blob: e4c91b7efd2bd2c6c4712d7c24683ab1e2e1a42b [file] [log] [blame] [edit]
use clippy_config::Conf;
use clippy_utils::consts::{ConstEvalCtxt, Constant};
use clippy_utils::diagnostics::{span_lint, span_lint_and_sugg, span_lint_and_then};
use clippy_utils::msrvs::{self, Msrv};
use clippy_utils::res::{MaybeQPath, MaybeResPath};
use clippy_utils::source::{SpanRangeExt, snippet, snippet_with_applicability};
use clippy_utils::sugg::Sugg;
use clippy_utils::ty::implements_trait;
use clippy_utils::{expr_use_ctxt, fn_def_id, get_parent_expr, higher, is_in_const_context, is_integer_const};
use rustc_ast::Mutability;
use rustc_ast::ast::RangeLimits;
use rustc_errors::Applicability;
use rustc_hir::{BinOpKind, Expr, ExprKind, HirId, LangItem, Node};
use rustc_lint::{LateContext, LateLintPass, Lint};
use rustc_middle::ty::{self, ClauseKind, GenericArgKind, PredicatePolarity, Ty};
use rustc_session::impl_lint_pass;
use rustc_span::source_map::Spanned;
use rustc_span::{Span, sym};
use std::cmp::Ordering;
declare_clippy_lint! {
/// ### What it does
/// Checks for exclusive ranges where 1 is added to the
/// upper bound, e.g., `x..(y+1)`.
///
/// ### Why is this bad?
/// The code is more readable with an inclusive range
/// like `x..=y`.
///
/// ### Limitations
/// The lint is conservative and will trigger only when switching
/// from an exclusive to an inclusive range is provably safe from
/// a typing point of view. This corresponds to situations where
/// the range is used as an iterator, or for indexing.
///
/// ### Known problems
/// Will add unnecessary pair of parentheses when the
/// expression is not wrapped in a pair but starts with an opening parenthesis
/// and ends with a closing one.
/// I.e., `let _ = (f()+1)..(f()+1)` results in `let _ = ((f()+1)..=f())`.
///
/// Also in many cases, inclusive ranges are still slower to run than
/// exclusive ranges, because they essentially add an extra branch that
/// LLVM may fail to hoist out of the loop.
///
/// ### Example
/// ```no_run
/// # let x = 0;
/// # let y = 1;
/// for i in x..(y+1) {
/// // ..
/// }
/// ```
///
/// Use instead:
/// ```no_run
/// # let x = 0;
/// # let y = 1;
/// for i in x..=y {
/// // ..
/// }
/// ```
#[clippy::version = "pre 1.29.0"]
pub RANGE_PLUS_ONE,
pedantic,
"`x..(y+1)` reads better as `x..=y`"
}
declare_clippy_lint! {
/// ### What it does
/// Checks for inclusive ranges where 1 is subtracted from
/// the upper bound, e.g., `x..=(y-1)`.
///
/// ### Why is this bad?
/// The code is more readable with an exclusive range
/// like `x..y`.
///
/// ### Limitations
/// The lint is conservative and will trigger only when switching
/// from an inclusive to an exclusive range is provably safe from
/// a typing point of view. This corresponds to situations where
/// the range is used as an iterator, or for indexing.
///
/// ### Example
/// ```no_run
/// # let x = 0;
/// # let y = 1;
/// for i in x..=(y-1) {
/// // ..
/// }
/// ```
///
/// Use instead:
/// ```no_run
/// # let x = 0;
/// # let y = 1;
/// for i in x..y {
/// // ..
/// }
/// ```
#[clippy::version = "pre 1.29.0"]
pub RANGE_MINUS_ONE,
pedantic,
"`x..=(y-1)` reads better as `x..y`"
}
declare_clippy_lint! {
/// ### What it does
/// Checks for range expressions `x..y` where both `x` and `y`
/// are constant and `x` is greater to `y`. Also triggers if `x` is equal to `y` when they are conditions to a `for` loop.
///
/// ### Why is this bad?
/// Empty ranges yield no values so iterating them is a no-op.
/// Moreover, trying to use a reversed range to index a slice will panic at run-time.
