1 use crate::consts::{constant, Constant};
2 use clippy_utils::diagnostics::{span_lint, span_lint_and_help};
3 use clippy_utils::{match_def_path, paths};
4 use if_chain::if_chain;
5 use rustc_ast::ast::{LitKind, StrStyle};
6 use rustc_data_structures::fx::FxHashSet;
7 use rustc_hir::{BorrowKind, Expr, ExprKind, HirId};
8 use rustc_lint::{LateContext, LateLintPass};
9 use rustc_session::{declare_tool_lint, impl_lint_pass};
10 use rustc_span::source_map::{BytePos, Span};
11 use std::convert::TryFrom;
13 declare_clippy_lint! {
14 /// **What it does:** Checks [regex](https://crates.io/crates/regex) creation
15 /// (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`) for correct
18 /// **Why is this bad?** This will lead to a runtime panic.
20 /// **Known problems:** None.
28 "invalid regular expressions"
31 declare_clippy_lint! {
32 /// **What it does:** Checks for trivial [regex](https://crates.io/crates/regex)
33 /// creation (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`).
35 /// **Why is this bad?** Matching the regex can likely be replaced by `==` or
36 /// `str::starts_with`, `str::ends_with` or `std::contains` or other `str`
39 /// **Known problems:** If the same regex is going to be applied to multiple
40 /// inputs, the precomputations done by `Regex` construction can give
41 /// significantly better performance than any of the `str`-based methods.
45 /// Regex::new("^foobar")
49 "trivial regular expressions"
52 #[derive(Clone, Default)]
54 spans: FxHashSet<Span>,
58 impl_lint_pass!(Regex => [INVALID_REGEX, TRIVIAL_REGEX]);
60 impl<'tcx> LateLintPass<'tcx> for Regex {
61 fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
63 if let ExprKind::Call(ref fun, ref args) = expr.kind;
64 if let ExprKind::Path(ref qpath) = fun.kind;
66 if let Some(def_id) = cx.qpath_res(qpath, fun.hir_id).opt_def_id();
68 if match_def_path(cx, def_id, &paths::REGEX_NEW) ||
69 match_def_path(cx, def_id, &paths::REGEX_BUILDER_NEW) {
70 check_regex(cx, &args[0], true);
71 } else if match_def_path(cx, def_id, &paths::REGEX_BYTES_NEW) ||
72 match_def_path(cx, def_id, &paths::REGEX_BYTES_BUILDER_NEW) {
73 check_regex(cx, &args[0], false);
74 } else if match_def_path(cx, def_id, &paths::REGEX_SET_NEW) {
75 check_set(cx, &args[0], true);
76 } else if match_def_path(cx, def_id, &paths::REGEX_BYTES_SET_NEW) {
77 check_set(cx, &args[0], false);
84 #[allow(clippy::cast_possible_truncation)] // truncation very unlikely here
86 fn str_span(base: Span, c: regex_syntax::ast::Span, offset: u16) -> Span {
87 let offset = u32::from(offset);
88 let end = base.lo() + BytePos(u32::try_from(c.end.offset).expect("offset too large") + offset);
89 let start = base.lo() + BytePos(u32::try_from(c.start.offset).expect("offset too large") + offset);
90 assert!(start <= end);
91 Span::new(start, end, base.ctxt())
94 fn const_str<'tcx>(cx: &LateContext<'tcx>, e: &'tcx Expr<'_>) -> Option<String> {
95 constant(cx, cx.typeck_results(), e).and_then(|(c, _)| match c {
96 Constant::Str(s) => Some(s),
101 fn is_trivial_regex(s: ®ex_syntax::hir::Hir) -> Option<&'static str> {
102 use regex_syntax::hir::Anchor::{EndText, StartText};
103 use regex_syntax::hir::HirKind::{Alternation, Anchor, Concat, Empty, Literal};
105 let is_literal = |e: &[regex_syntax::hir::Hir]| e.iter().all(|e| matches!(*e.kind(), Literal(_)));
108 Empty | Anchor(_) => Some("the regex is unlikely to be useful as it is"),
109 Literal(_) => Some("consider using `str::contains`"),
110 Alternation(ref exprs) => {
111 if exprs.iter().all(|e| e.kind().is_empty()) {
112 Some("the regex is unlikely to be useful as it is")
117 Concat(ref exprs) => match (exprs[0].kind(), exprs[exprs.len() - 1].kind()) {
118 (&Anchor(StartText), &Anchor(EndText)) if exprs[1..(exprs.len() - 1)].is_empty() => {
119 Some("consider using `str::is_empty`")
121 (&Anchor(StartText), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
122 Some("consider using `==` on `str`s")
124 (&Anchor(StartText), &Literal(_)) if is_literal(&exprs[1..]) => Some("consider using `str::starts_with`"),
125 (&Literal(_), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
126 Some("consider using `str::ends_with`")
128 _ if is_literal(exprs) => Some("consider using `str::contains`"),
135 fn check_set<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
137 if let ExprKind::AddrOf(BorrowKind::Ref, _, ref expr) = expr.kind;
138 if let ExprKind::Array(exprs) = expr.kind;
141 check_regex(cx, expr, utf8);
147 fn check_regex<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
148 let mut parser = regex_syntax::ParserBuilder::new()
150 .allow_invalid_utf8(!utf8)
153 if let ExprKind::Lit(ref lit) = expr.kind {
154 if let LitKind::Str(ref r, style) = lit.node {
156 let offset = if let StrStyle::Raw(n) = style { 2 + n } else { 1 };
157 match parser.parse(r) {
159 if let Some(repl) = is_trivial_regex(&r) {
160 span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
163 Err(regex_syntax::Error::Parse(e)) => {
167 str_span(expr.span, *e.span(), offset),
168 &format!("regex syntax error: {}", e.kind()),
171 Err(regex_syntax::Error::Translate(e)) => {
175 str_span(expr.span, *e.span(), offset),
176 &format!("regex syntax error: {}", e.kind()),
180 span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));
184 } else if let Some(r) = const_str(cx, expr) {
185 match parser.parse(&r) {
187 if let Some(repl) = is_trivial_regex(&r) {
188 span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
191 Err(regex_syntax::Error::Parse(e)) => {
196 &format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
199 Err(regex_syntax::Error::Translate(e)) => {
204 &format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
208 span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));