CPlusPlus: Refactor FindUsages::getType()

Use a dedicated class instead of a collection of lambdas. We had
difficulties debugging this code, as gdb appears to have problems stepping
into lambdas. The new structure is also easier to read.

Change-Id: Icc88b5b884f1d60458c7c3254c2d13d7ab4592de
Reviewed-by: Christian Stenger <christian.stenger@qt.io>
This commit is contained in:
Christian Kandeler
2021-01-12 16:10:23 +01:00
parent 4e5676e8c2
commit 5afdb17212
2 changed files with 167 additions and 135 deletions

View File

@@ -145,13 +145,136 @@ void FindUsages::reportResult(unsigned tokenIndex, const QList<LookupItem> &cand
_references.append(tokenIndex);
}
Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
class FindUsages::GetUsageType
{
const auto containsToken = [tokenIndex](const AST *ast) {
return ast && ast->firstToken() <= tokenIndex && ast->lastToken() > tokenIndex;
};
const auto isAssignment = [this](int token) {
switch (tokenKind(token)) {
public:
GetUsageType(FindUsages *findUsages, const QList<AST *> &astPath, int tokenIndex)
: m_findUsages(findUsages), m_astPath(astPath), m_tokenIndex(tokenIndex)
{
}
Usage::Type getUsageType() const
{
if (m_astPath.size() < 2 || !m_astPath.last()->asSimpleName())
return Usage::Type::Other;
for (auto it = m_astPath.rbegin() + 1; it != m_astPath.rend(); ++it) {
if ((*it)->asExpressionStatement())
return Usage::Type::Read;
if ((*it)->asSwitchStatement())
return Usage::Type::Read;
if ((*it)->asCaseStatement())
return Usage::Type::Read;
if ((*it)->asIfStatement())
return Usage::Type::Read;
if ((*it)->asLambdaCapture())
return Usage::Type::Other;
if ((*it)->asTypenameTypeParameter())
return Usage::Type::Declaration;
if ((*it)->asNewExpression())
return Usage::Type::Other;
if (ClassSpecifierAST *classSpec = (*it)->asClassSpecifier()) {
if (classSpec->name == *(it - 1))
return Usage::Type::Declaration;
continue;
}
if (const auto memInitAst = (*it)->asMemInitializer()) {
if (memInitAst->name == *(it - 1))
return Usage::Type::Write;
return Usage::Type::Read;
}
if ((*it)->asCall())
return checkPotentialWrite(getUsageTypeForCall(it), it + 1);
if ((*it)->asDeleteExpression())
return Usage::Type::Write;
if (const auto binExpr = (*it)->asBinaryExpression()) {
if (binExpr->left_expression == *(it - 1) && isAssignment(binExpr->binary_op_token))
return checkPotentialWrite(Usage::Type::Write, it + 1);
const Utils::optional<LookupItem> item = getTypeOfExpr(binExpr->left_expression,
it + 1);
if (!item)
return Usage::Type::Other;
return checkPotentialWrite(getUsageTypeFromLhsAndRhs(
item->type(), binExpr->right_expression, it),
it + 1);
}
if (const auto unaryOp = (*it)->asUnaryExpression()) {
switch (m_findUsages->tokenKind(unaryOp->unary_op_token)) {
case T_PLUS_PLUS: case T_MINUS_MINUS:
return checkPotentialWrite(Usage::Type::Write, it + 1);
case T_AMPER: case T_STAR:
continue;
default:
return Usage::Type::Read;
}
}
if (const auto sizeofExpr = (*it)->asSizeofExpression()) {
if (containsToken(sizeofExpr->expression))
return Usage::Type::Read;
return Usage::Type::Other;
}
if (const auto arrayExpr = (*it)->asArrayAccess()) {
if (containsToken(arrayExpr->expression))
return Usage::Type::Read;
continue;
}
if (const auto postIncrDecrOp = (*it)->asPostIncrDecr())
return checkPotentialWrite(Usage::Type::Write, it + 1);
if (const auto declaratorId = (*it)->asDeclaratorId()) {
// We don't want to classify constructors and destructors as declarations
// when listing class usages.
