rust/clippy_lints/src/regex.rs

232 lines
8.4 KiB
Rust
Raw Normal View History

use std::fmt::Display;
use clippy_utils::consts::{constant, Constant};
use clippy_utils::diagnostics::{span_lint, span_lint_and_help};
use clippy_utils::source::snippet_opt;
use clippy_utils::{def_path_def_ids, path_def_id, paths};
2020-02-29 21:23:33 -06:00
use rustc_ast::ast::{LitKind, StrStyle};
use rustc_hir::def_id::DefIdMap;
use rustc_hir::{BorrowKind, Expr, ExprKind};
2020-01-12 00:08:41 -06:00
use rustc_lint::{LateContext, LateLintPass};
use rustc_session::{declare_tool_lint, impl_lint_pass};
use rustc_span::{BytePos, Span};
2016-02-04 17:36:06 -06:00
2018-03-28 08:24:26 -05:00
declare_clippy_lint! {
/// ### What it does
/// Checks [regex](https://crates.io/crates/regex) creation
/// (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`) for correct
/// regex syntax.
///
/// ### Why is this bad?
/// This will lead to a runtime panic.
///
/// ### Example
2019-03-05 16:23:50 -06:00
/// ```ignore
/// Regex::new("(")
/// ```
#[clippy::version = "pre 1.29.0"]
2016-02-04 17:36:06 -06:00
pub INVALID_REGEX,
2018-03-28 08:24:26 -05:00
correctness,
"invalid regular expressions"
2016-02-04 17:36:06 -06:00
}
2018-03-28 08:24:26 -05:00
declare_clippy_lint! {
/// ### What it does
/// Checks for trivial [regex](https://crates.io/crates/regex)
/// creation (with `Regex::new`, `RegexBuilder::new`, or `RegexSet::new`).
///
/// ### Why is this bad?
/// Matching the regex can likely be replaced by `==` or
/// `str::starts_with`, `str::ends_with` or `std::contains` or other `str`
/// methods.
///
/// ### Known problems
/// If the same regex is going to be applied to multiple
/// inputs, the precomputations done by `Regex` construction can give
/// significantly better performance than any of the `str`-based methods.
///
/// ### Example
2019-03-05 16:23:50 -06:00
/// ```ignore
/// Regex::new("^foobar")
/// ```
#[clippy::version = "pre 1.29.0"]
2016-02-05 16:10:48 -06:00
pub TRIVIAL_REGEX,
nursery,
"trivial regular expressions"
2016-02-05 16:10:48 -06:00
}
#[derive(Copy, Clone)]
enum RegexKind {
Unicode,
UnicodeSet,
Bytes,
BytesSet,
}
#[derive(Default)]
pub struct Regex {
definitions: DefIdMap<RegexKind>,
}
impl_lint_pass!(Regex => [INVALID_REGEX, TRIVIAL_REGEX]);
2016-02-04 17:36:06 -06:00
impl<'tcx> LateLintPass<'tcx> for Regex {
fn check_crate(&mut self, cx: &LateContext<'tcx>) {
// We don't use `match_def_path` here because that relies on matching the exact path, which changed
// between regex 1.8 and 1.9
//
// `def_path_def_ids` will resolve through re-exports but is relatively heavy, so we only perform
// the operation once and store the results
let mut resolve = |path, kind| {
for id in def_path_def_ids(cx, path) {
self.definitions.insert(id, kind);
}
};
resolve(&paths::REGEX_NEW, RegexKind::Unicode);
resolve(&paths::REGEX_BUILDER_NEW, RegexKind::Unicode);
resolve(&paths::REGEX_SET_NEW, RegexKind::UnicodeSet);
resolve(&paths::REGEX_BYTES_NEW, RegexKind::Bytes);
resolve(&paths::REGEX_BYTES_BUILDER_NEW, RegexKind::Bytes);
resolve(&paths::REGEX_BYTES_SET_NEW, RegexKind::BytesSet);
}
fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
if let ExprKind::Call(fun, [arg]) = expr.kind
&& let Some(def_id) = path_def_id(cx, fun)
&& let Some(regex_kind) = self.definitions.get(&def_id)
{
match regex_kind {
RegexKind::Unicode => check_regex(cx, arg, true),
RegexKind::UnicodeSet => check_set(cx, arg, true),
RegexKind::Bytes => check_regex(cx, arg, false),
RegexKind::BytesSet => check_set(cx, arg, false),
2016-02-05 09:48:35 -06:00
}
}
2016-02-04 17:36:06 -06:00
}
}
2016-02-05 09:48:35 -06:00
fn lint_syntax_error(cx: &LateContext<'_>, error: &regex_syntax::Error, unescaped: &str, base: Span, offset: u8) {
let parts: Option<(_, _, &dyn Display)> = match &error {
regex_syntax::Error::Parse(e) => Some((e.span(), e.auxiliary_span(), e.kind())),
regex_syntax::Error::Translate(e) => Some((e.span(), None, e.kind())),
_ => None,
};
let convert_span = |regex_span: &regex_syntax::ast::Span| {
let offset = u32::from(offset);
let start = base.lo() + BytePos(u32::try_from(regex_span.start.offset).expect("offset too large") + offset);
let end = base.lo() + BytePos(u32::try_from(regex_span.end.offset).expect("offset too large") + offset);
Span::new(start, end, base.ctxt(), base.parent())
};
if let Some((primary, auxiliary, kind)) = parts
&& let Some(literal_snippet) = snippet_opt(cx, base)
&& let Some(inner) = literal_snippet.get(offset as usize..)
