rust-clippy/clippy_lints/src/suspicious_operation_groupings.rs

692 lines
24 KiB
Rust

use clippy_utils::ast_utils::{eq_id, is_useless_with_eq_exprs, IdentIter};
use clippy_utils::diagnostics::span_lint_and_sugg;
use clippy_utils::source::snippet_with_applicability;
use core::ops::{Add, AddAssign};
use if_chain::if_chain;
use rustc_ast::ast::{BinOpKind, Expr, ExprKind, StmtKind};
use rustc_data_structures::fx::FxHashSet;
use rustc_errors::Applicability;
use rustc_lint::{EarlyContext, EarlyLintPass};
use rustc_session::{declare_lint_pass, declare_tool_lint};
use rustc_span::source_map::Spanned;
use rustc_span::symbol::Ident;
use rustc_span::Span;
declare_clippy_lint! {
/// **What it does:**
/// Checks for unlikely usages of binary operators that are almost
/// certainly typos and/or copy/paste errors, given the other usages
/// of binary operators nearby.
/// **Why is this bad?**
/// They are probably bugs and if they aren't then they look like bugs
/// and you should add a comment explaining why you are doing such an
/// odd set of operations.
/// **Known problems:**
/// There may be some false positives if you are trying to do something
/// unusual that happens to look like a typo.
///
/// **Example:**
///
/// ```rust
/// struct Vec3 {
/// x: f64,
/// y: f64,
/// z: f64,
/// }
///
/// impl Eq for Vec3 {}
///
/// impl PartialEq for Vec3 {
/// fn eq(&self, other: &Self) -> bool {
/// // This should trigger the lint because `self.x` is compared to `other.y`
/// self.x == other.y && self.y == other.y && self.z == other.z
/// }
/// }
/// ```
/// Use instead:
/// ```rust
/// # struct Vec3 {
/// # x: f64,
/// # y: f64,
/// # z: f64,
/// # }
/// // same as above except:
/// impl PartialEq for Vec3 {
/// fn eq(&self, other: &Self) -> bool {
/// // Note we now compare other.x to self.x
/// self.x == other.x && self.y == other.y && self.z == other.z
/// }
/// }
/// ```
pub SUSPICIOUS_OPERATION_GROUPINGS,
nursery,
"groupings of binary operations that look suspiciously like typos"
}
declare_lint_pass!(SuspiciousOperationGroupings => [SUSPICIOUS_OPERATION_GROUPINGS]);
impl EarlyLintPass for SuspiciousOperationGroupings {
fn check_expr(&mut self, cx: &EarlyContext<'_>, expr: &Expr) {
if expr.span.from_expansion() {
return;
}
if let Some(binops) = extract_related_binops(&expr.kind) {
check_binops(cx, &binops.iter().collect::<Vec<_>>());
let mut op_types = Vec::with_capacity(binops.len());
// We could use a hashmap, etc. to avoid being O(n*m) here, but
// we want the lints to be emitted in a consistent order. Besides,
// m, (the number of distinct `BinOpKind`s in `binops`)
// will often be small, and does have an upper limit.
binops.iter().map(|b| b.op).for_each(|op| {
if !op_types.contains(&op) {
op_types.push(op);
}
});
for op_type in op_types {
let ops: Vec<_> = binops.iter().filter(|b| b.op == op_type).collect();
check_binops(cx, &ops);
}
}
}
}
fn check_binops(cx: &EarlyContext<'_>, binops: &[&BinaryOp<'_>]) {
let binop_count = binops.len();
if binop_count < 2 {
// Single binary operation expressions would likely be false
// positives.
return;
}
let mut one_ident_difference_count = 0;
let mut no_difference_info = None;
let mut double_difference_info = None;
let mut expected_ident_loc = None;
let mut paired_identifiers = FxHashSet::default();
for (i, BinaryOp { left, right, op, .. }) in binops.iter().enumerate() {
match ident_difference_expr(left, right) {
IdentDifference::NoDifference => {
if is_useless_with_eq_exprs(*op) {
// The `eq_op` lint should catch this in this case.
return;
}
no_difference_info = Some(i);
},
IdentDifference::Single(ident_loc) => {
one_ident_difference_count += 1;
if let Some(previous_expected) = expected_ident_loc {
if previous_expected != ident_loc {
// This expression doesn't match the form we're
// looking for.