if (m_findUsages->_declSymbol->asClass())
return Usage::Type::Other;
continue;
}
if (const auto declarator = (*it)->asDeclarator()) {
if (containsToken(declarator->core_declarator)) {
if (declarator->initializer && (!declarator->postfix_declarator_list
|| !declarator->postfix_declarator_list->value
|| !declarator->postfix_declarator_list->value->asFunctionDeclarator())) {
return Usage::Type::Initialization;
}
return Usage::Type::Declaration;
}
if (const auto decl = (*(it + 1))->asSimpleDeclaration()) {
if (decl->symbols && decl->symbols->value) {
return checkPotentialWrite(
getUsageTypeFromLhsAndRhs(decl->symbols->value->type(),
declarator->initializer, it + 1),
it + 1);
}
}
return Usage::Type::Other;
}
if (const auto retStmt = (*it)->asReturnStatement()) {
for (auto funcIt = it + 1; funcIt != m_astPath.rend(); ++funcIt) {
if (FunctionDefinitionAST * const funcAst = (*funcIt)->asFunctionDefinition()) {
if (funcAst->symbol) {
return checkPotentialWrite(
getUsageTypeFromLhsAndRhs(funcAst->symbol->type(),
retStmt->expression, funcIt),
funcIt + 1);
}
}
}
return Usage::Type::Other;
}
}
return Usage::Type::Other;
}
private:
using Iterator = QList<AST *>::const_reverse_iterator;
bool containsToken(const AST *ast) const
{
return ast && ast->firstToken() <= m_tokenIndex && ast->lastToken() > m_tokenIndex;
}
bool isAssignment(int token) const
{
switch (m_findUsages->tokenKind(token)) {
case T_AMPER_EQUAL: case T_CARET_EQUAL: case T_SLASH_EQUAL: case T_EQUAL:
case T_MINUS_EQUAL: case T_PERCENT_EQUAL: case T_PIPE_EQUAL: case T_PLUS_EQUAL:
case T_STAR_EQUAL: case T_TILDE_EQUAL:
@@ -159,12 +282,13 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
default:
return false;
}
};
}
// This is called for the type of the LHS of an (initialization) assignment.
// We consider the RHS to be writable through the LHS if the LHS is a pointer
// that is non-const at any element level, or if it is a a non-const reference.
static const auto getUsageTypeFromDataType = [](FullySpecifiedType type) {
static Usage::Type getUsageTypeFromDataType(FullySpecifiedType type)
{
if (type.isAuto())
return Usage::Type::Other;
if (const auto refType = type->asReferenceType())
@@ -175,18 +299,17 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
return Usage::Type::WritableRef;
}
return Usage::Type::Read;
};
const QList<AST *> astPath = ASTPath(_doc)(line, column);
}
// If we found a potential write access inside a lambda, we have to check whether the variable
// was captured by value. If so, it's not really a write access.
// FIXME: The parser does not record whether the capture was by reference.