///
/// ### Example
/// ```rust,no_run
/// fn main() {
/// (10..=0).for_each(|x| println!("{}", x));
///
/// let arr = [1, 2, 3, 4, 5];
/// let sub = &arr[3..1];
/// }
/// ```
/// Use instead:
/// ```no_run
/// fn main() {
/// (0..=10).rev().for_each(|x| println!("{}", x));
///
/// let arr = [1, 2, 3, 4, 5];
/// let sub = &arr[1..3];
/// }
/// ```
#[clippy::version = "1.45.0"]
pub REVERSED_EMPTY_RANGES,
correctness,
"reversing the limits of range expressions, resulting in empty ranges"
}
declare_clippy_lint! {
/// ### What it does
/// Checks for expressions like `x >= 3 && x < 8` that could
/// be more readably expressed as `(3..8).contains(x)`.
///
/// ### Why is this bad?
/// `contains` expresses the intent better and has less
/// failure modes (such as fencepost errors or using `||` instead of `&&`).
///
/// ### Example
/// ```no_run
/// // given
/// let x = 6;
///
/// assert!(x >= 3 && x < 8);
/// ```
/// Use instead:
/// ```no_run
///# let x = 6;
/// assert!((3..8).contains(&x));
/// ```
#[clippy::version = "1.49.0"]
pub MANUAL_RANGE_CONTAINS,
style,
"manually reimplementing {`Range`, `RangeInclusive`}`::contains`"
}
pub struct Ranges {
msrv: Msrv,
}
impl Ranges {
pub fn new(conf: &'static Conf) -> Self {
Self { msrv: conf.msrv }
}
}
impl_lint_pass!(Ranges => [
RANGE_PLUS_ONE,
RANGE_MINUS_ONE,
REVERSED_EMPTY_RANGES,
MANUAL_RANGE_CONTAINS,
]);
impl<'tcx> LateLintPass<'tcx> for Ranges {
fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
if let ExprKind::Binary(ref op, l, r) = expr.kind
&& self.msrv.meets(cx, msrvs::RANGE_CONTAINS)
{
check_possible_range_contains(cx, op.node, l, r, expr, expr.span);
}
check_exclusive_range_plus_one(cx, expr);
check_inclusive_range_minus_one(cx, expr);
check_reversed_empty_range(cx, expr);
}
}
fn check_possible_range_contains(
cx: &LateContext<'_>,
op: BinOpKind,
left: &Expr<'_>,
right: &Expr<'_>,
expr: &Expr<'_>,
span: Span,
) {
if is_in_const_context(cx) {
return;
}
let combine_and = match op {
BinOpKind::And | BinOpKind::BitAnd => true,
BinOpKind::Or | BinOpKind::BitOr => false,
_ => return,
};
// value, name, order (higher/lower), inclusiveness
if let (Some(l), Some(r)) = (check_range_bounds(cx, left), check_range_bounds(cx, right)) {
// we only lint comparisons on the same name and with different
// direction
if l.id != r.id || l.ord == r.ord {
return;
}
let ord = Constant::partial_cmp(cx.tcx, cx.typeck_results().expr_ty(l.expr), &l.val, &r.val);
if combine_and && ord == Some(r.ord) {
// order lower bound and upper bound
let (l_span, u_span, l_inc, u_inc) = if r.ord == Ordering::Less {
(l.val_span, r.val_span, l.inc, r.inc)
} else {
(r.val_span, l.val_span, r.inc, l.inc)
};
// we only lint inclusive lower bounds
if !l_inc {
return;
}
let (range_type, range_op) = if u_inc {
("RangeInclusive", "..=")
} else {
("Range", "..")
};
let mut applicability = Applicability::MachineApplicable;
let name = snippet_with_applicability(cx, l.name_span, "_", &mut applicability);
let lo = snippet_with_applicability(cx, l_span, "_", &mut applicability);
let hi = snippet_with_applicability(cx, u_span, "_", &mut applicability);
let space = if lo.ends_with('.') { " " } else { "" };
span_lint_and_sugg(
cx,
MANUAL_RANGE_CONTAINS,
span,
format!("manual `{range_type}::contains` implementation"),
"use",
format!("({lo}{space}{range_op}{hi}).contains(&{name})"),
applicability,
);
} else if !combine_and && ord == Some(l.ord) {
// `!_.contains(_)`
// order lower bound and upper bound
let (l_span, u_span, l_inc, u_inc) = if l.ord == Ordering::Less {
(l.val_span, r.val_span, l.inc, r.inc)
} else {
(r.val_span, l.val_span, r.inc, l.inc)
};
if l_inc {
return;
}
let (range_type, range_op) = if u_inc {
("Range", "..")
