
Implement the UNSIGNED extension type and operations under control of a language feature flag (-funsigned). This is nearly identical to the UNSIGNED feature that has been available in Sun Fortran for years, and now implemented in GNU Fortran for gfortran 15, and proposed for ISO standardization in J3/24-116.txt. See the new documentation for details; but in short, this is C's unsigned type, with guaranteed modular arithmetic for +, -, and *, and the related transformational intrinsic functions SUM & al.
278 lines
9.8 KiB
C++
278 lines
9.8 KiB
C++
//===-- lib/Semantics/check-case.cpp --------------------------------------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "check-case.h"
|
|
#include "flang/Common/idioms.h"
|
|
#include "flang/Common/reference.h"
|
|
#include "flang/Common/template.h"
|
|
#include "flang/Evaluate/fold.h"
|
|
#include "flang/Evaluate/type.h"
|
|
#include "flang/Parser/parse-tree.h"
|
|
#include "flang/Semantics/semantics.h"
|
|
#include "flang/Semantics/tools.h"
|
|
#include <tuple>
|
|
|
|
namespace Fortran::semantics {
|
|
|
|
template <typename T> class CaseValues {
|
|
public:
|
|
CaseValues(SemanticsContext &c, const evaluate::DynamicType &t)
|
|
: context_{c}, caseExprType_{t} {}
|
|
|
|
void Check(const std::list<parser::CaseConstruct::Case> &cases) {
|
|
for (const parser::CaseConstruct::Case &c : cases) {
|
|
AddCase(c);
|
|
}
|
|
if (!hasErrors_) {
|
|
cases_.sort(Comparator{});
|
|
if (!AreCasesDisjoint()) { // C1149
|
|
ReportConflictingCases();
|
|
}
|
|
}
|
|
}
|
|
|
|
private:
|
|
using Value = evaluate::Scalar<T>;
|
|
|
|
void AddCase(const parser::CaseConstruct::Case &c) {
|
|
const auto &stmt{std::get<parser::Statement<parser::CaseStmt>>(c.t)};
|
|
const parser::CaseStmt &caseStmt{stmt.statement};
|
|
const auto &selector{std::get<parser::CaseSelector>(caseStmt.t)};
|
|
common::visit(
|
|
common::visitors{
|
|
[&](const std::list<parser::CaseValueRange> &ranges) {
|
|
for (const auto &range : ranges) {
|
|
auto pair{ComputeBounds(range)};
|
|
if (pair.first && pair.second && *pair.first > *pair.second) {
|
|
context_.Warn(common::UsageWarning::EmptyCase, stmt.source,
|
|
"CASE has lower bound greater than upper bound"_warn_en_US);
|
|
} else {
|
|
if constexpr (T::category == TypeCategory::Logical) { // C1148
|
|
if ((pair.first || pair.second) &&
|
|
(!pair.first || !pair.second ||
|
|
*pair.first != *pair.second)) {
|
|
context_.Say(stmt.source,
|
|
"CASE range is not allowed for LOGICAL"_err_en_US);
|
|
}
|
|
}
|
|
cases_.emplace_back(stmt);
|
|
cases_.back().lower = std::move(pair.first);
|
|
cases_.back().upper = std::move(pair.second);
|
|
}
|
|
}
|
|
},
|
|
[&](const parser::Default &) { cases_.emplace_front(stmt); },
|
|
},
|
|
selector.u);
|
|
}
|
|
|
|
std::optional<Value> GetValue(const parser::CaseValue &caseValue) {
|
|
const parser::Expr &expr{caseValue.thing.thing.value()};
|
|
auto *x{expr.typedExpr.get()};
|
|
if (x && x->v) { // C1147
|
|
auto type{x->v->GetType()};
|
|
if (type && type->category() == caseExprType_.category() &&
|
|
(type->category() != TypeCategory::Character ||
|
|
type->kind() == caseExprType_.kind())) {
