rust-clippy/clippy_lints/src/regex.rs

272 lines
9.6 KiB
Rust
Raw Normal View History

2018-11-27 20:14:15 +00:00
use crate::consts::{constant, Constant};
use crate::utils::{is_expn_of, match_def_path, match_type, paths, span_help_and_lint, span_lint};
2018-11-27 20:14:15 +00:00
use if_chain::if_chain;
use regex_syntax;
use rustc::hir::*;
use rustc::lint::{LateContext, LateLintPass, LintArray, LintPass};
use rustc::{declare_tool_lint, lint_array};
use rustc_data_structures::fx::FxHashSet;
use std::convert::TryFrom;
2019-02-24 18:43:15 +00:00
use syntax::ast::{LitKind, StrStyle};
use syntax::source_map::{BytePos, Span};
2016-02-04 23:36:06 +00:00
2018-03-28 13:24:26 +00:00
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.
///
/// **Known problems:** None.
///
/// **Example:**
2019-03-05 22:23:50 +00:00
/// ```ignore
/// Regex::new("|")
/// ```
2016-02-04 23:36:06 +00:00
pub INVALID_REGEX,
2018-03-28 13:24:26 +00:00
correctness,
"invalid regular expressions"
2016-02-04 23:36:06 +00:00
}
2018-03-28 13:24:26 +00:00
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:** None.
///
/// **Example:**
2019-03-05 22:23:50 +00:00
/// ```ignore
/// Regex::new("^foobar")
/// ```
2016-02-05 22:10:48 +00:00
pub TRIVIAL_REGEX,
2018-03-28 13:24:26 +00:00
style,
"trivial regular expressions"
2016-02-05 22:10:48 +00:00
}
2018-03-28 13:24:26 +00:00
declare_clippy_lint! {
/// **What it does:** Checks for usage of `regex!(_)` which (as of now) is
/// usually slower than `Regex::new(_)` unless called in a loop (which is a bad
/// idea anyway).
///
/// **Why is this bad?** Performance, at least for now. The macro version is
/// likely to catch up long-term, but for now the dynamic version is faster.
///
/// **Known problems:** None.
///
/// **Example:**
2019-03-05 22:23:50 +00:00
/// ```ignore
/// regex!("foo|bar")
/// ```
2016-02-07 21:50:54 +00:00
pub REGEX_MACRO,
2018-03-28 13:24:26 +00:00
style,
"use of `regex!(_)` instead of `Regex::new(_)`"
2016-02-07 21:50:54 +00:00
}
2016-02-14 15:55:02 +00:00
#[derive(Clone, Default)]
2016-06-10 14:17:20 +00:00
pub struct Pass {
spans: FxHashSet<Span>,
2019-02-24 18:43:15 +00:00
last: Option<HirId>,
2016-02-14 15:55:02 +00:00
}
2016-02-04 23:36:06 +00:00
2016-06-10 14:17:20 +00:00
impl LintPass for Pass {
2016-02-04 23:36:06 +00:00
fn get_lints(&self) -> LintArray {
2016-02-07 21:50:54 +00:00
lint_array!(INVALID_REGEX, REGEX_MACRO, TRIVIAL_REGEX)
2016-02-04 23:36:06 +00:00
}
fn name(&self) -> &'static str {
"Regex"
}
2016-02-04 23:36:06 +00:00
}
impl<'a, 'tcx> LateLintPass<'a, 'tcx> for Pass {
fn check_crate(&mut self, _: &LateContext<'a, 'tcx>, _: &'tcx Crate) {
2016-02-14 15:55:02 +00:00
self.spans.clear();
2016-02-07 21:50:54 +00:00
}
fn check_block(&mut self, cx: &LateContext<'a, 'tcx>, block: &'tcx Block) {
if_chain! {
if self.last.is_none();
if let Some(ref expr) = block.expr;
if match_type(cx, cx.tables.expr_ty(expr), &paths::REGEX);
if let Some(span) = is_expn_of(expr.span, "regex");
then {
if !self.spans.contains(&span) {
span_lint(cx,
REGEX_MACRO,
span,
"`regex!(_)` found. \
Please use `Regex::new(_)`, which is faster for now.");
self.spans.insert(span);
}
2019-02-24 18:43:15 +00:00
self.last = Some(block.hir_id);
