#include "query.h" #include "indexer.h" #include #include #include #include #include #include #include #include #include #include // TODO: Make all copy constructors explicit. namespace { optional ToQuery(const IdMap& id_map, const IndexType::Def& type) { if (type.detailed_name.empty()) return nullopt; QueryType::DefUpdate result(type.usr); result.short_name = type.short_name; result.detailed_name = type.detailed_name; result.definition_spelling = id_map.ToQuery(type.definition_spelling); result.definition_extent = id_map.ToQuery(type.definition_extent); result.alias_of = id_map.ToQuery(type.alias_of); result.parents = id_map.ToQuery(type.parents); result.types = id_map.ToQuery(type.types); result.funcs = id_map.ToQuery(type.funcs); result.vars = id_map.ToQuery(type.vars); return result; } optional ToQuery(const IdMap& id_map, const IndexFunc::Def& func) { if (func.detailed_name.empty()) return nullopt; QueryFunc::DefUpdate result(func.usr); result.short_name = func.short_name; result.detailed_name = func.detailed_name; result.definition_spelling = id_map.ToQuery(func.definition_spelling); result.definition_extent = id_map.ToQuery(func.definition_extent); result.declaring_type = id_map.ToQuery(func.declaring_type); result.base = id_map.ToQuery(func.base); result.locals = id_map.ToQuery(func.locals); result.callees = id_map.ToQuery(func.callees); return result; } optional ToQuery(const IdMap& id_map, const IndexVar::Def& var) { if (var.detailed_name.empty()) return nullopt; QueryVar::DefUpdate result(var.usr); result.short_name = var.short_name; result.detailed_name = var.detailed_name; result.declaration = id_map.ToQuery(var.declaration); result.definition_spelling = id_map.ToQuery(var.definition_spelling); result.definition_extent = id_map.ToQuery(var.definition_extent); result.variable_type = id_map.ToQuery(var.variable_type); result.declaring_type = id_map.ToQuery(var.declaring_type); result.is_local = var.is_local; result.is_macro = var.is_macro; return result; } // Adds the mergeable updates in |source| to |dest|. If a mergeable update for // the destination type already exists, it will be combined. This makes merging // updates take longer but reduces import time on the querydb thread. template void AddMergeableRange( std::vector>* dest, const std::vector>& source) { // TODO: Consider caching the lookup table. It can probably save even more // time at the cost of some additional memory. // Build lookup table. spp::sparse_hash_map id_to_index; id_to_index.resize(dest->size()); for (size_t i = 0; i < dest->size(); ++i) id_to_index[(*dest)[i].id] = i; // Add entries. Try to add them to an existing entry. for (const auto& entry : source) { auto it = id_to_index.find(entry.id); if (it != id_to_index.end()) { AddRange(&(*dest)[it->second].to_add, entry.to_add); AddRange(&(*dest)[it->second].to_remove, entry.to_remove); } else { dest->push_back(entry); } } } // Compares |previous| and |current|, adding all elements that are // in |previous| but not |current| to |removed|, and all elements // that are in |current| but not |previous| to |added|. // // Returns true iff |removed| or |added| are non-empty. template bool ComputeDifferenceForUpdate(std::vector& previous, std::vector& current, std::vector* removed, std::vector* added) { // We need to sort to use std::set_difference. std::sort(previous.begin(), previous.end()); std::sort(current.begin(), current.end()); // Returns the elements in |previous| that are not in |current|. std::set_difference(previous.begin(), previous.end(), current.begin(), current.end(), std::back_inserter(*removed)); // Returns the elements in |current| that are not in |previous|. std::set_difference(current.begin(), current.end(), previous.begin(), previous.end(), std::back_inserter(*added)); return !removed->empty() || !added->empty(); } template void CompareGroups(std::vector& previous_data, std::vector& current_data, std::function on_removed, std::function on_added, std::function on_found) { std::sort(previous_data.begin(), previous_data.end()); std::sort(current_data.begin(), current_data.end()); auto prev_it = previous_data.begin(); auto curr_it = current_data.begin(); while (prev_it != previous_data.end() && curr_it != current_data.end()) { // same id if (prev_it->def.usr == curr_it->def.usr) { on_found(&*prev_it, &*curr_it); ++prev_it; ++curr_it; } // prev_id is smaller - prev_it has data curr_it does not have. else if (prev_it->def.usr < curr_it->def.usr) { on_removed(&*prev_it); ++prev_it; } // prev_id is bigger - curr_it has data prev_it does not have. else { on_added(&*curr_it); ++curr_it; } } // if prev_it still has data, that means it is not in curr_it and was removed. while (prev_it != previous_data.end()) { on_removed(&*prev_it); ++prev_it; } // if curr_it still has data, that means it is not in prev_it and was added. while (curr_it != current_data.end()) { on_added(&*curr_it); ++curr_it; } } QueryFile::Def BuildFileDef(const IdMap& id_map, const IndexFile& indexed) { QueryFile::Def def; def.path = indexed.path; def.includes = indexed.includes; def.inactive_regions = indexed.skipped_by_preprocessor; // Convert enum to markdown compatible strings def.language = [indexed] () { switch (indexed.language) { case LanguageId::C: return "c"; case LanguageId::Cpp: return "cpp"; case LanguageId::ObjC: return "objectivec"; default: return ""; } } (); auto add_outline = [&def, &id_map](SymbolIdx idx, Range range) { def.outline.push_back(SymbolRef(idx, id_map.ToQuery(range))); }; auto add_all_symbols = [&def, &id_map](SymbolIdx idx, Range range) { def.all_symbols.push_back(SymbolRef(idx, id_map.ToQuery(range))); }; for (const IndexType& type : indexed.types) { if (type.def.definition_spelling.has_value()) add_all_symbols(id_map.ToSymbol(type.id), type.def.definition_spelling.value()); if (type.def.definition_extent.has_value()) add_outline(id_map.ToSymbol(type.id), type.def.definition_extent.value()); for (const Range& use : type.uses) add_all_symbols(id_map.ToSymbol(type.id), use); } for (const IndexFunc& func : indexed.funcs) { if (func.def.definition_spelling.has_value()) add_all_symbols(id_map.ToSymbol(func.id), func.def.definition_spelling.value()); if (func.def.definition_extent.has_value()) add_outline(id_map.ToSymbol(func.id), func.def.definition_extent.value()); for (const IndexFunc::Declaration& decl : func.declarations) { // TODO: add more outline info? add_all_symbols(id_map.ToSymbol(func.id), decl.spelling); add_outline(id_map.ToSymbol(func.id), decl.spelling); } for (const IndexFuncRef& caller : func.callers) { if (caller.is_implicit) continue; add_all_symbols(id_map.ToSymbol(func.id), caller.loc); } } for (const IndexVar& var : indexed.vars) { if (var.def.definition_spelling.has_value()) add_all_symbols(id_map.ToSymbol(var.id), var.def.definition_spelling.value()); if (var.def.definition_extent.has_value()) add_outline(id_map.ToSymbol(var.id), var.def.definition_extent.value()); for (const Range& use : var.uses) add_all_symbols(id_map.ToSymbol(var.id), use); } std::sort(def.outline.begin(), def.outline.end(), [](const SymbolRef& a, const SymbolRef& b) { return a.loc.range.start < b.loc.range.start; }); std::sort(def.all_symbols.begin(), def.all_symbols.end(), [](const SymbolRef& a, const SymbolRef& b) { return a.loc.range.start < b.loc.range.start; }); return def; } } // namespace QueryFileId GetQueryFileIdFromPath(QueryDatabase* query_db, const