refactored misc to reduce passes

related to #1062

Before: 960MB After: 956MB

So while this reduces code size somewhat, I don't see much memory
improvement here. Still, it's probably worthwile to consider reducing
our passes at least within modules.
This commit is contained in:
Andre Bogus 2016-08-24 21:47:46 +02:00
parent ddf84ea845
commit 06235b0310
2 changed files with 208 additions and 303 deletions

View file

@ -176,8 +176,6 @@ pub fn register_plugins(reg: &mut rustc_plugin::Registry) {
reg.register_late_lint_pass(box types::TypePass);
reg.register_late_lint_pass(box booleans::NonminimalBool);
reg.register_early_lint_pass(box module_inception::Pass);
reg.register_late_lint_pass(box misc::TopLevelRefPass);
reg.register_late_lint_pass(box misc::CmpNan);
reg.register_late_lint_pass(box eq_op::EqOp);
reg.register_early_lint_pass(box enum_variants::EnumVariantNames::new(conf.enum_variant_name_threshold));
reg.register_late_lint_pass(box enum_glob_use::EnumGlobUse);
@ -187,7 +185,7 @@ pub fn register_plugins(reg: &mut rustc_plugin::Registry) {
reg.register_late_lint_pass(box needless_bool::NeedlessBool);
reg.register_late_lint_pass(box needless_bool::BoolComparison);
reg.register_late_lint_pass(box approx_const::Pass);
reg.register_late_lint_pass(box misc::FloatCmp);
reg.register_late_lint_pass(box misc::Pass);
reg.register_early_lint_pass(box precedence::Precedence);
reg.register_late_lint_pass(box eta_reduction::EtaPass);
reg.register_late_lint_pass(box identity_op::IdentityOp);
@ -195,11 +193,9 @@ pub fn register_plugins(reg: &mut rustc_plugin::Registry) {
reg.register_late_lint_pass(box mut_mut::MutMut);
reg.register_late_lint_pass(box mut_reference::UnnecessaryMutPassed);
reg.register_late_lint_pass(box len_zero::LenZero);
reg.register_late_lint_pass(box misc::CmpOwned);
reg.register_late_lint_pass(box attrs::AttrPass);
reg.register_early_lint_pass(box collapsible_if::CollapsibleIf);
reg.register_late_lint_pass(box block_in_if_condition::BlockInIfCondition);
reg.register_late_lint_pass(box misc::ModuloOne);
reg.register_late_lint_pass(box unicode::Unicode);
reg.register_late_lint_pass(box strings::StringAdd);
reg.register_early_lint_pass(box returns::ReturnPass);
@ -214,7 +210,6 @@ pub fn register_plugins(reg: &mut rustc_plugin::Registry) {
reg.register_late_lint_pass(box types::CastPass);
reg.register_late_lint_pass(box types::TypeComplexityPass::new(conf.type_complexity_threshold));
reg.register_late_lint_pass(box matches::MatchPass);
reg.register_late_lint_pass(box misc::PatternPass);
reg.register_late_lint_pass(box minmax::MinMaxPass);
reg.register_late_lint_pass(box open_options::NonSensical);
reg.register_late_lint_pass(box zero_div_zero::Pass);
@ -228,7 +223,6 @@ pub fn register_plugins(reg: &mut rustc_plugin::Registry) {
reg.register_late_lint_pass(box cyclomatic_complexity::CyclomaticComplexity::new(conf.cyclomatic_complexity_threshold));
reg.register_late_lint_pass(box escape::Pass{too_large_for_stack: conf.too_large_for_stack});
reg.register_early_lint_pass(box misc_early::MiscEarly);
reg.register_late_lint_pass(box misc::UsedUnderscoreBinding);
reg.register_late_lint_pass(box array_indexing::ArrayIndexing);
reg.register_late_lint_pass(box panic::Pass);
reg.register_late_lint_pass(box strings::StringLitAsBytes);

View file

@ -40,16 +40,134 @@ declare_lint! {
"an entire binding declared as `ref`, in a function argument or a `let` statement"
}
#[allow(missing_copy_implementations)]
pub struct TopLevelRefPass;
/// **What it does:** Checks for comparisons to NaN.