return;
}
} else {
expected_ident_loc = Some(ident_loc);
}
// If there was only a single difference, all other idents
// must have been the same, and thus were paired.
for id in skip_index(IdentIter::from(*left), ident_loc.index) {
paired_identifiers.insert(id);
}
},
IdentDifference::Double(ident_loc1, ident_loc2) => {
double_difference_info = Some((i, ident_loc1, ident_loc2));
},
IdentDifference::Multiple | IdentDifference::NonIdent => {
// It's too hard to know whether this is a bug or not.
return;
},
}
}
let mut applicability = Applicability::MachineApplicable;
if let Some(expected_loc) = expected_ident_loc {
match (no_difference_info, double_difference_info) {
(Some(i), None) => attempt_to_emit_no_difference_lint(cx, binops, i, expected_loc),
(None, Some((double_difference_index, ident_loc1, ident_loc2))) => {
if_chain! {
if one_ident_difference_count == binop_count - 1;
if let Some(binop) = binops.get(double_difference_index);
then {
let changed_loc = if ident_loc1 == expected_loc {
ident_loc2
} else if ident_loc2 == expected_loc {
ident_loc1
} else {
// This expression doesn't match the form we're
// looking for.
return;
};
if let Some(sugg) = ident_swap_sugg(
cx,
&paired_identifiers,
binop,
changed_loc,
&mut applicability,
) {
emit_suggestion(
cx,
binop.span,
sugg,
applicability,
);
}
}
}
},
_ => {},
}
}
}
fn attempt_to_emit_no_difference_lint(
cx: &EarlyContext<'_>,
binops: &[&BinaryOp<'_>],
i: usize,
expected_loc: IdentLocation,
) {
if let Some(binop) = binops.get(i).copied() {
// We need to try and figure out which identifier we should
// suggest using instead. Since there could be multiple
// replacement candidates in a given expression, and we're
// just taking the first one, we may get some bad lint
// messages.
let mut applicability = Applicability::MaybeIncorrect;
// We assume that the correct ident is one used elsewhere in
// the other binops, in a place that there was a single
// difference between idents before.
let old_left_ident = get_ident(binop.left, expected_loc);
let old_right_ident = get_ident(binop.right, expected_loc);
for b in skip_index(binops.iter(), i) {
if_chain! {
if let (Some(old_ident), Some(new_ident)) =
(old_left_ident, get_ident(b.left, expected_loc));
if old_ident != new_ident;
if let Some(sugg) = suggestion_with_swapped_ident(
cx,
binop.left,
expected_loc,
new_ident,
&mut applicability,
);
then {
emit_suggestion(
cx,
binop.span,
replace_left_sugg(cx, binop, &sugg, &mut applicability),
applicability,
);
return;
}
}
if_chain! {
if let (Some(old_ident), Some(new_ident)) =
(old_right_ident, get_ident(b.right, expected_loc));
if old_ident != new_ident;
if let Some(sugg) = suggestion_with_swapped_ident(
cx,
binop.right,
expected_loc,
new_ident,
&mut applicability,
);
then {
emit_suggestion(
cx,
binop.span,
replace_right_sugg(cx, binop, &sugg, &mut applicability),
applicability,
);
return;
}
}
}
}
}
fn emit_suggestion(cx: &EarlyContext<'_>, span: Span, sugg: String, applicability: Applicability) {
span_lint_and_sugg(
cx,
SUSPICIOUS_OPERATION_GROUPINGS,
span,
"this sequence of operators looks suspiciously like a bug",
"did you mean",
sugg,
applicability,
);
}
fn ident_swap_sugg(
cx: &EarlyContext<'_>,
paired_identifiers: &FxHashSet<Ident>,
binop: &BinaryOp<'_>,
location: IdentLocation,
applicability: &mut Applicability,
) -> Option<String> {
let left_ident = get_ident(binop.left, location)?;
let right_ident = get_ident(binop.right, location)?;
let sugg = match (
paired_identifiers.contains(&left_ident),
paired_identifiers.contains(&right_ident),
) {
(true, true) | (false, false) => {
// We don't have a good guess of what ident should be
// used instead, in these cases.