} else {
("RangeInclusive", "..=")
};
let mut applicability = Applicability::MachineApplicable;
let name = snippet_with_applicability(cx, l.name_span, "_", &mut applicability);
let lo = snippet_with_applicability(cx, l_span, "_", &mut applicability);
let hi = snippet_with_applicability(cx, u_span, "_", &mut applicability);
let space = if lo.ends_with('.') { " " } else { "" };
span_lint_and_sugg(
cx,
MANUAL_RANGE_CONTAINS,
span,
format!("manual `!{range_type}::contains` implementation"),
"use",
format!("!({lo}{space}{range_op}{hi}).contains(&{name})"),
applicability,
);
}
}
// If the LHS is the same operator, we have to recurse to get the "real" RHS, since they have
// the same operator precedence
if let ExprKind::Binary(ref lhs_op, _left, new_lhs) = left.kind
&& op == lhs_op.node
&& let new_span = Span::new(new_lhs.span.lo(), right.span.hi(), expr.span.ctxt(), expr.span.parent())
&& new_span.check_source_text(cx, |src| {
// Do not continue if we have mismatched number of parens, otherwise the suggestion is wrong
src.matches('(').count() == src.matches(')').count()
})
{
check_possible_range_contains(cx, op, new_lhs, right, expr, new_span);
}
}
struct RangeBounds<'a> {
val: Constant,
expr: &'a Expr<'a>,
id: HirId,
name_span: Span,
val_span: Span,
ord: Ordering,
inc: bool,
}
// Takes a binary expression such as x <= 2 as input
// Breaks apart into various pieces, such as the value of the number,
// hir id of the variable, and direction/inclusiveness of the operator
fn check_range_bounds<'a>(cx: &'a LateContext<'_>, ex: &'a Expr<'_>) -> Option<RangeBounds<'a>> {
if let ExprKind::Binary(ref op, l, r) = ex.kind {
let (inclusive, ordering) = match op.node {
BinOpKind::Gt => (false, Ordering::Greater),
BinOpKind::Ge => (true, Ordering::Greater),
BinOpKind::Lt => (false, Ordering::Less),
BinOpKind::Le => (true, Ordering::Less),
_ => return None,
};
if let Some(id) = l.res_local_id() {
if let Some(c) = ConstEvalCtxt::new(cx).eval(r) {
return Some(RangeBounds {
val: c,
expr: r,
id,
name_span: l.span,
val_span: r.span,
ord: ordering,
inc: inclusive,
});
}
} else if let Some(id) = r.res_local_id()
&& let Some(c) = ConstEvalCtxt::new(cx).eval(l)
{
return Some(RangeBounds {
val: c,
expr: l,
id,
name_span: r.span,
val_span: l.span,
ord: ordering.reverse(),
inc: inclusive,
});
}
}
None
}
/// Check whether `expr` could switch range types without breaking the typing requirements. This is
/// generally the case when `expr` is used as an iterator for example, or as a slice or `&str`
/// index.
///
/// FIXME: Note that the current implementation may still return false positives. A proper fix would
/// check that the obligations are still satisfied after switching the range type.
fn can_switch_ranges<'tcx>(
cx: &LateContext<'tcx>,
expr: &'tcx Expr<'_>,
original: RangeLimits,
inner_ty: Ty<'tcx>,
) -> bool {
let use_ctxt = expr_use_ctxt(cx, expr);
let (Node::Expr(parent_expr), false) = (use_ctxt.node, use_ctxt.is_ty_unified) else {
return false;
};
// Check if `expr` is the argument of a compiler-generated `IntoIter::into_iter(expr)`
if let ExprKind::Call(func, [arg]) = parent_expr.kind
&& arg.hir_id == use_ctxt.child_id
&& func.opt_lang_path() == Some(LangItem::IntoIterIntoIter)
{
return true;
}
// Check if `expr` is used as the receiver of a method of the `Iterator`, `IntoIterator`,
// or `RangeBounds` traits.
if let ExprKind::MethodCall(_, receiver, _, _) = parent_expr.kind
&& receiver.hir_id == use_ctxt.child_id
&& let Some(method_did) = cx.typeck_results().type_dependent_def_id(parent_expr.hir_id)
&& let Some(trait_did) = cx.tcx.trait_of_assoc(method_did)
&& matches!(
cx.tcx.get_diagnostic_name(trait_did),
Some(sym::Iterator | sym::IntoIterator | sym::RangeBounds)
)
{
return true;
}
// Check if `expr` is an argument of a call which requires an `Iterator`, `IntoIterator`,
// or `RangeBounds` trait.