///
/// **Why is this bad?** NaN does not compare meaningfully to anything not
/// even itself so those comparisons are simply wrong.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x == NAN
/// ```
declare_lint! {
pub CMP_NAN,
Deny,
"comparisons to NAN, which will always return false, probably not intended"
}
impl LintPass for TopLevelRefPass {
/// **What it does:** Checks for (in-)equality comparisons on floating-point
/// values (apart from zero), except in functions called `*eq*` (which probably
/// implement equality for a type involving floats).
///
/// **Why is this bad?** Floating point calculations are usually imprecise, so
/// asking if two values are *exactly* equal is asking for trouble. For a good
/// guide on what to do, see [the floating point
/// guide](http://www.floating-point-gui.de/errors/comparison).
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// y == 1.23f64
/// y != x // where both are floats
/// ```
declare_lint! {
pub FLOAT_CMP,
Warn,
"using `==` or `!=` on float values instead of comparing difference with an epsilon"
}
/// **What it does:** Checks for conversions to owned values just for the sake
/// of a comparison.
///
/// **Why is this bad?** The comparison can operate on a reference, so creating
/// an owned value effectively throws it away directly afterwards, which is
/// needlessly consuming code and heap space.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x.to_owned() == y
/// ```
declare_lint! {
pub CMP_OWNED,
Warn,
"creating owned instances for comparing with others, e.g. `x == \"foo\".to_string()`"
}
/// **What it does:** Checks for getting the remainder of a division by one.
///
/// **Why is this bad?** The result can only ever be zero. No one will write
/// such code deliberately, unless trying to win an Underhanded Rust
/// Contest. Even for that contest, it's probably a bad idea. Use something more
/// underhanded.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x % 1
/// ```
declare_lint! {
pub MODULO_ONE,
Warn,
"taking a number modulo 1, which always returns 0"
}
/// **What it does:** Checks for patterns in the form `name @ _`.
///
/// **Why is this bad?** It's almost always more readable to just use direct bindings.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// match v {
/// Some(x) => (),
/// y @ _ => (), // easier written as `y`,
/// }
/// ```
declare_lint! {
pub REDUNDANT_PATTERN,
Warn,
"using `name @ _` in a pattern"
}
/// **What it does:** Checks for the use of bindings with a single leading underscore.
///
/// **Why is this bad?** A single leading underscore is usually used to indicate
/// that a binding will not be used. Using such a binding breaks this
/// expectation.
///
/// **Known problems:** The lint does not work properly with desugaring and
/// macro, it has been allowed in the mean time.
///
/// **Example:**
/// ```rust
/// let _x = 0;
/// let y = _x + 1; // Here we are using `_x`, even though it has a leading underscore.
/// // We should rename `_x` to `x`
/// ```
declare_lint! {
pub USED_UNDERSCORE_BINDING,
Allow,
"using a binding which is prefixed with an underscore"
}
#[derive(Copy, Clone)]
pub struct Pass;
impl LintPass for Pass {
fn get_lints(&self) -> LintArray {
lint_array!(TOPLEVEL_REF_ARG)
lint_array!(TOPLEVEL_REF_ARG, CMP_NAN, FLOAT_CMP, CMP_OWNED, MODULO_ONE, REDUNDANT_PATTERN,
USED_UNDERSCORE_BINDING)
}
}
impl LateLintPass for TopLevelRefPass {
impl LateLintPass for Pass {
fn check_fn(&mut self, cx: &LateContext, k: FnKind, decl: &FnDecl, _: &Block, _: Span, _: NodeId) {
if let FnKind::Closure(_) = k {
// Does not apply to closures
@ -64,6 +182,7 @@ impl LateLintPass for TopLevelRefPass {
}
}
}
fn check_stmt(&mut self, cx: &LateContext, s: &Stmt) {
if_let_chain! {[
let StmtDecl(ref d, _) = s.node,
@ -97,95 +216,20 @@ impl LateLintPass for TopLevelRefPass {
);
}}
}
}
/// **What it does:** Checks for comparisons to NaN.