const auto checkPotentialWrite = [&](Usage::Type usageType, auto startIt) {
Usage::Type checkPotentialWrite(Usage::Type usageType, Iterator startIt) const
{
if (usageType != Usage::Type::Write && usageType != Usage::Type::WritableRef)
return usageType;
for (auto it = startIt; it != astPath.rend(); ++it) {
if ((*it)->firstToken() > tokenIndex)
for (auto it = startIt; it != m_astPath.rend(); ++it) {
if ((*it)->firstToken() > m_tokenIndex)
break;
const auto lambdaExpr = (*it)->asLambdaExpression();
if (!lambdaExpr)
@@ -198,20 +321,22 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
capList = capList->next) {
if (!capList->value || !capList->value->identifier)
continue;
if (!Matcher::match(_declSymbol->name(), capList->value->identifier->name))
if (!Matcher::match(m_findUsages->_declSymbol->name(),
capList->value->identifier->name)) {
continue;
}
return capList->value->amper_token ? usageType : Usage::Type::Read;
}
}
return usageType;
};
}
const auto getTypesOfExpr = [&](ExpressionAST *expr, auto scopeSearchPos)
-> const QList<LookupItem> {
const QList<LookupItem> getTypesOfExpr(ExpressionAST *expr, Iterator scopeSearchPos) const
{
if (!expr)
return {};
Scope *scope = nullptr;
for (auto it = scopeSearchPos; !scope && it != astPath.rend(); ++it) {
for (auto it = scopeSearchPos; !scope && it != m_astPath.rend(); ++it) {
if (const auto stmt = (*it)->asCompoundStatement())
scope = stmt->symbol;
else if (const auto klass = (*it)->asClassSpecifier())
@@ -220,20 +345,21 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
scope = ns->symbol;
}
if (!scope)
scope = _doc->globalNamespace();
return typeofExpression(expr, _doc, scope);
};
scope = m_findUsages->_doc->globalNamespace();
return m_findUsages->typeofExpression(expr, m_findUsages->_doc, scope);
}
const auto getTypeOfExpr = [&](ExpressionAST *expr, auto scopeSearchPos)
-> Utils::optional<LookupItem> {
Utils::optional<LookupItem> getTypeOfExpr(ExpressionAST *expr, Iterator scopeSearchPos) const
{
const QList<LookupItem> items = getTypesOfExpr(expr, scopeSearchPos);
if (items.isEmpty())
return {};
return Utils::optional<LookupItem>(items.first());
};
}
const auto getUsageTypeFromLhsAndRhs
= [&](const FullySpecifiedType &lhsType, ExpressionAST *rhs, auto scopeSearchPos) {
Usage::Type getUsageTypeFromLhsAndRhs(const FullySpecifiedType &lhsType, ExpressionAST *rhs,
Iterator scopeSearchPos) const
{
const Usage::Type usageType = getUsageTypeFromDataType(lhsType);
if (usageType != Usage::Type::Other)
return usageType;
@@ -243,16 +369,17 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
if (!item)
return Usage::Type::Other;
return getUsageTypeFromDataType(item->type());
};
}
const auto getUsageTypeForCall = [&](auto callIt) {
Usage::Type getUsageTypeForCall(Iterator callIt) const
{
CallAST * const call = (*callIt)->asCall();
// Check whether this is a member function call on the symbol we are looking for
// (possibly indirectly via a data member).
// If it is and the function is not const, then this is a potential write.
if (call->base_expression == *(callIt - 1)) {
for (auto it = callIt; it != astPath.rbegin(); --it) {
for (auto it = callIt; it != m_astPath.rbegin(); --it) {
const auto memberAccess = (*it)->asMemberAccess();
if (!memberAccess || !memberAccess->member_name || !memberAccess->member_name->name)
continue;
@@ -268,7 +395,7 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
while (const auto ptrType = baseExprType->asPointerType())
baseExprType = ptrType->elementType();
Class *klass = baseExprType->asClassType();
const LookupContext context(_doc, _snapshot);
const LookupContext context(m_findUsages->_doc, m_findUsages->_snapshot);
QList<LookupItem> items;
if (!klass) {
if (const auto namedType = baseExprType->asNamedType()) {
