1 use crate::consts::{constant, Constant};
2 use crate::utils::{match_def_path, paths, span_lint, span_lint_and_help};
3 use if_chain::if_chain;
4 use rustc_ast::ast::{LitKind, StrStyle};
5 use rustc_data_structures::fx::FxHashSet;
6 use rustc_hir::{BorrowKind, Expr, ExprKind, HirId};
7 use rustc_lint::{LateContext, LateLintPass};
8 use rustc_session::{declare_tool_lint, impl_lint_pass};
9 use rustc_span::source_map::{BytePos, Span};
10 use std::convert::TryFrom;
12 declare_clippy_lint! {
13 /// **What it does:** Checks [regex](https://crates.io/crates/regex) creation
14 /// (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`) for correct
17 /// **Why is this bad?** This will lead to a runtime panic.
19 /// **Known problems:** None.
27 "invalid regular expressions"
30 declare_clippy_lint! {
31 /// **What it does:** Checks for trivial [regex](https://crates.io/crates/regex)
32 /// creation (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`).
34 /// **Why is this bad?** Matching the regex can likely be replaced by `==` or
35 /// `str::starts_with`, `str::ends_with` or `std::contains` or other `str`
38 /// **Known problems:** If the same regex is going to be applied to multiple
39 /// inputs, the precomputations done by `Regex` construction can give
40 /// significantly better performance than any of the `str`-based methods.
44 /// Regex::new("^foobar")
48 "trivial regular expressions"
51 #[derive(Clone, Default)]
53 spans: FxHashSet<Span>,
57 impl_lint_pass!(Regex => [INVALID_REGEX, TRIVIAL_REGEX]);
59 impl<'tcx> LateLintPass<'tcx> for Regex {
60 fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
62 if let ExprKind::Call(ref fun, ref args) = expr.kind;
63 if let ExprKind::Path(ref qpath) = fun.kind;
65 if let Some(def_id) = cx.qpath_res(qpath, fun.hir_id).opt_def_id();
67 if match_def_path(cx, def_id, &paths::REGEX_NEW) ||
68 match_def_path(cx, def_id, &paths::REGEX_BUILDER_NEW) {
69 check_regex(cx, &args[0], true);
70 } else if match_def_path(cx, def_id, &paths::REGEX_BYTES_NEW) ||
71 match_def_path(cx, def_id, &paths::REGEX_BYTES_BUILDER_NEW) {
72 check_regex(cx, &args[0], false);
73 } else if match_def_path(cx, def_id, &paths::REGEX_SET_NEW) {
74 check_set(cx, &args[0], true);
75 } else if match_def_path(cx, def_id, &paths::REGEX_BYTES_SET_NEW) {
76 check_set(cx, &args[0], false);
83 #[allow(clippy::cast_possible_truncation)] // truncation very unlikely here
85 fn str_span(base: Span, c: regex_syntax::ast::Span, offset: u16) -> Span {
86 let offset = u32::from(offset);
87 let end = base.lo() + BytePos(u32::try_from(c.end.offset).expect("offset too large") + offset);
88 let start = base.lo() + BytePos(u32::try_from(c.start.offset).expect("offset too large") + offset);
89 assert!(start <= end);
90 Span::new(start, end, base.ctxt())
93 fn const_str<'tcx>(cx: &LateContext<'tcx>, e: &'tcx Expr<'_>) -> Option<String> {
94 constant(cx, cx.typeck_results(), e).and_then(|(c, _)| match c {
95 Constant::Str(s) => Some(s),
100 fn is_trivial_regex(s: ®ex_syntax::hir::Hir) -> Option<&'static str> {
101 use regex_syntax::hir::Anchor::{EndText, StartText};
102 use regex_syntax::hir::HirKind::{Alternation, Anchor, Concat, Empty, Literal};
104 let is_literal = |e: &[regex_syntax::hir::Hir]| e.iter().all(|e| matches!(*e.kind(), Literal(_)));
107 Empty | Anchor(_) => Some("the regex is unlikely to be useful as it is"),
108 Literal(_) => Some("consider using `str::contains`"),
109 Alternation(ref exprs) => {
110 if exprs.iter().all(|e| e.kind().is_empty()) {
111 Some("the regex is unlikely to be useful as it is")
116 Concat(ref exprs) => match (exprs[0].kind(), exprs[exprs.len() - 1].kind()) {
117 (&Anchor(StartText), &Anchor(EndText)) if exprs[1..(exprs.len() - 1)].is_empty() => {
118 Some("consider using `str::is_empty`")
120 (&Anchor(StartText), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
121 Some("consider using `==` on `str`s")
123 (&Anchor(StartText), &Literal(_)) if is_literal(&exprs[1..]) => Some("consider using `str::starts_with`"),
124 (&Literal(_), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
125 Some("consider using `str::ends_with`")
127 _ if is_literal(exprs) => Some("consider using `str::contains`"),
134 fn check_set<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
136 if let ExprKind::AddrOf(BorrowKind::Ref, _, ref expr) = expr.kind;
137 if let ExprKind::Array(exprs) = expr.kind;
140 check_regex(cx, expr, utf8);
146 fn check_regex<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
147 let mut parser = regex_syntax::ParserBuilder::new()
149 .allow_invalid_utf8(!utf8)
152 if let ExprKind::Lit(ref lit) = expr.kind {
153 if let LitKind::Str(ref r, style) = lit.node {
155 let offset = if let StrStyle::Raw(n) = style { 2 + n } else { 1 };
156 match parser.parse(r) {
158 if let Some(repl) = is_trivial_regex(&r) {
159 span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
162 Err(regex_syntax::Error::Parse(e)) => {
166 str_span(expr.span, *e.span(), offset),
167 &format!("regex syntax error: {}", e.kind()),
170 Err(regex_syntax::Error::Translate(e)) => {
174 str_span(expr.span, *e.span(), offset),
175 &format!("regex syntax error: {}", e.kind()),
179 span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));
183 } else if let Some(r) = const_str(cx, expr) {
184 match parser.parse(&r) {
186 if let Some(repl) = is_trivial_regex(&r) {
187 span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
190 Err(regex_syntax::Error::Parse(e)) => {
195 &format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
198 Err(regex_syntax::Error::Translate(e)) => {
203 &format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
207 span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));