// Only convert to native rustc spans if the parsed regex matches the
// source snippet exactly, to ensure the span offsets are correct
&& inner.get(..unescaped.len()) == Some(unescaped)
{
let spans = if let Some(auxiliary) = auxiliary {
vec![convert_span(primary), convert_span(auxiliary)]
} else {
vec![convert_span(primary)]
};
span_lint(cx, INVALID_REGEX, spans, &format!("regex syntax error: {kind}"));
} else {
span_lint_and_help(
cx,
INVALID_REGEX,
base,
&error.to_string(),
None,
"consider using a raw string literal: `r\"..\"`",
);
}
2016-02-05 09:48:35 -06:00
}
fn const_str<'tcx>(cx: &LateContext<'tcx>, e: &'tcx Expr<'_>) -> Option<String> {
constant(cx, cx.typeck_results(), e).and_then(|c| match c {
2018-03-13 05:38:11 -05:00
Constant::Str(s) => Some(s),
2016-02-24 10:38:57 -06:00
_ => None,
2018-03-13 05:38:11 -05:00
})
2016-02-05 09:48:35 -06:00
}
2016-02-05 16:10:48 -06:00
2018-03-13 09:02:40 -05:00
fn is_trivial_regex(s: &regex_syntax::hir::Hir) -> Option<&'static str> {
use regex_syntax::hir::HirKind::{Alternation, Concat, Empty, Literal, Look};
use regex_syntax::hir::Look as HirLook;
2018-03-13 09:02:40 -05:00
let is_literal = |e: &[regex_syntax::hir::Hir]| e.iter().all(|e| matches!(*e.kind(), Literal(_)));
2018-03-13 09:02:40 -05:00
match *s.kind() {
Empty | Look(_) => Some("the regex is unlikely to be useful as it is"),
2018-03-13 09:02:40 -05:00
Literal(_) => Some("consider using `str::contains`"),
2018-11-27 14:14:15 -06:00
Alternation(ref exprs) => {
if exprs.iter().all(|e| matches!(e.kind(), Empty)) {
2018-11-27 14:14:15 -06:00
Some("the regex is unlikely to be useful as it is")
} else {
None
}
2018-03-13 09:02:40 -05:00
},
Concat(ref exprs) => match (exprs[0].kind(), exprs[exprs.len() - 1].kind()) {
(&Look(HirLook::Start), &Look(HirLook::End)) if exprs[1..(exprs.len() - 1)].is_empty() => {
2018-11-27 14:14:15 -06:00
Some("consider using `str::is_empty`")
},
(&Look(HirLook::Start), &Look(HirLook::End)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
2018-11-27 14:14:15 -06:00
Some("consider using `==` on `str`s")
},
(&Look(HirLook::Start), &Literal(_)) if is_literal(&exprs[1..]) => {
Some("consider using `str::starts_with`")
},
(&Literal(_), &Look(HirLook::End)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
2018-11-27 14:14:15 -06:00
Some("consider using `str::ends_with`")
},
2018-03-13 09:02:40 -05:00
_ if is_literal(exprs) => Some("consider using `str::contains`"),
2017-09-05 04:33:04 -05:00
_ => None,
2016-12-20 11:21:30 -06:00
},
2016-02-06 11:06:39 -06:00
_ => None,
2016-02-05 16:10:48 -06:00
}
}
2016-05-25 10:15:19 -05:00
fn check_set<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
if_chain! {
if let ExprKind::AddrOf(BorrowKind::Ref, _, expr) = expr.kind;
2019-12-27 01:12:26 -06:00
if let ExprKind::Array(exprs) = expr.kind;
then {
for expr in exprs {
check_regex(cx, expr, utf8);
}
2016-05-25 10:15:19 -05:00
}
}
2016-05-25 10:15:19 -05:00
}
fn check_regex<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>, utf8: bool) {
let mut parser = regex_syntax::ParserBuilder::new().unicode(true).utf8(utf8).build();
2016-05-25 10:15:19 -05:00
if let ExprKind::Lit(lit) = expr.kind {
if let LitKind::Str(ref r, style) = lit.node {
let r = r.as_str();
2018-03-13 09:02:40 -05:00
let offset = if let StrStyle::Raw(n) = style { 2 + n } else { 1 };
match parser.parse(r) {
2018-11-27 14:14:15 -06:00
Ok(r) => {
if let Some(repl) = is_trivial_regex(&r) {
span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
2018-11-27 14:14:15 -06:00
}
2018-03-13 09:02:40 -05:00
},
Err(e) => lint_syntax_error(cx, &e, r, expr.span, offset),
2016-05-25 10:15:19 -05:00
}
}
} else if let Some(r) = const_str(cx, expr) {
2018-03-13 09:02:40 -05:00
match parser.parse(&r) {
2018-11-27 14:14:15 -06:00
Ok(r) => {
if let Some(repl) = is_trivial_regex(&r) {
span_lint_and_help(cx, TRIVIAL_REGEX, expr.span, "trivial regex", None, repl);
2018-11-27 14:14:15 -06:00
}
2018-03-13 09:02:40 -05:00
},
Err(e) => span_lint(cx, INVALID_REGEX, expr.span, &e.to_string()),
2016-05-25 10:15:19 -05:00
}
}
}