|
|
parser::Messages buffer; // discarded folding messages
|
|
parser::ContextualMessages foldingMessages{expr.source, &buffer};
|
|
evaluate::FoldingContext foldingContext{
|
|
context_.foldingContext(), foldingMessages};
|
|
auto folded{evaluate::Fold(foldingContext, SomeExpr{*x->v})};
|
|
if (auto converted{evaluate::Fold(foldingContext,
|
|
evaluate::ConvertToType(T::GetType(), SomeExpr{folded}))}) {
|
|
if (auto value{evaluate::GetScalarConstantValue<T>(*converted)}) {
|
|
auto back{evaluate::Fold(foldingContext,
|
|
evaluate::ConvertToType(*type, SomeExpr{*converted}))};
|
|
if (back == folded) {
|
|
x->v = converted;
|
|
return value;
|
|
} else {
|
|
context_.Warn(common::UsageWarning::CaseOverflow, expr.source,
|
|
"CASE value (%s) overflows type (%s) of SELECT CASE expression"_warn_en_US,
|
|
folded.AsFortran(), caseExprType_.AsFortran());
|
|
hasErrors_ = true;
|
|
return std::nullopt;
|
|
}
|
|
}
|
|
}
|
|
context_.Say(expr.source,
|
|
"CASE value (%s) must be a constant scalar"_err_en_US,
|
|
x->v->AsFortran());
|
|
} else {
|
|
std::string typeStr{type ? type->AsFortran() : "typeless"s};
|
|
context_.Say(expr.source,
|
|
"CASE value has type '%s' which is not compatible with the SELECT CASE expression's type '%s'"_err_en_US,
|
|
typeStr, caseExprType_.AsFortran());
|
|
}
|
|
hasErrors_ = true;
|
|
}
|
|
return std::nullopt;
|
|
}
|
|
|
|
using PairOfValues = std::pair<std::optional<Value>, std::optional<Value>>;
|
|
PairOfValues ComputeBounds(const parser::CaseValueRange &range) {
|
|
return common::visit(
|
|
common::visitors{
|
|
[&](const parser::CaseValue &x) {
|
|
auto value{GetValue(x)};
|
|
return PairOfValues{value, value};
|
|
},
|
|
[&](const parser::CaseValueRange::Range &x) {
|
|
std::optional<Value> lo, hi;
|
|
if (x.lower) {
|
|
lo = GetValue(*x.lower);
|
|
}
|
|
if (x.upper) {
|
|
hi = GetValue(*x.upper);
|
|
}
|
|
if ((x.lower && !lo) || (x.upper && !hi)) {
|
|
return PairOfValues{}; // error case
|
|
}
|
|
return PairOfValues{std::move(lo), std::move(hi)};
|
|
},
|
|
},
|
|
range.u);
|
|
}
|
|
|
|
struct Case {
|
|
explicit Case(const parser::Statement<parser::CaseStmt> &s) : stmt{s} {}
|
|
bool IsDefault() const { return !lower && !upper; }
|
|
std::string AsFortran() const {
|
|
std::string result;
|
|
{
|
|
llvm::raw_string_ostream bs{result};
|
|
if (lower) {
|
|
evaluate::Constant<T>{*lower}.AsFortran(bs << '(');
|
|
if (!upper) {
|
|
bs << ':';
|
|
} else if (*lower != *upper) {
|
|
evaluate::Constant<T>{*upper}.AsFortran(bs << ':');
|
|
}
|
|
bs << ')';
|
|
} else if (upper) {
|
|
evaluate::Constant<T>{*upper}.AsFortran(bs << "(:") << ')';
|
|
} else {
|
|
bs << "DEFAULT";
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
|
|
const parser::Statement<parser::CaseStmt> &stmt;
|
|
std::optional<Value> lower, upper;
|
|
};
|
|
|
|
// Defines a comparator for use with std::list<>::sort().
|
|
// Returns true if and only if the highest value in range x is less
|
|
// than the least value in range y. The DEFAULT case is arbitrarily
|
|
// defined to be less than all others. When two ranges overlap,
|
|
// neither is less than the other.