///
/// **Why is this bad?** NaN does not compare meaningfully to anything not
/// even itself so those comparisons are simply wrong.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x == NAN
/// ```
declare_lint! {
pub CMP_NAN,
Deny,
"comparisons to NAN, which will always return false, probably not intended"
}
#[derive(Copy,Clone)]
pub struct CmpNan;
impl LintPass for CmpNan {
fn get_lints(&self) -> LintArray {
lint_array!(CMP_NAN)
}
}
impl LateLintPass for CmpNan {
fn check_expr(&mut self, cx: &LateContext, expr: &Expr) {
if let ExprBinary(ref cmp, ref left, ref right) = expr.node {
if cmp.node.is_comparison() {
let op = cmp.node;
if op.is_comparison() {
if let ExprPath(_, ref path) = left.node {
check_nan(cx, path, expr.span);
}
if let ExprPath(_, ref path) = right.node {
check_nan(cx, path, expr.span);
}
check_to_owned(cx, left, right, true, cmp.span);
check_to_owned(cx, right, left, false, cmp.span)
}
}
}
}
fn check_nan(cx: &LateContext, path: &Path, span: Span) {
path.segments.last().map(|seg| {
if seg.name.as_str() == "NAN" {
span_lint(cx,
CMP_NAN,
span,
"doomed comparison with NAN, use `std::{f32,f64}::is_nan()` instead");
}
});
}
/// **What it does:** Checks for (in-)equality comparisons on floating-point
/// values (apart from zero), except in functions called `*eq*` (which probably
/// implement equality for a type involving floats).
///
/// **Why is this bad?** Floating point calculations are usually imprecise, so
/// asking if two values are *exactly* equal is asking for trouble. For a good
/// guide on what to do, see [the floating point
/// guide](http://www.floating-point-gui.de/errors/comparison).
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// y == 1.23f64
/// y != x // where both are floats
/// ```
declare_lint! {
pub FLOAT_CMP,
Warn,
"using `==` or `!=` on float values instead of comparing difference with an epsilon"
}
#[derive(Copy,Clone)]
pub struct FloatCmp;
impl LintPass for FloatCmp {
fn get_lints(&self) -> LintArray {
lint_array!(FLOAT_CMP)
}
}
impl LateLintPass for FloatCmp {
fn check_expr(&mut self, cx: &LateContext, expr: &Expr) {
if let ExprBinary(ref cmp, ref left, ref right) = expr.node {
let op = cmp.node;
if (op == BiEq || op == BiNe) && (is_float(cx, left) || is_float(cx, right)) {
if is_allowed(cx, left) || is_allowed(cx, right) {
return;
@ -210,11 +254,76 @@ impl LateLintPass for FloatCmp {
format!("({}).abs() < error", lhs - rhs));
db.span_note(expr.span, "std::f32::EPSILON and std::f64::EPSILON are available.");
});
} else if op == BiRem && is_integer_literal(right, 1) {
span_lint(cx, MODULO_ONE, expr.span, "any number modulo 1 will be 0");
}
}
if in_attributes_expansion(cx, expr) {
// Don't lint things expanded by #[derive(...)], etc
return;
}
let binding = match expr.node {
ExprPath(_, ref path) => {
let binding = path.segments
.last()
.expect("path should always have at least one segment")
.name
.as_str();
if binding.starts_with('_') &&
!binding.starts_with("__") &&
binding != "_result" && // FIXME: #944
is_used(cx, expr) &&
// don't lint if the declaration is in a macro
non_macro_local(cx, &cx.tcx.expect_def(expr.id)) {
Some(binding)
} else {
None
}
}
ExprField(_, spanned) => {
let name = spanned.node.as_str();
if name.starts_with('_') && !name.starts_with("__") {
Some(name)
} else {
None
}
}
_ => None,
};
