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};
7 use std::convert::TryFrom;
10 /// **What it does:** Checks doc comments for usage of tab characters.
12 /// **Why is this bad?** The rust style-guide promotes spaces instead of tabs for indentation.
13 /// To keep a consistent view on the source, also doc comments should not have tabs.
14 /// Also, explaining ascii-diagrams containing tabs can get displayed incorrectly when the
15 /// display settings of the author and reader differ.
17 /// **Known problems:** None.
22 /// /// Struct to hold two strings:
25 /// pub struct DoubleString {
27 /// /// - First String:
28 /// /// - needs to be inside here
29 /// first_string: String,
31 /// /// - Second String:
32 /// /// - needs to be inside here
33 /// second_string: String,
37 /// Will be converted to:
40 /// /// Struct to hold two strings:
43 /// pub struct DoubleString {
45 /// /// - First String:
46 /// /// - needs to be inside here
47 /// first_string: String,
49 /// /// - Second String:
50 /// /// - needs to be inside here
51 /// second_string: String,
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),
77 "using tabs in doc comments is not recommended",
78 "consider using four spaces per tab",
79 " ".repeat((hi - lo) as usize),
80 Applicability::MaybeIncorrect,
87 impl EarlyLintPass for TabsInDocComments {
88 fn check_attribute(&mut self, cx: &EarlyContext<'_>, attribute: &ast::Attribute) {
89 Self::warn_if_tabs_in_doc(cx, attribute);
94 /// scans the string for groups of tabs and returns the start(inclusive) and end positions
95 /// (exclusive) of all groups
96 /// e.g. "sd\tasd\t\taa" will be converted to [(2, 3), (6, 8)] as
99 fn get_chunks_of_tabs(the_str: &str) -> Vec<(u32, u32)> {
100 let line_length_way_to_long = "doc comment longer than 2^32 chars";
101 let mut spans: Vec<(u32, u32)> = vec![];
102 let mut current_start: u32 = 0;
104 // tracker to decide if the last group of tabs is not closed by a non-tab character
105 let mut is_active = false;
107 // Note that we specifically need the char _byte_ indices here, not the positional indexes
108 // within the char array to deal with multi-byte characters properly. `char_indices` does
109 // exactly that. It provides an iterator over tuples of the form `(byte position, char)`.
110 let char_indices: Vec<_> = the_str.char_indices().collect();
112 if let [(_, '\t')] = char_indices.as_slice() {
116 for entry in char_indices.windows(2) {
118 [(_, '\t'), (_, '\t')] => {
119 // either string starts with double tab, then we have to set it active,
120 // otherwise is_active is true anyway
123 [(_, _), (index_b, '\t')] => {
124 // as ['\t', '\t'] is excluded, this has to be a start of a tab group,
125 // set indices accordingly
127 current_start = u32::try_from(*index_b).unwrap();
129 [(_, '\t'), (index_b, _)] => {
130 // this now has to be an end of the group, hence we have to push a new tuple
132 spans.push((current_start, u32::try_from(*index_b).unwrap()));
138 // only possible when tabs are at the end, insert last group
142 u32::try_from(char_indices.last().unwrap().0 + 1).expect(line_length_way_to_long),
150 mod tests_for_get_chunks_of_tabs {
151 use super::get_chunks_of_tabs;
154 fn test_unicode_han_string() {
155 let res = get_chunks_of_tabs(" \u{4f4d}\t");
157 assert_eq!(res, vec![(4, 5)]);
161 fn test_empty_string() {
162 let res = get_chunks_of_tabs("");
164 assert_eq!(res, vec![]);
169 let res = get_chunks_of_tabs("sd\t\t\taa");
171 assert_eq!(res, vec![(2, 5)]);
176 let res = get_chunks_of_tabs("\t\t");
178 assert_eq!(res, vec![(0, 2)]);
182 fn test_only_one_t() {
183 let res = get_chunks_of_tabs("\t");
185 assert_eq!(res, vec![(0, 1)]);
190 let res = get_chunks_of_tabs("sd\tasd\t\taa");
192 assert_eq!(res, vec![(2, 3), (6, 8)]);
197 let res = get_chunks_of_tabs("\t\taa");
199 assert_eq!(res, vec![(0, 2)]);
204 let res = get_chunks_of_tabs("aa\t\t");
206 assert_eq!(res, vec![(2, 4)]);
210 fn test_start_single() {
211 let res = get_chunks_of_tabs("\taa");
213 assert_eq!(res, vec![(0, 1)]);
217 fn test_end_single() {
218 let res = get_chunks_of_tabs("aa\t");
220 assert_eq!(res, vec![(2, 3)]);
225 let res = get_chunks_of_tabs("dsfs");
227 assert_eq!(res, vec![]);