if let ExprKind::Call(_, args) | ExprKind::MethodCall(_, _, args, _) = parent_expr.kind
&& let Some(id) = fn_def_id(cx, parent_expr)
&& let Some(arg_idx) = args.iter().position(|e| e.hir_id == use_ctxt.child_id)
{
let input_idx = if matches!(parent_expr.kind, ExprKind::MethodCall(..)) {
arg_idx + 1
} else {
arg_idx
};
let inputs = cx
.tcx
.liberate_late_bound_regions(id, cx.tcx.fn_sig(id).instantiate_identity())
.inputs();
let expr_ty = inputs[input_idx];
// Check that the `expr` type is present only once, otherwise modifying just one of them might be
// risky if they are referenced using the same generic type for example.
if inputs.iter().enumerate().all(|(n, ty)|
n == input_idx
|| !ty.walk().any(|arg| matches!(arg.kind(),
GenericArgKind::Type(ty) if ty == expr_ty)))
// Look for a clause requiring `Iterator`, `IntoIterator`, or `RangeBounds`, and resolving to `expr_type`.
&& cx
.tcx
.param_env(id)
.caller_bounds()
.into_iter()
.any(|p| {
if let ClauseKind::Trait(t) = p.kind().skip_binder()
&& t.polarity == PredicatePolarity::Positive
&& matches!(
cx.tcx.get_diagnostic_name(t.trait_ref.def_id),
Some(sym::Iterator | sym::IntoIterator | sym::RangeBounds)
)
{
t.self_ty() == expr_ty
} else {
false
}
})
{
return true;
}
}
// Check if `expr` is used for indexing, and if the switched range type could be used
// as well.
if let ExprKind::Index(outer_expr, index, _) = parent_expr.kind
&& index.hir_id == expr.hir_id
// Build the switched range type (for example `RangeInclusive<usize>`).
&& let Some(switched_range_def_id) = match original {
RangeLimits::HalfOpen => cx.tcx.lang_items().range_inclusive_struct(),
RangeLimits::Closed => cx.tcx.lang_items().range_struct(),
}
&& let switched_range_ty = cx
.tcx
.type_of(switched_range_def_id)
.instantiate(cx.tcx, &[inner_ty.into()])
// Check that the switched range type can be used for indexing the original expression
// through the `Index` or `IndexMut` trait.
&& let ty::Ref(_, outer_ty, mutability) = cx.typeck_results().expr_ty_adjusted(outer_expr).kind()
&& let Some(index_def_id) = match mutability {
Mutability::Not => cx.tcx.lang_items().index_trait(),
Mutability::Mut => cx.tcx.lang_items().index_mut_trait(),
}
&& implements_trait(cx, *outer_ty, index_def_id, &[switched_range_ty.into()])
// We could also check that the associated item of the `index_def_id` trait with the switched range type
// return the same type, but it is reasonable to expect so. We can't check that the result is identical
// in both `Index<Range<…>>` and `Index<RangeInclusive<…>>` anyway.
{
return true;
}
false
}
// exclusive range plus one: `x..(y+1)`
fn check_exclusive_range_plus_one<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
check_range_switch(
cx,
expr,
RangeLimits::HalfOpen,
y_plus_one,
RANGE_PLUS_ONE,
"an inclusive range would be more readable",
"..=",
);
}
// inclusive range minus one: `x..=(y-1)`
fn check_inclusive_range_minus_one<'tcx>(cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) {
check_range_switch(
cx,
expr,
RangeLimits::Closed,
y_minus_one,
RANGE_MINUS_ONE,
"an exclusive range would be more readable",
"..",
);
}
/// Check for a `kind` of range in `expr`, check for `predicate` on the end,
/// and emit the `lint` with `msg` and the `operator`.