2016-02-14 15:55:02 +00:00
}
}
2016-02-14 15:55:02 +00:00
}
2016-02-24 16:38:57 +00:00
fn check_block_post(&mut self, _: &LateContext<'a, 'tcx>, block: &'tcx Block) {
2019-02-24 18:43:15 +00:00
if self.last.map_or(false, |id| block.hir_id == id) {
2016-02-24 16:38:57 +00:00
self.last = None;
2016-02-14 15:55:02 +00:00
}
}
2016-02-07 21:50:54 +00:00
fn check_expr(&mut self, cx: &LateContext<'a, 'tcx>, expr: &'tcx Expr) {
if_chain! {
2018-07-12 07:30:57 +00:00
if let ExprKind::Call(ref fun, ref args) = expr.node;
if let ExprKind::Path(ref qpath) = fun.node;
if args.len() == 1;
if let Some(def_id) = cx.tables.qpath_def(qpath, fun.hir_id).opt_def_id();
then {
if match_def_path(cx.tcx, def_id, &paths::REGEX_NEW) ||
match_def_path(cx.tcx, def_id, &paths::REGEX_BUILDER_NEW) {
check_regex(cx, &args[0], true);
} else if match_def_path(cx.tcx, def_id, &paths::REGEX_BYTES_NEW) ||
match_def_path(cx.tcx, def_id, &paths::REGEX_BYTES_BUILDER_NEW) {
check_regex(cx, &args[0], false);
} else if match_def_path(cx.tcx, def_id, &paths::REGEX_SET_NEW) {
check_set(cx, &args[0], true);
} else if match_def_path(cx.tcx, def_id, &paths::REGEX_BYTES_SET_NEW) {
check_set(cx, &args[0], false);
}
2016-02-05 15:48:35 +00:00
}
}
2016-02-04 23:36:06 +00:00
}
}
2016-02-05 15:48:35 +00:00
#[allow(clippy::cast_possible_truncation)] // truncation very unlikely here
2018-04-19 06:30:07 +00:00
fn str_span(base: Span, c: regex_syntax::ast::Span, offset: u16) -> Span {
let offset = u32::from(offset);
let end = base.lo() + BytePos(u32::try_from(c.end.offset).expect("offset too large") + offset);
let start = base.lo() + BytePos(u32::try_from(c.start.offset).expect("offset too large") + offset);
2018-03-13 14:02:40 +00:00
assert!(start <= end);
Span::new(start, end, base.ctxt())
2016-02-05 15:48:35 +00:00
}
2018-03-13 10:38:11 +00:00
fn const_str<'a, 'tcx>(cx: &LateContext<'a, 'tcx>, e: &'tcx Expr) -> Option<String> {
2018-05-13 11:16:31 +00:00
constant(cx, cx.tables, e).and_then(|(c, _)| match c {
2018-03-13 10:38:11 +00:00
Constant::Str(s) => Some(s),
2016-02-24 16:38:57 +00:00
_ => None,
2018-03-13 10:38:11 +00:00
})
2016-02-05 15:48:35 +00:00
}
2016-02-05 22:10:48 +00:00
2018-03-13 14:02:40 +00:00
fn is_trivial_regex(s: &regex_syntax::hir::Hir) -> Option<&'static str> {
use regex_syntax::hir::Anchor::*;
2018-11-27 20:14:15 +00:00
use regex_syntax::hir::HirKind::*;
2018-03-13 14:02:40 +00:00
2018-11-27 20:14:15 +00:00
let is_literal = |e: &[regex_syntax::hir::Hir]| {
e.iter().all(|e| match *e.kind() {
Literal(_) => true,
_ => false,
})
};
2018-03-13 14:02:40 +00:00
match *s.kind() {
2018-11-27 20:14:15 +00:00
Empty | Anchor(_) => Some("the regex is unlikely to be useful as it is"),
2018-03-13 14:02:40 +00:00
Literal(_) => Some("consider using `str::contains`"),
2018-11-27 20:14:15 +00:00
Alternation(ref exprs) => {
if exprs.iter().all(|e| e.kind().is_empty()) {
Some("the regex is unlikely to be useful as it is")
} else {
None
}
2018-03-13 14:02:40 +00:00
},
Concat(ref exprs) => match (exprs[0].kind(), exprs[exprs.len() - 1].kind()) {
2018-11-27 20:14:15 +00:00
(&Anchor(StartText), &Anchor(EndText)) if exprs[1..(exprs.len() - 1)].is_empty() => {
Some("consider using `str::is_empty`")