std::string& path) { auto it = query_db->usr_to_file.find(LowerPathIfCaseInsensitive(path)); if (it != query_db->usr_to_file.end()) return QueryFileId(it->second.id); size_t idx = query_db->files.size(); query_db->usr_to_file[LowerPathIfCaseInsensitive(path)] = QueryFileId(idx); query_db->files.push_back(QueryFile(path)); return QueryFileId(idx); } QueryTypeId GetQueryTypeIdFromUsr(QueryDatabase* query_db, const Usr& usr) { auto it = query_db->usr_to_type.find(usr); if (it != query_db->usr_to_type.end()) return QueryTypeId(it->second.id); size_t idx = query_db->types.size(); query_db->usr_to_type[usr] = QueryTypeId(idx); query_db->types.push_back(QueryType(usr)); return QueryTypeId(idx); } QueryFuncId GetQueryFuncIdFromUsr(QueryDatabase* query_db, const Usr& usr) { auto it = query_db->usr_to_func.find(usr); if (it != query_db->usr_to_func.end()) return QueryFuncId(it->second.id); size_t idx = query_db->funcs.size(); query_db->usr_to_func[usr] = QueryFuncId(idx); query_db->funcs.push_back(QueryFunc(usr)); return QueryFuncId(idx); } QueryVarId GetQueryVarIdFromUsr(QueryDatabase* query_db, const Usr& usr) { auto it = query_db->usr_to_var.find(usr); if (it != query_db->usr_to_var.end()) return QueryVarId(it->second.id); size_t idx = query_db->vars.size(); query_db->usr_to_var[usr] = QueryVarId(idx); query_db->vars.push_back(QueryVar(usr)); return QueryVarId(idx); } IdMap::IdMap(QueryDatabase* query_db, const IdCache& local_ids) : local_ids(local_ids) { // LOG_S(INFO) << "Creating IdMap for " << local_ids.primary_file; primary_file = GetQueryFileIdFromPath(query_db, local_ids.primary_file); cached_type_ids_.resize(local_ids.type_id_to_usr.size()); for (const auto& entry : local_ids.type_id_to_usr) cached_type_ids_[entry.first] = GetQueryTypeIdFromUsr(query_db, entry.second); cached_func_ids_.resize(local_ids.func_id_to_usr.size()); for (const auto& entry : local_ids.func_id_to_usr) cached_func_ids_[entry.first] = GetQueryFuncIdFromUsr(query_db, entry.second); cached_var_ids_.resize(local_ids.var_id_to_usr.size()); for (const auto& entry : local_ids.var_id_to_usr) cached_var_ids_[entry.first] = GetQueryVarIdFromUsr(query_db, entry.second); } QueryLocation IdMap::ToQuery(Range range) const { return QueryLocation(primary_file, range); } QueryTypeId IdMap::ToQuery(IndexTypeId id) const { assert(cached_type_ids_.find(id) != cached_type_ids_.end()); return QueryTypeId(cached_type_ids_.find(id)->second); } QueryFuncId IdMap::ToQuery(IndexFuncId id) const { if (id.id == -1) return QueryFuncId((size_t)-1); assert(cached_func_ids_.find(id) != cached_func_ids_.end()); return QueryFuncId(cached_func_ids_.find(id)->second); } QueryVarId IdMap::ToQuery(IndexVarId id) const { assert(cached_var_ids_.find(id) != cached_var_ids_.end()); return QueryVarId(cached_var_ids_.find(id)->second); } QueryFuncRef IdMap::ToQuery(IndexFuncRef ref) const { return QueryFuncRef(ToQuery(ref.id), ToQuery(ref.loc), ref.is_implicit); } QueryLocation IdMap::ToQuery(IndexFunc::Declaration decl) const { // TODO: expose more than just QueryLocation. return QueryLocation(primary_file, decl.spelling); } optional IdMap::ToQuery(optional range) const { if (!range) return nullopt; return ToQuery(range.value()); } optional IdMap::ToQuery(optional id) const { if (!id) return nullopt; return ToQuery(id.value()); } optional IdMap::ToQuery(optional id) const { if (!id) return nullopt; return ToQuery(id.value()); } optional IdMap::ToQuery(optional id) const { if (!id) return nullopt; return ToQuery(id.value()); } optional IdMap::ToQuery(optional ref) const { if (!ref) return nullopt; return ToQuery(ref.value()); } optional IdMap::ToQuery( optional decl) const { if (!decl) return nullopt; return