter.peek() {
1554 Some(&(_, (end_idx, _))) => &self.slice[start_idx..end_idx],
1555 None => &self.slice[start_idx..],
1558 if kind.is_comment() {
1559 CodeCharKind::Comment
1561 CodeCharKind::Normal
1569 /// Iterator over an alternating sequence of functional and commented parts of
1570 /// a string. The first item is always a, possibly zero length, subslice of
1571 /// functional text. Line style comments contain their ending newlines.
1572 pub(crate) struct CommentCodeSlices<'a> {
1574 last_slice_kind: CodeCharKind,
1575 last_slice_end: usize,
1578 impl<'a> CommentCodeSlices<'a> {
1579 pub(crate) fn new(slice: &'a str) -> CommentCodeSlices<'a> {
1582 last_slice_kind: CodeCharKind::Comment,
1588 impl<'a> Iterator for CommentCodeSlices<'a> {
1589 type Item = (CodeCharKind, usize, &'a str);
1591 fn next(&mut self) -> Option<Self::Item> {
1592 if self.last_slice_end == self.slice.len() {
1596 let mut sub_slice_end = self.last_slice_end;
1597 let mut first_whitespace = None;
1598 let subslice = &self.slice[self.last_slice_end..];
1599 let mut iter = CharClasses::new(subslice.char_indices());
1601 for (kind, (i, c)) in &mut iter {
1602 let is_comment_connector = self.last_slice_kind == CodeCharKind::Normal
1603 && &subslice[..2] == "//"
1604 && [' ', '\t'].contains(&c);
1606 if is_comment_connector && first_whitespace.is_none() {
1607 first_whitespace = Some(i);
1610 if kind.to_codecharkind() == self.last_slice_kind && !is_comment_connector {
1611 let last_index = match first_whitespace {
1615 sub_slice_end = self.last_slice_end + last_index;
1619 if !is_comment_connector {
1620 first_whitespace = None;
1624 if let (None, true) = (iter.next(), sub_slice_end == self.last_slice_end) {
1625 // This was the last subslice.
1626 sub_slice_end = match first_whitespace {
1627 Some(i) => self.last_slice_end + i,
1628 None => self.slice.len(),
1632 let kind = match self.last_slice_kind {
1633 CodeCharKind::Comment => CodeCharKind::Normal,
1634 CodeCharKind::Normal => CodeCharKind::Comment,
1638 self.last_slice_end,
1639 &self.slice[self.last_slice_end..sub_slice_end],
1641 self.last_slice_end = sub_slice_end;
1642 self.last_slice_kind = kind;
1648 /// Checks is `new` didn't miss any comment from `span`, if it removed any, return previous text
1649 /// (if it fits in the width/offset, else return `None`), else return `new`
1650 pub(crate) fn recover_comment_removed(
1653 context: &RewriteContext<'_>,
1654 ) -> Option<String> {
1655 let snippet = context.snippet(span);
1656 if snippet != new && changed_comment_content(snippet, &new) {
1657 // We missed some comments. Warn and keep the original text.
1658 if context.config.error_on_unformatted() {
1659 context.report.append(
1660 context.parse_sess.span_to_filename(span),
1661 vec![FormattingError::from_span(
1664 ErrorKind::LostComment,
1668 Some(snippet.to_owned())
1674 pub(crate) fn filter_normal_code(code: &str) -> String {
1675 let mut buffer = String::with_capacity(code.len());
1676 LineClasses::new(code).for_each(|(kind, line)| match kind {
1677 FullCodeCharKind::Normal
1678 | FullCodeCharKind::StartString
1679 | FullCodeCharKind::InString
1680 | FullCodeCharKind::EndString => {
1681 buffer.push_str(&line);
1686 if !code.ends_with('\n') && buffer.ends_with('\n') {
1692 /// Returns `true` if the two strings of code have the same payload of comments.
1693 /// The payload of comments is everything in the string except:
1694 /// - actual code (not comments),
1695 /// - comment start/end marks,
1697 /// - '*' at the beginning of lines in block comments.
1698 fn changed_comment_content(orig: &str, new: &str) -> bool {
1699 // Cannot write this as a fn since we cannot return types containing closures.
1700 let code_comment_content = |code| {
1701 let slices = UngroupedCommentCodeSlices::new(code);
1703 .filter(|&(ref kind, _, _)| *kind == CodeCharKind::Comment)
1704 .flat_map(|(_, _, s)| CommentReducer::new(s))
1706 let res = code_comment_content(orig).ne(code_comment_content(new));
1708 "comment::changed_comment_content: {}\norig: '{}'\nnew: '{}'\nraw_old: {}\nraw_new: {}",
1712 code_comment_content(orig).collect::<String>(),
1713 code_comment_content(new).collect::<String>()
1718 /// Iterator over the 'payload' characters of a comment.