fn check_range_switch<'tcx>(
cx: &LateContext<'tcx>,
expr: &'tcx Expr<'_>,
kind: RangeLimits,
predicate: impl for<'hir> FnOnce(&LateContext<'_>, &Expr<'hir>) -> Option<&'hir Expr<'hir>>,
lint: &'static Lint,
msg: &'static str,
operator: &str,
) {
if expr.span.can_be_used_for_suggestions()
&& let Some(higher::Range {
start,
end: Some(end),
limits,
}) = higher::Range::hir(expr)
&& limits == kind
&& let Some(y) = predicate(cx, end)
&& can_switch_ranges(cx, expr, kind, cx.typeck_results().expr_ty(y))
{
let span = expr.span;
span_lint_and_then(cx, lint, span, msg, |diag| {
let mut app = Applicability::MachineApplicable;
let start = start.map_or(String::new(), |x| {
Sugg::hir_with_applicability(cx, x, "<x>", &mut app)
.maybe_paren()
.to_string()
});
let end = Sugg::hir_with_applicability(cx, y, "<y>", &mut app).maybe_paren();
match span.with_source_text(cx, |src| src.starts_with('(') && src.ends_with(')')) {
Some(true) => {
diag.span_suggestion(span, "use", format!("({start}{operator}{end})"), app);
},
Some(false) => {
diag.span_suggestion(span, "use", format!("{start}{operator}{end}"), app);
},
None => {},
}
});
}
}
fn check_reversed_empty_range(cx: &LateContext<'_>, expr: &Expr<'_>) {
fn inside_indexing_expr(cx: &LateContext<'_>, expr: &Expr<'_>) -> bool {
matches!(
get_parent_expr(cx, expr),
Some(Expr {
kind: ExprKind::Index(..),
..
})
)
}
fn is_for_loop_arg(cx: &LateContext<'_>, expr: &Expr<'_>) -> bool {
let mut cur_expr = expr;
while let Some(parent_expr) = get_parent_expr(cx, cur_expr) {
match higher::ForLoop::hir(parent_expr) {
Some(higher::ForLoop { arg, .. }) if arg.hir_id == expr.hir_id => return true,
_ => cur_expr = parent_expr,
}
}
false
}
fn is_empty_range(limits: RangeLimits, ordering: Ordering) -> bool {
match limits {
RangeLimits::HalfOpen => ordering != Ordering::Less,
RangeLimits::Closed => ordering == Ordering::Greater,
}
}
if let Some(higher::Range {
start: Some(start),
end: Some(end),
limits,
}) = higher::Range::hir(expr)
&& let ty = cx.typeck_results().expr_ty(start)
&& let ty::Int(_) | ty::Uint(_) = ty.kind()
&& let ecx = ConstEvalCtxt::new(cx)
&& let Some(start_idx) = ecx.eval(start)
&& let Some(end_idx) = ecx.eval(end)
&& let Some(ordering) = Constant::partial_cmp(cx.tcx, ty, &start_idx, &end_idx)
&& is_empty_range(limits, ordering)
{
if inside_indexing_expr(cx, expr) {
// Avoid linting `N..N` as it has proven to be useful, see #5689 and #5628 ...
if ordering != Ordering::Equal {
span_lint(
cx,
REVERSED_EMPTY_RANGES,
expr.span,
"this range is reversed and using it to index a slice will panic at run-time",
);
}
// ... except in for loop arguments for backwards compatibility with `reverse_range_loop`
} else if ordering != Ordering::Equal || is_for_loop_arg(cx, expr) {
span_lint_and_then(
cx,
REVERSED_EMPTY_RANGES,
expr.span,
"this range is empty so it will yield no values",
|diag| {
if ordering != Ordering::Equal {
let start_snippet = snippet(cx, start.span, "_");
let end_snippet = snippet(cx, end.span, "_");
let dots = match limits {
RangeLimits::HalfOpen => "..",
RangeLimits::Closed => "..=",
};
diag.span_suggestion(
expr.span,
"consider using the following if you are attempting to iterate over this \
range in reverse",
format!("({end_snippet}{dots}{start_snippet}).rev()"),
Applicability::MaybeIncorrect,
);
}
},
);
}
}
}
fn y_plus_one<'tcx>(cx: &LateContext<'_>, expr: &Expr<'tcx>) -> Option<&'tcx Expr<'tcx>> {
match expr.kind {
ExprKind::Binary(
Spanned {
node: BinOpKind::Add, ..
},
lhs,
rhs,
) => {
if is_integer_const(cx, lhs, 1) {
Some(rhs)
} else if is_integer_const(cx, rhs, 1) {
Some(lhs)
} else {
None
}
},
_ => None,
}
}
fn y_minus_one<'tcx>(cx: &LateContext<'_>, expr: &Expr<'tcx>) -> Option<&'tcx Expr<'tcx>> {
match expr.kind {
ExprKind::Binary(
Spanned {
node: BinOpKind::Sub, ..
},
lhs,
rhs,
) if is_integer_const(cx, rhs, 1) => Some(lhs),
_ => None,
}
}