*applicability = Applicability::MaybeIncorrect;
// We arbitraily choose one side to suggest changing,
// since we don't have a better guess. If the user
// ends up duplicating a clause, the `logic_bug` lint
// should catch it.
let right_suggestion = suggestion_with_swapped_ident(cx, binop.right, location, left_ident, applicability)?;
replace_right_sugg(cx, binop, &right_suggestion, applicability)
},
(false, true) => {
// We haven't seen a pair involving the left one, so
// it's probably what is wanted.
let right_suggestion = suggestion_with_swapped_ident(cx, binop.right, location, left_ident, applicability)?;
replace_right_sugg(cx, binop, &right_suggestion, applicability)
},
(true, false) => {
// We haven't seen a pair involving the right one, so
// it's probably what is wanted.
let left_suggestion = suggestion_with_swapped_ident(cx, binop.left, location, right_ident, applicability)?;
replace_left_sugg(cx, binop, &left_suggestion, applicability)
},
};
Some(sugg)
}
fn replace_left_sugg(
cx: &EarlyContext<'_>,
binop: &BinaryOp<'_>,
left_suggestion: &str,
applicability: &mut Applicability,
) -> String {
format!(
"{} {} {}",
left_suggestion,
binop.op.to_string(),
snippet_with_applicability(cx, binop.right.span, "..", applicability),
)
}
fn replace_right_sugg(
cx: &EarlyContext<'_>,
binop: &BinaryOp<'_>,
right_suggestion: &str,
applicability: &mut Applicability,
) -> String {
format!(
"{} {} {}",
snippet_with_applicability(cx, binop.left.span, "..", applicability),
binop.op.to_string(),
right_suggestion,
)
}
#[derive(Clone, Debug)]
struct BinaryOp<'exprs> {
op: BinOpKind,
span: Span,
left: &'exprs Expr,
right: &'exprs Expr,
}
impl BinaryOp<'exprs> {
fn new(op: BinOpKind, span: Span, (left, right): (&'exprs Expr, &'exprs Expr)) -> Self {
Self { op, span, left, right }
}
}
fn strip_non_ident_wrappers(expr: &Expr) -> &Expr {
let mut output = expr;
loop {
output = match &output.kind {
ExprKind::Paren(ref inner) | ExprKind::Unary(_, ref inner) => inner,
_ => {
return output;
},
};
}
}
fn extract_related_binops(kind: &ExprKind) -> Option<Vec<BinaryOp<'_>>> {
append_opt_vecs(chained_binops(kind), if_statment_binops(kind))
}
fn if_statment_binops(kind: &ExprKind) -> Option<Vec<BinaryOp<'_>>> {
match kind {
ExprKind::If(ref condition, _, _) => chained_binops(&condition.kind),
ExprKind::Paren(ref e) => if_statment_binops(&e.kind),
ExprKind::Block(ref block, _) => {
let mut output = None;
for stmt in &block.stmts {
match stmt.kind {
StmtKind::Expr(ref e) | StmtKind::Semi(ref e) => {
output = append_opt_vecs(output, if_statment_binops(&e.kind));
},
_ => {},
}
}
output
},
_ => None,
}
}
fn append_opt_vecs<A>(target_opt: Option<Vec<A>>, source_opt: Option<Vec<A>>) -> Option<Vec<A>> {
match (target_opt, source_opt) {
(Some(mut target), Some(mut source)) => {
target.reserve(source.len());
for op in source.drain(..) {
target.push(op);
}
Some(target)
},
(Some(v), None) | (None, Some(v)) => Some(v),
(None, None) => None,
}
}
fn chained_binops(kind: &ExprKind) -> Option<Vec<BinaryOp<'_>>> {
match kind {
ExprKind::Binary(_, left_outer, right_outer) => chained_binops_helper(left_outer, right_outer),
ExprKind::Paren(ref e) | ExprKind::Unary(_, ref e) => chained_binops(&e.kind),
_ => None,
}
}
fn chained_binops_helper(left_outer: &'expr Expr, right_outer: &'expr Expr) -> Option<Vec<BinaryOp<'expr>>> {
match (&left_outer.kind, &right_outer.kind) {
(
ExprKind::Paren(ref left_e) | ExprKind::Unary(_, ref left_e),
ExprKind::Paren(ref right_e) | ExprKind::Unary(_, ref right_e),
) => chained_binops_helper(left_e, right_e),
(ExprKind::Paren(ref left_e) | ExprKind::Unary(_, ref left_e), _) => chained_binops_helper(left_e, right_outer),
(_, ExprKind::Paren(ref right_e) | ExprKind::Unary(_, ref right_e)) => {
chained_binops_helper(left_outer, right_e)
},
(