if let Some(binding) = binding {
span_lint(cx,
USED_UNDERSCORE_BINDING,
expr.span,
&format!("used binding `{}` which is prefixed with an underscore. A leading \
underscore signals that a binding will not be used.", binding));
}
}
fn check_pat(&mut self, cx: &LateContext, pat: &Pat) {
if let PatKind::Binding(_, ref ident, Some(ref right)) = pat.node {
if right.node == PatKind::Wild {
span_lint(cx,
REDUNDANT_PATTERN,
pat.span,
&format!("the `{} @ _` pattern can be written as just `{}`",
ident.node,
ident.node));
}
}
}
}
fn check_nan(cx: &LateContext, path: &Path, span: Span) {
path.segments.last().map(|seg| {
if seg.name.as_str() == "NAN" {
span_lint(cx,
CMP_NAN,
span,
"doomed comparison with NAN, use `std::{f32,f64}::is_nan()` instead");
}
});
}
fn is_allowed(cx: &LateContext, expr: &Expr) -> bool {
let res = eval_const_expr_partial(cx.tcx, expr, ExprTypeChecked, None);
if let Ok(ConstVal::Float(val)) = res {
@ -247,45 +356,6 @@ fn is_float(cx: &LateContext, expr: &Expr) -> bool {
matches!(walk_ptrs_ty(cx.tcx.expr_ty(expr)).sty, ty::TyFloat(_))
}
/// **What it does:** Checks for conversions to owned values just for the sake
/// of a comparison.
///
/// **Why is this bad?** The comparison can operate on a reference, so creating
/// an owned value effectively throws it away directly afterwards, which is
/// needlessly consuming code and heap space.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x.to_owned() == y
/// ```
declare_lint! {
pub CMP_OWNED,
Warn,
"creating owned instances for comparing with others, e.g. `x == \"foo\".to_string()`"
}
#[derive(Copy,Clone)]
pub struct CmpOwned;
impl LintPass for CmpOwned {
fn get_lints(&self) -> LintArray {
lint_array!(CMP_OWNED)
}
}
impl LateLintPass for CmpOwned {
fn check_expr(&mut self, cx: &LateContext, expr: &Expr) {
if let ExprBinary(ref cmp, ref left, ref right) = expr.node {
if cmp.node.is_comparison() {
check_to_owned(cx, left, right, true, cmp.span);
check_to_owned(cx, right, left, false, cmp.span)
}
}
}
}
fn check_to_owned(cx: &LateContext, expr: &Expr, other: &Expr, left: bool, op: Span) {
let (arg_ty, snip) = match expr.node {
ExprMethodCall(Spanned { node: ref name, .. }, _, ref args) if args.len() == 1 => {
@ -346,165 +416,6 @@ fn is_str_arg(cx: &LateContext, args: &[P<Expr>]) -> bool {
matches!(walk_ptrs_ty(cx.tcx.expr_ty(&args[0])).sty, ty::TyStr)
}
/// **What it does:** Checks for getting the remainder of a division by one.
///
/// **Why is this bad?** The result can only ever be zero. No one will write
/// such code deliberately, unless trying to win an Underhanded Rust
/// Contest. Even for that contest, it's probably a bad idea. Use something more
/// underhanded.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// x % 1
/// ```
declare_lint! {
pub MODULO_ONE,
Warn,
"taking a number modulo 1, which always returns 0"
}
#[derive(Copy,Clone)]
pub struct ModuloOne;
impl LintPass for ModuloOne {
fn get_lints(&self) -> LintArray {
lint_array!(MODULO_ONE)
}
}
impl LateLintPass for ModuloOne {
fn check_expr(&mut self, cx: &LateContext, expr: &Expr) {
if let ExprBinary(ref cmp, _, ref right) = expr.node {
if let Spanned { node: BinOp_::BiRem, .. } = *cmp {
if is_integer_literal(right, 1) {
span_lint(cx, MODULO_ONE, expr.span, "any number modulo 1 will be 0");
}
}
}
}
}
/// **What it does:** Checks for patterns in the form `name @ _`.