1719 /// It skips whitespace, comment start/end marks, and '*' at the beginning of lines.
1720 /// The comment must be one comment, ie not more than one start mark (no multiple line comments,
1722 struct CommentReducer<'a> {
1724 at_start_line: bool,
1725 iter: std::str::Chars<'a>,
1728 impl<'a> CommentReducer<'a> {
1729 fn new(comment: &'a str) -> CommentReducer<'a> {
1730 let is_block = comment.starts_with("/*");
1731 let comment = remove_comment_header(comment);
1734 // There are no supplementary '*' on the first line.
1735 at_start_line: false,
1736 iter: comment.chars(),
1741 impl<'a> Iterator for CommentReducer<'a> {
1744 fn next(&mut self) -> Option<Self::Item> {
1746 let mut c = self.iter.next()?;
1747 if self.is_block && self.at_start_line {
1748 while c.is_whitespace() {
1749 c = self.iter.next()?;
1751 // Ignore leading '*'.
1753 c = self.iter.next()?;
1755 } else if c == '\n' {
1756 self.at_start_line = true;
1758 if !c.is_whitespace() {
1765 fn remove_comment_header(comment: &str) -> &str {
1766 if comment.starts_with("///") || comment.starts_with("//!") {
1768 } else if let Some(stripped) = comment.strip_prefix("//") {
1770 } else if (comment.starts_with("/**") && !comment.starts_with("/**/"))
1771 || comment.starts_with("/*!")
1773 &comment[3..comment.len() - 2]
1776 comment.starts_with("/*"),
1777 "string '{}' is not a comment",
1780 &comment[2..comment.len() - 2]
1787 use crate::shape::{Indent, Shape};
1791 let mut iter = CharClasses::new("//\n\n".chars());
1793 assert_eq!((FullCodeCharKind::StartComment, '/'), iter.next().unwrap());
1794 assert_eq!((FullCodeCharKind::InComment, '/'), iter.next().unwrap());
1795 assert_eq!((FullCodeCharKind::EndComment, '\n'), iter.next().unwrap());
1796 assert_eq!((FullCodeCharKind::Normal, '\n'), iter.next().unwrap());
1797 assert_eq!(None, iter.next());
1801 fn comment_code_slices() {
1802 let input = "code(); /* test */ 1 + 1";
1803 let mut iter = CommentCodeSlices::new(input);
1805 assert_eq!((CodeCharKind::Normal, 0, "code(); "), iter.next().unwrap());
1807 (CodeCharKind::Comment, 8, "/* test */"),
1808 iter.next().unwrap()
1810 assert_eq!((CodeCharKind::Normal, 18, " 1 + 1"), iter.next().unwrap());
1811 assert_eq!(None, iter.next());
1815 fn comment_code_slices_two() {
1816 let input = "// comment\n test();";
1817 let mut iter = CommentCodeSlices::new(input);
1819 assert_eq!((CodeCharKind::Normal, 0, ""), iter.next().unwrap());
1821 (CodeCharKind::Comment, 0, "// comment\n"),
1822 iter.next().unwrap()
1825 (CodeCharKind::Normal, 11, " test();"),
1826 iter.next().unwrap()
1828 assert_eq!(None, iter.next());
1832 fn comment_code_slices_three() {
1833 let input = "1 // comment\n // comment2\n\n";
1834 let mut iter = CommentCodeSlices::new(input);
1836 assert_eq!((CodeCharKind::Normal, 0, "1 "), iter.next().unwrap());
1838 (CodeCharKind::Comment, 2, "// comment\n // comment2\n"),
1839 iter.next().unwrap()
1841 assert_eq!((CodeCharKind::Normal, 29, "\n"), iter.next().unwrap());
1842 assert_eq!(None, iter.next());
1847 fn format_doc_comments() {
1848 let mut wrap_normalize_config: crate::config::Config = Default::default();
1849 wrap_normalize_config.set().wrap_comments(true);
1850 wrap_normalize_config.set().normalize_comments(true);
1852 let mut wrap_config: crate::config::Config = Default::default();
1853 wrap_config.set().wrap_comments(true);
1855 let comment = rewrite_comment(" //test",
1857 Shape::legacy(100, Indent::new(0, 100)),
1858 &wrap_normalize_config).unwrap();
1859 assert_eq!("/* test */", comment);
1861 let comment = rewrite_comment("// comment on a",
1863 Shape::legacy(10, Indent::empty()),
1864 &wrap_normalize_config).unwrap();
1865 assert_eq!("// comment\n// on a", comment);
1867 let comment = rewrite_comment("// A multi line comment\n // between args.",
1869 Shape::legacy(60, Indent::new(0, 12)),
1870 &wrap_normalize_config).unwrap();
1871 assert_eq!("// A multi line comment\n // between args.", comment);
1873 let input = "// comment";
1876 let comment = rewrite_comment(input,
1878 Shape::legacy(9, Indent::new(0, 69)),
1879 &wrap_normalize_config).unwrap();
1880 assert_eq!(expected, comment);
1882 let comment = rewrite_comment("/* trimmed */",
1884 Shape::legacy(100, Indent::new(0, 100)),
1885 &wrap_normalize_config).unwrap();
1886 assert_eq!("/* trimmed */", comment);
1888 // Check that different comment style are properly recognised.