ToQuery(decl.value()); } template std::vector ToQueryTransform(const IdMap& id_map, const std::vector& input) { std::vector result; result.reserve(input.size()); for (const In& in : input) result.push_back(id_map.ToQuery(in)); return result; } std::vector IdMap::ToQuery(std::vector ranges) const { return ToQueryTransform(*this, ranges); } std::vector IdMap::ToQuery(std::vector ids) const { return ToQueryTransform(*this, ids); } std::vector IdMap::ToQuery(std::vector ids) const { return ToQueryTransform(*this, ids); } std::vector IdMap::ToQuery(std::vector ids) const { return ToQueryTransform(*this, ids); } std::vector IdMap::ToQuery(std::vector refs) const { return ToQueryTransform(*this, refs); } std::vector IdMap::ToQuery( std::vector decls) const { return ToQueryTransform(*this, decls); } SymbolIdx IdMap::ToSymbol(IndexTypeId id) const { return SymbolIdx(SymbolKind::Type, ToQuery(id).id); } SymbolIdx IdMap::ToSymbol(IndexFuncId id) const { return SymbolIdx(SymbolKind::Func, ToQuery(id).id); } SymbolIdx IdMap::ToSymbol(IndexVarId id) const { return SymbolIdx(SymbolKind::Var, ToQuery(id).id); } // ---------------------- // INDEX THREAD FUNCTIONS // ---------------------- // static IndexUpdate IndexUpdate::CreateDelta(const IdMap* previous_id_map, const IdMap* current_id_map, IndexFile* previous, IndexFile* current) { // This function runs on an indexer thread. if (!previous_id_map) { assert(!previous); IndexFile empty(current->path); return IndexUpdate(*current_id_map, *current_id_map, empty, *current); } return IndexUpdate(*previous_id_map, *current_id_map, *previous, *current); } IndexUpdate::IndexUpdate(const IdMap& previous_id_map, const IdMap& current_id_map, IndexFile& previous_file, IndexFile& current_file) { // This function runs on an indexer thread. // |query_name| is the name of the variable on the query type. // |index_name| is the name of the variable on the index type. // |type| is the type of the variable. #define PROCESS_UPDATE_DIFF(type_id, query_name, index_name, type) \ { \ /* Check for changes. */ \ std::vector removed, added; \ auto previous = previous_id_map.ToQuery(previous_def->index_name); \ auto current = current_id_map.ToQuery(current_def->index_name); \ bool did_add = \ ComputeDifferenceForUpdate(previous, current, &removed, &added); \ if (did_add) { \ query_name.push_back(MergeableUpdate( \ current_id_map.ToQuery(current_def->id), added, removed)); \ } \ } // File files_def_update.push_back(BuildFileDef(current_id_map, current_file)); // **NOTE** We only remove entries if they were defined in the previous index. // For example, if a type is included from another file it will be defined // simply so we can attribute the usage/reference to it. If the reference goes // away we don't want to remove the type/func/var usage. // Types CompareGroups( previous_file.types, current_file.types, /*onRemoved:*/ [this, &previous_id_map](IndexType* type) { if (type->def.definition_spelling) types_removed.push_back(type->def.usr); else { if (!type->derived.empty()) types_derived.push_back(QueryType::DerivedUpdate( previous_id_map.ToQuery(type->id), {}, previous_id_map.ToQuery(type->derived))); if (!type->instances.empty()) types_instances.push_back(QueryType::InstancesUpdate( previous_id_map.ToQuery(type->id), {}, previous_id_map.ToQuery(type->instances))); if (!type->uses.empty()) types_uses.push_back( QueryType::UsesUpdate(previous_id_map.ToQuery(type->id), {}, previous_id_map.ToQuery(type->uses))); } }, /*onAdded:*/ [this, ¤t_id_map](IndexType* type) { optional def_update = ToQuery(current_id_map, type->def); if (def_update) types_def_update.push_back(*def_update); if (!type->derived.empty()) types_derived.push_back( QueryType::DerivedUpdate(current_id_map.ToQuery(type->id), current_id_map.ToQuery(type->derived))); if (!type->instances.empty()) types_instances.push_back(QueryType::InstancesUpdate( current_id_map.ToQuery(type->id), current_id_map.ToQuery(type->instances))); if (!type->uses.empty()) types_uses.push_back( QueryType::UsesUpdate(current_id_map.ToQuery(type->id), current_id_map.ToQuery(type->uses))); }, /*onFound:*/ [this, &previous_id_map, ¤t_id_map](IndexType* previous_def, IndexType* current_def) { optional previous_remapped_def = ToQuery(previous_id_map, previous_def->def); optional current_remapped_def = ToQuery(current_id_map, current_def->def); if (current_remapped_def && previous_remapped_def != current_remapped_def && !current_remapped_def->detailed_name.empty()) types_def_update.push_back(*current_remapped_def); PROCESS_UPDATE_DIFF(QueryTypeId, types_derived, derived, QueryTypeId); PROCESS_UPDATE_DIFF(QueryTypeId, types_instances, instances, QueryVarId); PROCESS_UPDATE_DIFF(QueryTypeId, types_uses, uses, QueryLocation); }); // Functions CompareGroups( previous_file.funcs, current_file.funcs, /*onRemoved:*/ [this, &previous_id_map](IndexFunc* func) { if (func->def.definition_spelling) { funcs_removed.push_back(func->def.usr); } else { if (!func->declarations.empty()) funcs_declarations.push_back(QueryFunc::DeclarationsUpdate( previous_id_map.ToQuery(func->id), {}, previous_id_map.ToQuery(func->declarations))); if (!func->derived.empty()) funcs_derived.push_back(QueryFunc::DerivedUpdate( previous_id_map.ToQuery(func->id), {}, previous_id_map.ToQuery(func->derived))); if (!func->callers.empty()) funcs_callers.push_back(QueryFunc::CallersUpdate( previous_id_map.ToQuery(func->id), {}, previous_id_map.ToQuery(func->callers))); } }, /*onAdded:*/ [this, ¤t_id_map](IndexFunc* func) { optional def_update = ToQuery(current_id_map, func->def); if (def_update) funcs_def_update.push_back(*def_update); if (!func->declarations.empty()) funcs_declarations.push_back(QueryFunc::DeclarationsUpdate( current_id_map.ToQuery(func->id), current_id_map.ToQuery(func->declarations))); if (!func->derived.empty()) funcs_derived.push_back( QueryFunc::DerivedUpdate(current_id_map.ToQuery(func->id), current_id_map.ToQuery(func->derived))); if (!func->callers.empty()) funcs_callers.push_back( QueryFunc::CallersUpdate(current_id_map.ToQuery(func->id), current_id_map.ToQuery(func->callers))); }, /*onFound:*/ [this, &previous_id_map, ¤t_id_map](IndexFunc* previous_def, IndexFunc* current_def) { optional previous_remapped_def = ToQuery(previous_id_map, previous_def->def); optional current_remapped_def = ToQuery(current_id_map, current_def->def); if (current_remapped_def && previous_remapped_def != current_remapped_def && !current_remapped_def->detailed_name.empty()) funcs_def_update.push_back(*current_remapped_def); PROCESS_UPDATE_DIFF(QueryFuncId, funcs_declarations, declarations, QueryLocation); PROCESS_UPDATE_DIFF(QueryFuncId, funcs_derived, derived, QueryFuncId); PROCESS_UPDATE_DIFF(QueryFuncId, funcs_callers, callers, QueryFuncRef); }); // Variables CompareGroups( previous_file.vars, current_file.vars, /*onRemoved:*/ [this, &previous_id_map](IndexVar* var) { if (var->def.definition_spelling) { vars_removed.push_back(var->def.usr); } else { if (!var->uses.empty()) vars_uses.push_back( QueryVar::UsesUpdate(previous_id_map.ToQuery(var->id), {}, previous_id_map.ToQuery(var->uses))); } }, /*onAdded:*/ [this, ¤t_id_map](IndexVar* var) { optional def_update = ToQuery(current_id_map, var->def); if (def_update) vars_def_update.push_back(*def_update); if (!var->uses.empty()) vars_uses.push_back( QueryVar::UsesUpdate(current_id_map.ToQuery(var->id), current_id_map.ToQuery(var->uses))); }, /*onFound:*/ [this, &previous_id_map, ¤t_id_map](IndexVar* previous_def, IndexVar* current_def) { optional previous_remapped_def = ToQuery(previous_id_map, previous_def->def); optional current_remapped_def = ToQuery(current_id_map, current_def->def); if (current_remapped_def && previous_remapped_def != current_remapped_def && !current_remapped_def->detailed_name.empty()) vars_def_update.push_back(*current_remapped_def); PROCESS_UPDATE_DIFF(QueryVarId, vars_uses, uses, QueryLocation); }); #undef PROCESS_UPDATE_DIFF } void IndexUpdate::Merge(const IndexUpdate& update) { // This function runs on an indexer thread. #define INDEX_UPDATE_APPEND(name) AddRange(&name, update.name); #define INDEX_UPDATE_MERGE(name) AddMergeableRange(&name, update.name); INDEX_UPDATE_APPEND(files_removed); INDEX_UPDATE_APPEND(files_def_update); INDEX_UPDATE_APPEND(types_removed); INDEX_UPDATE_APPEND(types_def_update); INDEX_UPDATE_MERGE(types_derived); INDEX_UPDATE_MERGE(types_instances); INDEX_UPDATE_MERGE(types_uses); INDEX_UPDATE_APPEND(funcs_removed); INDEX_UPDATE_APPEND(funcs_def_update); INDEX_UPDATE_MERGE(funcs_declarations); INDEX_UPDATE_MERGE(funcs_derived); INDEX_UPDATE_MERGE(funcs_callers); INDEX_UPDATE_APPEND(vars_removed); INDEX_UPDATE_APPEND(vars_def_update); INDEX_UPDATE_MERGE(vars_uses); #undef INDEX_UPDATE_APPEND #undef INDEX_UPDATE_MERGE } std::string IndexUpdate::ToString() { rapidjson::StringBuffer output; Writer writer(output); IndexUpdate& update = *this; Reflect(writer, update); return output.GetString(); } // ------------------------ // QUERYDB THREAD FUNCTIONS // ------------------------ void QueryDatabase::RemoveUsrs(SymbolKind usr_kind, const std::vector& to_remove) { // This function runs on the querydb thread. // When we remove an element, we just erase the state from the storage. We do // not update array indices because that would take a huge amount of time for // a very large index. // // There means that there is some memory growth that will never be reclaimed, // but it should be pretty minimal and is solved by simply restarting the // indexer and loading from cache, which is a fast operation. // // TODO: Add "cquery: Reload Index" command which unloads all querydb state // and fully reloads from cache. This will address the memory leak above. switch (usr_kind) { case SymbolKind::File: { for (const Usr& usr : to_remove) files[usr_to_file[LowerPathIfCaseInsensitive(usr)].id].def = nullopt; break; } case SymbolKind::Type: { for (const Usr& usr : to_remove) types[usr_to_type[usr].id].def = nullopt; break; } case SymbolKind::Func: { for (const Usr& usr : to_remove) funcs[usr_to_func[usr].id].def = nullopt; break; } case SymbolKind::Var: { for (const Usr& usr : to_remove) vars[usr_to_var[usr].id].def = nullopt; break; } case SymbolKind::Invalid: break; } } void QueryDatabase::ApplyIndexUpdate(IndexUpdate* update) { // This function runs on the querydb thread. // Example types: // storage_name => std::vector> // merge_update => QueryType::DerivedUpdate => // MergeableUpdate def => QueryType // def->def_var_name => std::vector #define HANDLE_MERGEABLE(update_var_name, def_var_name, storage_name) \ for (auto merge_update : update->update_var_name) { \ auto& def = storage_name[merge_update.id.id]; \ AddRange(&def.def_var_name, merge_update.to_add); \ RemoveRange(&def.def_var_name, merge_update.to_remove); \ } RemoveUsrs(SymbolKind::File, update->files_removed); ImportOrUpdate(update->files_def_update); RemoveUsrs(SymbolKind::Type, update->types_removed); ImportOrUpdate(update->types_def_update); HANDLE_MERGEABLE(types_derived, derived, types); HANDLE_MERGEABLE(types_instances, instances, types); HANDLE_MERGEABLE(types_uses, uses, types); RemoveUsrs(SymbolKind::Func, update->funcs_removed); ImportOrUpdate(update->funcs_def_update); HANDLE_MERGEABLE(funcs_declarations, declarations, funcs); HANDLE_MERGEABLE(funcs_derived, derived, funcs); HANDLE_MERGEABLE(funcs_callers, callers, funcs); RemoveUsrs(SymbolKind::Var, update->vars_removed); ImportOrUpdate(update->vars_def_update); HANDLE_MERGEABLE(vars_uses, uses, vars); #undef HANDLE_MERGEABLE } void QueryDatabase::ImportOrUpdate( const std::vector& updates) { // This function runs on the querydb thread. for (auto& def : updates) { auto it = usr_to_file.find(LowerPathIfCaseInsensitive(def.path)); assert(it != usr_to_file.end()); QueryFile& existing = files[it->second.id]; existing.def = def; UpdateDetailedNames(&existing.detailed_name_idx, SymbolKind::File, it->second.id, def.path); } } void QueryDatabase::ImportOrUpdate( const std::vector& updates) { // This function runs on the querydb thread. for (auto& def : updates) { assert(!def.detailed_name.empty()); auto it = usr_to_type.find(def.usr); assert(it != usr_to_type.end()); assert(it->second.id >= 0 && it->second.id < types.size()); QueryType& existing = types[it->second.id]; // Keep the existing definition if it is higher quality. if (existing.def && existing.def->definition_spelling && !def.definition_spelling) continue; existing.def = def; UpdateDetailedNames(&existing.detailed_name_idx, SymbolKind::Type, it->second.id, def.detailed_name); } } void QueryDatabase::ImportOrUpdate( const std::vector& updates) { // This function runs on the querydb thread. for (auto& def : updates) { assert(!def.detailed_name.empty()); auto it = usr_to_func.find(def.usr); assert(it != usr_to_func.end()); assert(it->second.id >= 0 && it->second.id < funcs.size()); QueryFunc& existing = funcs[it->second.id]; // Keep the existing definition if it is higher quality. if (existing.def && existing.def->definition_spelling && !def.definition_spelling) continue; existing.def = def; UpdateDetailedNames(&existing.detailed_name_idx, SymbolKind::Func, it->second.id, def.detailed_name); } } void QueryDatabase::ImportOrUpdate( const std::vector& updates) { // This function runs on the querydb thread. for (auto& def : updates) { assert(!def.detailed_name.empty()); auto it = usr_to_var.find(def.usr); assert(it != usr_to_var.end()); assert(it->second.id >= 0 && it->second.id < vars.size()); QueryVar& existing = vars[it->second.id]; // Keep the existing definition if it is higher quality. if (existing.def && existing.def->definition_spelling && !def.definition_spelling) continue; existing.def = def; if (!def.is_local) UpdateDetailedNames(&existing.detailed_name_idx, SymbolKind::Var, it->second.id, def.detailed_name); } } void QueryDatabase::UpdateDetailedNames(size_t* qualified_name_index, SymbolKind kind, size_t symbol_index, const std::string& name) { if (*qualified_name_index == -1) { detailed_names.push_back(name); symbols.push_back(SymbolIdx(kind, symbol_index)); *qualified_name_index = detailed_names.size() - 1; } else { detailed_names[*qualified_name_index] = name; } } TEST_SUITE("query") { IndexUpdate GetDelta(IndexFile previous, IndexFile current) { QueryDatabase db; IdMap previous_map(&db, previous.id_cache); IdMap current_map(&db, current.id_cache); return IndexUpdate::CreateDelta(&previous_map, ¤t_map, &previous, ¤t); } TEST_CASE("remove defs") { IndexFile previous("foo.cc"); IndexFile current("foo.cc"); previous.Resolve(previous.ToTypeId("usr1"))->def.definition_spelling = Range(Position(1, 0)); previous.Resolve(previous.ToFuncId("usr2"))->def.definition_spelling = Range(Position(2, 0)); previous.Resolve(previous.ToVarId("usr3"))->def.definition_spelling = Range(Position(3, 0)); IndexUpdate