@@ -328,112 +455,16 @@ Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
}
}
return currentType;
};
if (astPath.size() < 2 || !astPath.last()->asSimpleName())
return Usage::Type::Other;
for (auto it = astPath.rbegin() + 1; it != astPath.rend(); ++it) {
if ((*it)->asExpressionStatement())
return Usage::Type::Read;
if ((*it)->asSwitchStatement())
return Usage::Type::Read;
if ((*it)->asCaseStatement())
return Usage::Type::Read;
if ((*it)->asIfStatement())
return Usage::Type::Read;
if ((*it)->asLambdaCapture())
return Usage::Type::Other;
if ((*it)->asTypenameTypeParameter())
return Usage::Type::Declaration;
if ((*it)->asNewExpression())
return Usage::Type::Other;
if (ClassSpecifierAST *classSpec = (*it)->asClassSpecifier()) {
if (classSpec->name == *(it - 1))
return Usage::Type::Declaration;
continue;
}
if (const auto memInitAst = (*it)->asMemInitializer()) {
if (memInitAst->name == *(it - 1))
return Usage::Type::Write;
return Usage::Type::Read;
}
if ((*it)->asCall())
return checkPotentialWrite(getUsageTypeForCall(it), it + 1);
if ((*it)->asDeleteExpression())
return Usage::Type::Write;
if (const auto binExpr = (*it)->asBinaryExpression()) {
if (binExpr->left_expression == *(it - 1) && isAssignment(binExpr->binary_op_token))
return checkPotentialWrite(Usage::Type::Write, it + 1);
const Utils::optional<LookupItem> item = getTypeOfExpr(binExpr->left_expression, it + 1);
if (!item)
return Usage::Type::Other;
return checkPotentialWrite(getUsageTypeFromLhsAndRhs(
item->type(), binExpr->right_expression, it), it + 1);
}
if (const auto unaryOp = (*it)->asUnaryExpression()) {
switch (tokenKind(unaryOp->unary_op_token)) {
case T_PLUS_PLUS: case T_MINUS_MINUS:
return checkPotentialWrite(Usage::Type::Write, it + 1);
case T_AMPER: case T_STAR:
continue;
default:
return Usage::Type::Read;
}
}
if (const auto sizeofExpr = (*it)->asSizeofExpression()) {
if (containsToken(sizeofExpr->expression))
return Usage::Type::Read;
return Usage::Type::Other;
}
if (const auto arrayExpr = (*it)->asArrayAccess()) {
if (containsToken(arrayExpr->expression))
return Usage::Type::Read;
continue;
}
if (const auto postIncrDecrOp = (*it)->asPostIncrDecr())
return checkPotentialWrite(Usage::Type::Write, it + 1);
if (const auto declaratorId = (*it)->asDeclaratorId()) {
// We don't want to classify constructors and destructors as declarations
// when listing class usages.
if (_declSymbol->asClass())
return Usage::Type::Other;
continue;
}
if (const auto declarator = (*it)->asDeclarator()) {
if (containsToken(declarator->core_declarator)) {
if (declarator->initializer && (!declarator->postfix_declarator_list
|| !declarator->postfix_declarator_list->value
|| !declarator->postfix_declarator_list->value->asFunctionDeclarator())) {
return Usage::Type::Initialization;
}
return Usage::Type::Declaration;
}
if (const auto decl = (*(it + 1))->asSimpleDeclaration()) {
if (decl->symbols && decl->symbols->value) {
return checkPotentialWrite(
getUsageTypeFromLhsAndRhs(decl->symbols->value->type(),
declarator->initializer, it + 1), it + 1);
}
}
return Usage::Type::Other;
}
if (const auto retStmt = (*it)->asReturnStatement()) {
for (auto funcIt = it + 1; funcIt != astPath.rend(); ++funcIt) {
if (FunctionDefinitionAST * const funcAst = (*funcIt)->asFunctionDefinition()) {
if (funcAst->symbol) {
return checkPotentialWrite(
getUsageTypeFromLhsAndRhs(funcAst->symbol->type(),
retStmt->expression, funcIt),
funcIt + 1);
}
}
}
return Usage::Type::Other;
}
}
return Usage::Type::Other;
FindUsages * const m_findUsages;
const QList<AST *> &m_astPath;
const int m_tokenIndex;
};
Usage::Type FindUsages::getType(int line, int column, int tokenIndex)
{
return GetUsageType(this, ASTPath(_doc)(line, column), tokenIndex).getUsageType();
}
QString FindUsages::matchingLine(const Token &tk) const