///
/// **Why is this bad?** It's almost always more readable to just use direct bindings.
///
/// **Known problems:** None.
///
/// **Example:**
/// ```rust
/// match v {
/// Some(x) => (),
/// y @ _ => (), // easier written as `y`,
/// }
/// ```
declare_lint! {
pub REDUNDANT_PATTERN,
Warn,
"using `name @ _` in a pattern"
}
#[derive(Copy,Clone)]
pub struct PatternPass;
impl LintPass for PatternPass {
fn get_lints(&self) -> LintArray {
lint_array!(REDUNDANT_PATTERN)
}
}
impl LateLintPass for PatternPass {
fn check_pat(&mut self, cx: &LateContext, pat: &Pat) {
if let PatKind::Binding(_, ref ident, Some(ref right)) = pat.node {
if right.node == PatKind::Wild {
span_lint(cx,
REDUNDANT_PATTERN,
pat.span,
&format!("the `{} @ _` pattern can be written as just `{}`",
ident.node,
ident.node));
}
}
}
}
/// **What it does:** Checks for the use of bindings with a single leading underscore.
///
/// **Why is this bad?** A single leading underscore is usually used to indicate
/// that a binding will not be used. Using such a binding breaks this
/// expectation.
///
/// **Known problems:** The lint does not work properly with desugaring and
/// macro, it has been allowed in the mean time.
///
/// **Example:**
/// ```rust
/// let _x = 0;
/// let y = _x + 1; // Here we are using `_x`, even though it has a leading underscore.
/// // We should rename `_x` to `x`
/// ```
declare_lint! {
pub USED_UNDERSCORE_BINDING,
Allow,
"using a binding which is prefixed with an underscore"
}
#[derive(Copy, Clone)]
pub struct UsedUnderscoreBinding;
impl LintPass for UsedUnderscoreBinding {
fn get_lints(&self) -> LintArray {
lint_array!(USED_UNDERSCORE_BINDING)
}
}
impl LateLintPass for UsedUnderscoreBinding {
#[cfg_attr(rustfmt, rustfmt_skip)]
fn check_expr(&mut self, cx: &LateContext, expr: &Expr) {
if in_attributes_expansion(cx, expr) {
// Don't lint things expanded by #[derive(...)], etc
return;
}
let binding = match expr.node {
ExprPath(_, ref path) => {
let binding = path.segments
.last()
.expect("path should always have at least one segment")
.name
.as_str();
if binding.starts_with('_') &&
!binding.starts_with("__") &&
binding != "_result" && // FIXME: #944
is_used(cx, expr) &&
// don't lint if the declaration is in a macro
non_macro_local(cx, &cx.tcx.expect_def(expr.id)) {
Some(binding)
} else {
None
}
}
ExprField(_, spanned) => {
let name = spanned.node.as_str();
if name.starts_with('_') && !name.starts_with("__") {
Some(name)
} else {
None
}
}
_ => None,
};
if let Some(binding) = binding {
span_lint(cx,
USED_UNDERSCORE_BINDING,
expr.span,
&format!("used binding `{}` which is prefixed with an underscore. A leading \
underscore signals that a binding will not be used.", binding));
}
}
}
/// Heuristic to see if an expression is used. Should be compatible with `unused_variables`'s idea
/// of what it means for an expression to be "used".
fn is_used(cx: &LateContext, expr: &Expr) -> bool {