|
|
struct Comparator {
|
|
bool operator()(const Case &x, const Case &y) const {
|
|
if (x.IsDefault()) {
|
|
return !y.IsDefault();
|
|
} else {
|
|
return x.upper && y.lower && *x.upper < *y.lower;
|
|
}
|
|
}
|
|
};
|
|
|
|
bool AreCasesDisjoint() const {
|
|
auto endIter{cases_.end()};
|
|
for (auto iter{cases_.begin()}; iter != endIter; ++iter) {
|
|
auto next{iter};
|
|
if (++next != endIter && !Comparator{}(*iter, *next)) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
// This has quadratic time, but only runs in error cases
|
|
void ReportConflictingCases() {
|
|
for (auto iter{cases_.begin()}; iter != cases_.end(); ++iter) {
|
|
parser::Message *msg{nullptr};
|
|
for (auto p{cases_.begin()}; p != cases_.end(); ++p) {
|
|
if (p->stmt.source.begin() < iter->stmt.source.begin() &&
|
|
!Comparator{}(*p, *iter) && !Comparator{}(*iter, *p)) {
|
|
if (!msg) {
|
|
msg = &context_.Say(iter->stmt.source,
|
|
"CASE %s conflicts with previous cases"_err_en_US,
|
|
iter->AsFortran());
|
|
}
|
|
msg->Attach(
|
|
p->stmt.source, "Conflicting CASE %s"_en_US, p->AsFortran());
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
SemanticsContext &context_;
|
|
const evaluate::DynamicType &caseExprType_;
|
|
std::list<Case> cases_;
|
|
bool hasErrors_{false};
|
|
};
|
|
|
|
template <TypeCategory CAT> struct TypeVisitor {
|
|
using Result = bool;
|
|
using Types = evaluate::CategoryTypes<CAT>;
|
|
template <typename T> Result Test() {
|
|
if (T::kind == exprType.kind()) {
|
|
CaseValues<T>(context, exprType).Check(caseList);
|
|
return true;
|
|
} else {
|
|
return false;
|
|
}
|
|
}
|
|
SemanticsContext &context;
|
|
const evaluate::DynamicType &exprType;
|
|
const std::list<parser::CaseConstruct::Case> &caseList;
|
|
};
|
|
|
|
void CaseChecker::Enter(const parser::CaseConstruct &construct) {
|
|
const auto &selectCaseStmt{
|
|
std::get<parser::Statement<parser::SelectCaseStmt>>(construct.t)};
|
|
const auto &selectCase{selectCaseStmt.statement};
|
|
const auto &selectExpr{
|
|
std::get<parser::Scalar<parser::Expr>>(selectCase.t).thing};
|
|
const auto *x{GetExpr(context_, selectExpr)};
|
|
if (!x) {
|
|
return; // expression semantics failed
|
|
}
|
|
if (auto exprType{x->GetType()}) {
|
|
const auto &caseList{
|
|
std::get<std::list<parser::CaseConstruct::Case>>(construct.t)};
|
|
switch (exprType->category()) {
|
|
case TypeCategory::Integer:
|
|
common::SearchTypes(
|
|
TypeVisitor<TypeCategory::Integer>{context_, *exprType, caseList});
|
|
return;
|
|
case TypeCategory::Unsigned:
|
|
common::SearchTypes(
|
|
TypeVisitor<TypeCategory::Unsigned>{context_, *exprType, caseList});
|
|
return;
|
|
case TypeCategory::Logical:
|
|
CaseValues<evaluate::Type<TypeCategory::Logical, 1>>{context_, *exprType}
|
|
.Check(caseList);
|
|
return;
|
|
case TypeCategory::Character:
|
|
common::SearchTypes(
|
|
TypeVisitor<TypeCategory::Character>{context_, *exprType, caseList});
|
|
return;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
context_.Say(selectExpr.source,
|
|
context_.IsEnabled(common::LanguageFeature::Unsigned)
|
|
? "SELECT CASE expression must be integer, unsigned, logical, or character"_err_en_US
|
|
: "SELECT CASE expression must be integer, logical, or character"_err_en_US);
|
|
}
|
|
} // namespace Fortran::semantics
|