1 use clippy_utils::diagnostics::span_lint_and_sugg;
3 use rustc_errors::Applicability;
4 use rustc_lint::{EarlyContext, EarlyLintPass};
5 use rustc_session::{declare_lint_pass, declare_tool_lint};
6 use rustc_span::source_map::{BytePos, Span};
10 /// Checks doc comments for usage of tab characters.
12 /// ### Why is this bad?
13 /// The rust style-guide promotes spaces instead of tabs for indentation.
14 /// To keep a consistent view on the source, also doc comments should not have tabs.
15 /// Also, explaining ascii-diagrams containing tabs can get displayed incorrectly when the
16 /// display settings of the author and reader differ.
21 /// /// Struct to hold two strings:
24 /// pub struct DoubleString {
26 /// /// - First String:
27 /// /// - needs to be inside here
28 /// first_string: String,
30 /// /// - Second String:
31 /// /// - needs to be inside here
32 /// second_string: String,
36 /// Will be converted to:
39 /// /// Struct to hold two strings:
42 /// pub struct DoubleString {
44 /// /// - First String:
45 /// /// - needs to be inside here
46 /// first_string: String,
48 /// /// - Second String:
49 /// /// - needs to be inside here
50 /// second_string: String,
53 #[clippy::version = "1.41.0"]
54 pub TABS_IN_DOC_COMMENTS,
56 "using tabs in doc comments is not recommended"
59 declare_lint_pass!(TabsInDocComments => [TABS_IN_DOC_COMMENTS]);
61 impl TabsInDocComments {
62 fn warn_if_tabs_in_doc(cx: &EarlyContext<'_>, attr: &ast::Attribute) {
63 if let ast::AttrKind::DocComment(_, comment) = attr.kind {
64 let comment = comment.as_str();
66 for (lo, hi) in get_chunks_of_tabs(comment) {
67 // +3 skips the opening delimiter
68 let new_span = Span::new(
69 attr.span.lo() + BytePos(3 + lo),
70 attr.span.lo() + BytePos(3 + hi),
78 "using tabs in doc comments is not recommended",
79 "consider using four spaces per tab",
80 " ".repeat((hi - lo) as usize),
81 Applicability::MaybeIncorrect,
88 impl EarlyLintPass for TabsInDocComments {
89 fn check_attribute(&mut self, cx: &EarlyContext<'_>, attribute: &ast::Attribute) {
90 Self::warn_if_tabs_in_doc(cx, attribute);
95 /// scans the string for groups of tabs and returns the start(inclusive) and end positions
96 /// (exclusive) of all groups
97 /// e.g. "sd\tasd\t\taa" will be converted to [(2, 3), (6, 8)] as
100 fn get_chunks_of_tabs(the_str: &str) -> Vec<(u32, u32)> {
101 let line_length_way_to_long = "doc comment longer than 2^32 chars";
102 let mut spans: Vec<(u32, u32)> = vec![];
103 let mut current_start: u32 = 0;
105 // tracker to decide if the last group of tabs is not closed by a non-tab character
106 let mut is_active = false;
108 // Note that we specifically need the char _byte_ indices here, not the positional indexes
109 // within the char array to deal with multi-byte characters properly. `char_indices` does
110 // exactly that. It provides an iterator over tuples of the form `(byte position, char)`.
111 let char_indices: Vec<_> = the_str.char_indices().collect();
113 if let [(_, '\t')] = char_indices.as_slice() {
117 for entry in char_indices.windows(2) {
119 [(_, '\t'), (_, '\t')] => {
120 // either string starts with double tab, then we have to set it active,
121 // otherwise is_active is true anyway
124 [(_, _), (index_b, '\t')] => {
125 // as ['\t', '\t'] is excluded, this has to be a start of a tab group,
126 // set indices accordingly
128 current_start = u32::try_from(*index_b).unwrap();
130 [(_, '\t'), (index_b, _)] => {
131 // this now has to be an end of the group, hence we have to push a new tuple
133 spans.push((current_start, u32::try_from(*index_b).unwrap()));
139 // only possible when tabs are at the end, insert last group
143 u32::try_from(char_indices.last().unwrap().0 + 1).expect(line_length_way_to_long),
151 mod tests_for_get_chunks_of_tabs {
152 use super::get_chunks_of_tabs;
155 fn test_unicode_han_string() {
156 let res = get_chunks_of_tabs(" \u{4f4d}\t");
158 assert_eq!(res, vec![(4, 5)]);
162 fn test_empty_string() {
163 let res = get_chunks_of_tabs("");
165 assert_eq!(res, vec![]);
170 let res = get_chunks_of_tabs("sd\t\t\taa");
172 assert_eq!(res, vec![(2, 5)]);
177 let res = get_chunks_of_tabs("\t\t");
179 assert_eq!(res, vec![(0, 2)]);
183 fn test_only_one_t() {
184 let res = get_chunks_of_tabs("\t");
186 assert_eq!(res, vec![(0, 1)]);
191 let res = get_chunks_of_tabs("sd\tasd\t\taa");
193 assert_eq!(res, vec![(2, 3), (6, 8)]);
198 let res = get_chunks_of_tabs("\t\taa");
200 assert_eq!(res, vec![(0, 2)]);
205 let res = get_chunks_of_tabs("aa\t\t");
207 assert_eq!(res, vec![(2, 4)]);
211 fn test_start_single() {
212 let res = get_chunks_of_tabs("\taa");
214 assert_eq!(res, vec![(0, 1)]);
218 fn test_end_single() {
219 let res = get_chunks_of_tabs("aa\t");
221 assert_eq!(res, vec![(2, 3)]);
226 let res = get_chunks_of_tabs("dsfs");
228 assert_eq!(res, vec![]);