use rustc::hir::*; use rustc::lint::{LateContext, LateLintPass, LintArray, LintPass}; use rustc::{declare_lint_pass, declare_tool_lint}; use rustc_errors::Applicability; use crate::utils::*; declare_clippy_lint! { /// **What it does:** Checks for expressions of the form `a * b + c` /// or `c + a * b` where `a`, `b`, `c` are floats and suggests using /// `a.mul_add(b, c)` instead. /// /// **Why is this bad?** Calculating `a * b + c` may lead to slight /// numerical inaccuracies as `a * b` is rounded before being added to /// `c`. Depending on the target architecture, `mul_add()` may be more /// performant. /// /// **Known problems:** None. /// /// **Example:** /// /// ```rust /// # let a = 0_f32; /// # let b = 0_f32; /// # let c = 0_f32; /// let foo = (a * b) + c; /// ``` /// /// can be written as /// /// ```rust /// # let a = 0_f32; /// # let b = 0_f32; /// # let c = 0_f32; /// let foo = a.mul_add(b, c); /// ``` pub MANUAL_MUL_ADD, perf, "Using `a.mul_add(b, c)` for floating points has higher numerical precision than `a * b + c`" } declare_lint_pass!(MulAddCheck => [MANUAL_MUL_ADD]); fn is_float<'a, 'tcx>(cx: &LateContext<'a, 'tcx>, expr: &Expr) -> bool { cx.tables.expr_ty(expr).is_floating_point() } // Checks whether expression is multiplication of two floats fn is_float_mult_expr<'a, 'tcx, 'b>(cx: &LateContext<'a, 'tcx>, expr: &'b Expr) -> Option<(&'b Expr, &'b Expr)> { if let ExprKind::Binary(op, lhs, rhs) = &expr.kind { if let BinOpKind::Mul = op.node { if is_float(cx, &lhs) && is_float(cx, &rhs) { return Some((&lhs, &rhs)); } } } None } impl<'a, 'tcx> LateLintPass<'a, 'tcx> for MulAddCheck { fn check_expr(&mut self, cx: &LateContext<'a, 'tcx>, expr: &'tcx Expr) { if let ExprKind::Binary(op, lhs, rhs) = &expr.kind { if let BinOpKind::Add = op.node { //Converts mult_lhs * mult_rhs + rhs to mult_lhs.mult_add(mult_rhs, rhs) if let Some((mult_lhs, mult_rhs)) = is_float_mult_expr(cx, lhs) { if is_float(cx, rhs) { span_lint_and_sugg( cx, MANUAL_MUL_ADD, expr.span, "consider using mul_add() for better numerical precision", "try", format!( "{}.mul_add({}, {})", snippet(cx, mult_lhs.span, "_"), snippet(cx, mult_rhs.span, "_"), snippet(cx, rhs.span, "_"), ), Applicability::MaybeIncorrect, ); } } //Converts lhs + mult_lhs * mult_rhs to mult_lhs.mult_add(mult_rhs, lhs) if let Some((mult_lhs, mult_rhs)) = is_float_mult_expr(cx, rhs) { if is_float(cx, lhs) { span_lint_and_sugg( cx, MANUAL_MUL_ADD, expr.span, "consider using mul_add() for better numerical precision", "try", format!( "{}.mul_add({}, {})", snippet(cx, mult_lhs.span, "_"), snippet(cx, mult_rhs.span, "_"), snippet(cx, lhs.span, "_"), ), Applicability::MaybeIncorrect, ); } } } } } }