ExprKind::Binary(Spanned { node: left_op, .. }, ref left_left, ref left_right),
ExprKind::Binary(Spanned { node: right_op, .. }, ref right_left, ref right_right),
) => match (
chained_binops_helper(left_left, left_right),
chained_binops_helper(right_left, right_right),
) {
(Some(mut left_ops), Some(mut right_ops)) => {
left_ops.reserve(right_ops.len());
for op in right_ops.drain(..) {
left_ops.push(op);
}
Some(left_ops)
},
(Some(mut left_ops), _) => {
left_ops.push(BinaryOp::new(*right_op, right_outer.span, (right_left, right_right)));
Some(left_ops)
},
(_, Some(mut right_ops)) => {
right_ops.insert(0, BinaryOp::new(*left_op, left_outer.span, (left_left, left_right)));
Some(right_ops)
},
(None, None) => Some(vec![
BinaryOp::new(*left_op, left_outer.span, (left_left, left_right)),
BinaryOp::new(*right_op, right_outer.span, (right_left, right_right)),
]),
},
_ => None,
}
}
#[derive(Clone, Copy, PartialEq, Eq, Default, Debug)]
struct IdentLocation {
index: usize,
}
impl Add for IdentLocation {
type Output = IdentLocation;
fn add(self, other: Self) -> Self::Output {
Self {
index: self.index + other.index,
}
}
}
impl AddAssign for IdentLocation {
fn add_assign(&mut self, other: Self) {
*self = *self + other;
}
}
#[derive(Clone, Copy, Debug)]
enum IdentDifference {
NoDifference,
Single(IdentLocation),
Double(IdentLocation, IdentLocation),
Multiple,
NonIdent,
}
impl Add for IdentDifference {
type Output = IdentDifference;
fn add(self, other: Self) -> Self::Output {
match (self, other) {
(Self::NoDifference, output) | (output, Self::NoDifference) => output,
(Self::Multiple, _)
| (_, Self::Multiple)
| (Self::Double(_, _), Self::Single(_))
| (Self::Single(_) | Self::Double(_, _), Self::Double(_, _)) => Self::Multiple,
(Self::NonIdent, _) | (_, Self::NonIdent) => Self::NonIdent,
(Self::Single(il1), Self::Single(il2)) => Self::Double(il1, il2),
}
}
}
impl AddAssign for IdentDifference {
fn add_assign(&mut self, other: Self) {
*self = *self + other;
}
}
impl IdentDifference {
/// Returns true if learning about more differences will not change the value
/// of this `IdentDifference`, and false otherwise.
fn is_complete(&self) -> bool {
match self {
Self::NoDifference | Self::Single(_) | Self::Double(_, _) => false,
Self::Multiple | Self::NonIdent => true,
}
}
}
fn ident_difference_expr(left: &Expr, right: &Expr) -> IdentDifference {
ident_difference_expr_with_base_location(left, right, IdentLocation::default()).0
}
fn ident_difference_expr_with_base_location(
left: &Expr,
right: &Expr,
mut base: IdentLocation,
) -> (IdentDifference, IdentLocation) {
// Ideally, this function should not use IdentIter because it should return
// early if the expressions have any non-ident differences. We want that early
// return because if without that restriction the lint would lead to false
// positives.
//
// But, we cannot (easily?) use a `rustc_ast::visit::Visitor`, since we need
// the two expressions to be walked in lockstep. And without a `Visitor`, we'd
// have to do all the AST traversal ourselves, which is a lot of work, since to
// do it properly we'd need to be able to handle more or less every possible
// AST node since `Item`s can be written inside `Expr`s.
//
// In practice, it seems likely that expressions, above a certain size, that
// happen to use the exact same idents in the exact same order, and which are
// not structured the same, would be rare. Therefore it seems likely that if
// we do only the first layer of matching ourselves and eventually fallback on
// IdentIter, then the output of this function will be almost always be correct
// in practice.
//
// If it turns out that problematic cases are more prelavent than we assume,
// then we should be able to change this function to do the correct traversal,
// without needing to change the rest of the code.