update = GetDelta(previous, current); REQUIRE(update.types_removed == std::vector{"usr1"}); REQUIRE(update.funcs_removed == std::vector{"usr2"}); REQUIRE(update.vars_removed == std::vector{"usr3"}); } TEST_CASE("do not remove ref-only defs") { IndexFile previous("foo.cc"); IndexFile current("foo.cc"); previous.Resolve(previous.ToTypeId("usr1")) ->uses.push_back(Range(Position(1, 0))); previous.Resolve(previous.ToFuncId("usr2")) ->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(2, 0)), false /*is_implicit*/)); previous.Resolve(previous.ToVarId("usr3")) ->uses.push_back(Range(Position(3, 0))); IndexUpdate update = GetDelta(previous, current); REQUIRE(update.types_removed == std::vector{}); REQUIRE(update.funcs_removed == std::vector{}); REQUIRE(update.vars_removed == std::vector{}); } TEST_CASE("func callers") { IndexFile previous("foo.cc"); IndexFile current("foo.cc"); IndexFunc* pf = previous.Resolve(previous.ToFuncId("usr")); IndexFunc* cf = current.Resolve(current.ToFuncId("usr")); pf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(1, 0)), false /*is_implicit*/)); cf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(2, 0)), false /*is_implicit*/)); IndexUpdate update = GetDelta(previous, current); REQUIRE(update.funcs_removed == std::vector{}); REQUIRE(update.funcs_callers.size() == 1); REQUIRE(update.funcs_callers[0].id == QueryFuncId(0)); REQUIRE(update.funcs_callers[0].to_remove.size() == 1); REQUIRE(update.funcs_callers[0].to_remove[0].loc.range == Range(Position(1, 0))); REQUIRE(update.funcs_callers[0].to_add.size() == 1); REQUIRE(update.funcs_callers[0].to_add[0].loc.range == Range(Position(2, 0))); } TEST_CASE("type usages") { IndexFile previous("foo.cc"); IndexFile current("foo.cc"); IndexType* pt = previous.Resolve(previous.ToTypeId("usr")); IndexType* ct = current.Resolve(current.ToTypeId("usr")); pt->uses.push_back(Range(Position(1, 0))); ct->uses.push_back(Range(Position(2, 0))); IndexUpdate update = GetDelta(previous, current); REQUIRE(update.types_removed == std::vector{}); REQUIRE(update.types_def_update == std::vector{}); REQUIRE(update.types_uses.size() == 1); REQUIRE(update.types_uses[0].to_remove.size() == 1); REQUIRE(update.types_uses[0].to_remove[0].range == Range(Position(1, 0))); REQUIRE(update.types_uses[0].to_add.size() == 1); REQUIRE(update.types_uses[0].to_add[0].range == Range(Position(2, 0))); } TEST_CASE("apply delta") { IndexFile previous("foo.cc"); IndexFile current("foo.cc"); IndexFunc* pf = previous.Resolve(previous.ToFuncId("usr")); IndexFunc* cf = current.Resolve(current.ToFuncId("usr")); pf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(1, 0)), false /*is_implicit*/)); pf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(2, 0)), false /*is_implicit*/)); cf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(4, 0)), false /*is_implicit*/)); cf->callers.push_back(IndexFuncRef(IndexFuncId(0), Range(Position(5, 0)), false /*is_implicit*/)); QueryDatabase db; IdMap previous_map(&db, previous.id_cache); IdMap current_map(&db, current.id_cache); REQUIRE(db.funcs.size() == 1); IndexUpdate import_update = IndexUpdate::CreateDelta(nullptr, &previous_map, nullptr, &previous); IndexUpdate delta_update = IndexUpdate::CreateDelta( &previous_map, ¤t_map, &previous, ¤t); db.ApplyIndexUpdate(&import_update); REQUIRE(db.funcs[0].callers.size() == 2); REQUIRE(db.funcs[0].callers[0].loc.range == Range(Position(1, 0))); REQUIRE(db.funcs[0].callers[1].loc.range == Range(Position(2, 0))); db.ApplyIndexUpdate(&delta_update); REQUIRE(db.funcs[0].callers.size() == 2); REQUIRE(db.funcs[0].callers[0].loc.range == Range(Position(4, 0))); REQUIRE(db.funcs[0].callers[1].loc.range == Range(Position(5, 0))); } }