},
(&Anchor(StartText), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
Some("consider using `==` on `str`s")
},
2018-03-13 14:02:40 +00:00
(&Anchor(StartText), &Literal(_)) if is_literal(&exprs[1..]) => Some("consider using `str::starts_with`"),
2018-11-27 20:14:15 +00:00
(&Literal(_), &Anchor(EndText)) if is_literal(&exprs[1..(exprs.len() - 1)]) => {
Some("consider using `str::ends_with`")
},
2018-03-13 14:02:40 +00:00
_ if is_literal(exprs) => Some("consider using `str::contains`"),
2017-09-05 09:33:04 +00:00
_ => None,
2016-12-20 17:21:30 +00:00
},
2016-02-06 17:06:39 +00:00
_ => None,
2016-02-05 22:10:48 +00:00
}
}
2016-05-25 15:15:19 +00:00
2017-09-13 13:34:04 +00:00
fn check_set<'a, 'tcx>(cx: &LateContext<'a, 'tcx>, expr: &'tcx Expr, utf8: bool) {
if_chain! {
2018-07-12 07:30:57 +00:00
if let ExprKind::AddrOf(_, ref expr) = expr.node;
if let ExprKind::Array(ref exprs) = expr.node;
then {
for expr in exprs {
check_regex(cx, expr, utf8);
}
2016-05-25 15:15:19 +00:00
}
}
2016-05-25 15:15:19 +00:00
}
2017-09-13 13:34:04 +00:00
fn check_regex<'a, 'tcx>(cx: &LateContext<'a, 'tcx>, expr: &'tcx Expr, utf8: bool) {
2018-04-07 20:18:51 +00:00
let mut parser = regex_syntax::ParserBuilder::new()
.unicode(utf8)
.allow_invalid_utf8(!utf8)
.build();
2016-05-25 15:15:19 +00:00
2018-07-12 07:30:57 +00:00
if let ExprKind::Lit(ref lit) = expr.node {
if let LitKind::Str(ref r, style) = lit.node {
let r = &r.as_str();
2018-03-13 14:02:40 +00:00
let offset = if let StrStyle::Raw(n) = style { 2 + n } else { 1 };
match parser.parse(r) {
2018-11-27 20:14:15 +00:00
Ok(r) => {
if let Some(repl) = is_trivial_regex(&r) {
span_help_and_lint(cx, TRIVIAL_REGEX, expr.span, "trivial regex", repl);
}
2018-03-13 14:02:40 +00:00
},
Err(regex_syntax::Error::Parse(e)) => {
span_lint(
cx,
INVALID_REGEX,
str_span(expr.span, *e.span(), offset),
&format!("regex syntax error: {}", e.kind()),
);
},
Err(regex_syntax::Error::Translate(e)) => {
span_lint(
cx,
INVALID_REGEX,
str_span(expr.span, *e.span(), offset),
&format!("regex syntax error: {}", e.kind()),
2017-09-05 09:33:04 +00:00
);
2016-12-20 17:21:30 +00:00
},
2016-05-25 15:15:19 +00:00
Err(e) => {
2018-11-27 20:14:15 +00:00
span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));
2016-12-20 17:21:30 +00:00
},
2016-05-25 15:15:19 +00:00
}
}
} else if let Some(r) = const_str(cx, expr) {
2018-03-13 14:02:40 +00:00
match parser.parse(&r) {
2018-11-27 20:14:15 +00:00
Ok(r) => {
if let Some(repl) = is_trivial_regex(&r) {
span_help_and_lint(cx, TRIVIAL_REGEX, expr.span, "trivial regex", repl);
}
2018-03-13 14:02:40 +00:00
},
Err(regex_syntax::Error::Parse(e)) => {
span_lint(
cx,
INVALID_REGEX,
expr.span,
&format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
);
},
Err(regex_syntax::Error::Translate(e)) => {
span_lint(
cx,
INVALID_REGEX,
expr.span,
&format!("regex syntax error on position {}: {}", e.span().start.offset, e.kind()),
2017-09-05 09:33:04 +00:00
);
2016-12-20 17:21:30 +00:00
},
2016-05-25 15:15:19 +00:00
Err(e) => {
2018-11-27 20:14:15 +00:00
span_lint(cx, INVALID_REGEX, expr.span, &format!("regex syntax error: {}", e));
2016-12-20 17:21:30 +00:00
},
2016-05-25 15:15:19 +00:00
}
}
}