#![allow(clippy::enum_glob_use)]
use ExprKind::*;
match (
&strip_non_ident_wrappers(left).kind,
&strip_non_ident_wrappers(right).kind,
) {
(Yield(_), Yield(_))
| (Try(_), Try(_))
| (Paren(_), Paren(_))
| (Repeat(_, _), Repeat(_, _))
| (Struct(_), Struct(_))
| (MacCall(_), MacCall(_))
| (LlvmInlineAsm(_), LlvmInlineAsm(_))
| (InlineAsm(_), InlineAsm(_))
| (Ret(_), Ret(_))
| (Continue(_), Continue(_))
| (Break(_, _), Break(_, _))
| (AddrOf(_, _, _), AddrOf(_, _, _))
| (Path(_, _), Path(_, _))
| (Range(_, _, _), Range(_, _, _))
| (Index(_, _), Index(_, _))
| (Field(_, _), Field(_, _))
| (AssignOp(_, _, _), AssignOp(_, _, _))
| (Assign(_, _, _), Assign(_, _, _))
| (TryBlock(_), TryBlock(_))
| (Await(_), Await(_))
| (Async(_, _, _), Async(_, _, _))
| (Block(_, _), Block(_, _))
| (Closure(_, _, _, _, _, _), Closure(_, _, _, _, _, _))
| (Match(_, _), Match(_, _))
| (Loop(_, _), Loop(_, _))
| (ForLoop(_, _, _, _), ForLoop(_, _, _, _))
| (While(_, _, _), While(_, _, _))
| (If(_, _, _), If(_, _, _))
| (Let(_, _), Let(_, _))
| (Type(_, _), Type(_, _))
| (Cast(_, _), Cast(_, _))
| (Lit(_), Lit(_))
| (Unary(_, _), Unary(_, _))
| (Binary(_, _, _), Binary(_, _, _))
| (Tup(_), Tup(_))
| (MethodCall(_, _, _), MethodCall(_, _, _))
| (Call(_, _), Call(_, _))
| (ConstBlock(_), ConstBlock(_))
| (Array(_), Array(_))
| (Box(_), Box(_)) => {
// keep going
},
_ => {
return (IdentDifference::NonIdent, base);
},
}
let mut difference = IdentDifference::NoDifference;
for (left_attr, right_attr) in left.attrs.iter().zip(right.attrs.iter()) {
let (new_difference, new_base) =
ident_difference_via_ident_iter_with_base_location(left_attr, right_attr, base);
base = new_base;
difference += new_difference;
if difference.is_complete() {
return (difference, base);
}
}
let (new_difference, new_base) = ident_difference_via_ident_iter_with_base_location(left, right, base);
base = new_base;
difference += new_difference;
(difference, base)
}
fn ident_difference_via_ident_iter_with_base_location<Iterable: Into<IdentIter>>(
left: Iterable,
right: Iterable,
mut base: IdentLocation,
) -> (IdentDifference, IdentLocation) {
// See the note in `ident_difference_expr_with_base_location` about `IdentIter`
let mut difference = IdentDifference::NoDifference;
let mut left_iterator = left.into();
let mut right_iterator = right.into();
loop {
match (left_iterator.next(), right_iterator.next()) {
(Some(left_ident), Some(right_ident)) => {
if !eq_id(left_ident, right_ident) {
difference += IdentDifference::Single(base);
if difference.is_complete() {
return (difference, base);
}
}
},
(Some(_), None) | (None, Some(_)) => {
return (IdentDifference::NonIdent, base);
},
(None, None) => {
return (difference, base);
},
}
base += IdentLocation { index: 1 };
}
}
fn get_ident(expr: &Expr, location: IdentLocation) -> Option<Ident> {
IdentIter::from(expr).nth(location.index)
}
fn suggestion_with_swapped_ident(
cx: &EarlyContext<'_>,
expr: &Expr,
location: IdentLocation,
new_ident: Ident,
applicability: &mut Applicability,
) -> Option<String> {
get_ident(expr, location).and_then(|current_ident| {
if eq_id(current_ident, new_ident) {
// We never want to suggest a non-change
return None;
}
Some(format!(
"{}{}{}",
snippet_with_applicability(cx, expr.span.with_hi(current_ident.span.lo()), "..", applicability),
new_ident.to_string(),
snippet_with_applicability(cx, expr.span.with_lo(current_ident.span.hi()), "..", applicability),
))
})
}
fn skip_index<A, Iter>(iter: Iter, index: usize) -> impl Iterator<Item = A>
where
Iter: Iterator<Item = A>,
{
iter.enumerate()
.filter_map(move |(i, a)| if i == index { None } else { Some(a) })
}