1889 let comment = rewrite_comment(r#"/// test1
1895 Shape::legacy(100, Indent::new(0, 0)),
1896 &wrap_normalize_config).unwrap();
1897 assert_eq!("/// test1\n/// test2\n// test3", comment);
1899 // Check that the blank line marks the end of a commented paragraph.
1900 let comment = rewrite_comment(r#"// test1
1904 Shape::legacy(100, Indent::new(0, 0)),
1905 &wrap_normalize_config).unwrap();
1906 assert_eq!("// test1\n\n// test2", comment);
1908 // Check that the blank line marks the end of a custom-commented paragraph.
1909 let comment = rewrite_comment(r#"//@ test1
1913 Shape::legacy(100, Indent::new(0, 0)),
1914 &wrap_normalize_config).unwrap();
1915 assert_eq!("//@ test1\n\n//@ test2", comment);
1917 // Check that bare lines are just indented but otherwise left unchanged.
1918 let comment = rewrite_comment(r#"// test1
1925 Shape::legacy(100, Indent::new(0, 0)),
1926 &wrap_config).unwrap();
1927 assert_eq!("// test1\n/*\n a bare line!\n\n another bare line!\n*/", comment);
1930 // This is probably intended to be a non-test fn, but it is not used.
1931 // We should keep this around unless it helps us test stuff to remove it.
1932 fn uncommented(text: &str) -> String {
1933 CharClasses::new(text.chars())
1934 .filter_map(|(s, c)| match s {
1935 FullCodeCharKind::Normal | FullCodeCharKind::InString => Some(c),
1942 fn test_uncommented() {
1943 assert_eq!(&uncommented("abc/*...*/"), "abc");
1945 &uncommented("// .... /* \n../* /* *** / */ */a/* // */c\n"),
1948 assert_eq!(&uncommented("abc \" /* */\" qsdf"), "abc \" /* */\" qsdf");
1952 fn test_contains_comment() {
1953 assert_eq!(contains_comment("abc"), false);
1954 assert_eq!(contains_comment("abc // qsdf"), true);
1955 assert_eq!(contains_comment("abc /* kqsdf"), true);
1956 assert_eq!(contains_comment("abc \" /* */\" qsdf"), false);
1960 fn test_find_uncommented() {
1961 fn check(haystack: &str, needle: &str, expected: Option<usize>) {
1962 assert_eq!(expected, haystack.find_uncommented(needle));
1965 check("/*/ */test", "test", Some(6));
1966 check("//test\ntest", "test", Some(7));
1967 check("/* comment only */", "whatever", None);
1969 "/* comment */ some text /* more commentary */ result",
1973 check("sup // sup", "p", Some(2));
1974 check("sup", "x", None);
1975 check(r#"π? /**/ π is nice!"#, r#"π is nice"#, Some(9));
1976 check("/*sup yo? \n sup*/ sup", "p", Some(20));
1977 check("hel/*lohello*/lo", "hello", None);
1978 check("acb", "ab", None);
1979 check(",/*A*/ ", ",", Some(0));
1980 check("abc", "abc", Some(0));
1981 check("/* abc */", "abc", None);
1982 check("/**/abc/* */", "abc", Some(4));
1983 check("\"/* abc */\"", "abc", Some(4));
1984 check("\"/* abc", "abc", Some(4));
1988 fn test_filter_normal_code() {
1991 println!("hello, world");
1994 assert_eq!(s, filter_normal_code(s));
1995 let s_with_comment = r#"
1998 println!("hello, world");
2001 assert_eq!(s, filter_normal_code(s_with_comment));