#pragma once #include "indexer.h" #include "serializer.h" using Usr = std::string; // TODO: in types, store refs separately from irefs. Then we can drop // 'interesting' from location when that is cleaned up. // TODO: Switch over to QueryableLocation. Figure out if there is // a good way to get the indexer using it. I don't think so // since we may discover more files while indexing a file. // // We could also reuse planned USR caching system for file // paths. struct QueryableLocation { Usr path; int line; int column; bool interesting; QueryableLocation() : path(""), line(-1), column(-1), interesting(false) {} QueryableLocation(Usr path, int line, int column, bool interesting) : path(path), line(line), column(column), interesting(interesting) {} QueryableLocation OffsetColumn(int offset) const { return QueryableLocation(path, line, column + offset, interesting); } bool operator==(const QueryableLocation& other) const { // Note: We ignore |is_interesting|. return path == other.path && line == other.line && column == other.column; } bool operator!=(const QueryableLocation& other) const { return !(*this == other); } bool operator<(const QueryableLocation& o) const { return path < o.path && line < o.line && column < o.column && interesting < o.interesting; } }; struct QueryableRange { QueryableLocation start; QueryableLocation end; QueryableRange() {} QueryableRange(QueryableLocation start, QueryableLocation end) : start(start), end(end) {} QueryableRange OffsetStartColumn(int offset) const { return QueryableRange(start.OffsetColumn(offset), end); } bool operator==(const QueryableRange& other) const { // Note: We ignore |is_interesting|. return start == other.start && end == other.end; } bool operator!=(const QueryableRange& other) const { return !(*this == other); } bool operator<(const QueryableRange& o) const { return start < o.start; } }; struct UsrRef { Usr usr; QueryableRange loc; UsrRef() {} UsrRef(Usr usr, QueryableRange loc) : usr(usr), loc(loc) {} bool operator==(const UsrRef& other) const { return usr == other.usr && loc.start == other.loc.start; } bool operator!=(const UsrRef& other) const { return !(*this == other); } bool operator<(const UsrRef& other) const { return usr < other.usr && loc.start < other.loc.start; } }; // There are two sources of reindex updates: the (single) definition of a // symbol has changed, or one of many users of the symbol has changed. // // For simplicitly, if the single definition has changed, we update all of the // associated single-owner definition data. See |Update*DefId|. // // If one of the many symbol users submits an update, we store the update such // that it can be merged with other updates before actually being applied to // the main database. See |MergeableUpdate|. template struct MergeableUpdate { // The type/func/var which is getting new usages. Usr usr; // Entries to add and remove. std::vector to_add; std::vector to_remove; MergeableUpdate(Usr usr, const std::vector& to_add) : usr(usr), to_add(to_add) {} MergeableUpdate(Usr usr, const std::vector& to_add, const std::vector& to_remove) : usr(usr), to_add(to_add), to_remove(to_remove) {} }; template struct ReplacementUpdate { // The type/func/var which is getting new usages. Usr usr; // New entries. std::vector values; ReplacementUpdate(Usr usr, const std::vector& entries) : usr(usr), entries(entries) {} }; struct QueryableFile { struct Def { Usr usr; // Outline of the file (ie, for code lens). std::vector outline; // Every symbol found in the file (ie, for goto definition) std::vector all_symbols; }; using DefUpdate = Def; DefUpdate def; QueryableFile() {} QueryableFile(const IndexedFile& indexed); }; struct QueryableTypeDef { using DefUpdate = TypeDefDefinitionData; using DerivedUpdate = MergeableUpdate; using InstantiationsUpdate = MergeableUpdate; using UsesUpdate = MergeableUpdate; DefUpdate def; std::vector derived; std::vector instantiations; std::vector uses; QueryableTypeDef() : def("") {} QueryableTypeDef(IdCache& id_cache, const IndexedTypeDef& indexed); }; struct QueryableFuncDef { using DefUpdate = FuncDefDefinitionData; using DeclarationsUpdate = MergeableUpdate; using DerivedUpdate = MergeableUpdate; using CallersUpdate = MergeableUpdate; using UsesUpdate = MergeableUpdate; DefUpdate def; std::vector declarations; std::vector derived; std::vector callers; std::vector uses; QueryableFuncDef() : def("") {} QueryableFuncDef(IdCache& id_cache, const IndexedFuncDef& indexed); }; struct QueryableVarDef { using DefUpdate = VarDefDefinitionData; using UsesUpdate = MergeableUpdate; DefUpdate def; std::vector uses; QueryableVarDef() : def("") {} QueryableVarDef(IdCache& id_cache, const IndexedVarDef& indexed); }; enum class SymbolKind { Invalid, File, Type, Func, Var }; struct SymbolIdx { SymbolKind kind; uint64_t idx; SymbolIdx() : kind(SymbolKind::Invalid), idx(-1) {} // Default ctor needed by stdlib. Do not use. SymbolIdx(SymbolKind kind, uint64_t idx) : kind(kind), idx(idx) {} }; struct IndexUpdate { // Creates a new IndexUpdate that will import |file|. static IndexUpdate CreateImport(IndexedFile& file); static IndexUpdate CreateDelta(IndexedFile& current, IndexedFile& updated); // Merge |update| into this update; this can reduce overhead / index update // work can be parallelized. void Merge(const IndexUpdate& update); // File updates. std::vector files_removed; std::vector files_def_update; // Type updates. std::vector types_removed; std::vector types_def_update; std::vector types_derived; std::vector types_instantiations; std::vector types_uses; // Function updates. std::vector funcs_removed; std::vector funcs_def_update; std::vector funcs_declarations; std::vector funcs_derived; std::vector funcs_callers; std::vector funcs_uses; // Variable updates. std::vector vars_removed; std::vector vars_def_update; std::vector vars_uses; private: // Creates an index update assuming that |previous| is already // in the index, so only the delta between |previous| and |current| // will be applied. IndexUpdate(IndexedFile& previous, IndexedFile& current); }; // The query database is heavily optimized for fast queries. It is stored // in-memory. struct QueryableDatabase { // Indicies between lookup vectors are related to symbols, ie, index 5 in // |qualified_names| matches index 5 in |symbols|. std::vector qualified_names; std::vector symbols; // Raw data storage. std::vector files; // File path is stored as a Usr. std::vector types; std::vector funcs; std::vector vars; // Lookup symbol based on a usr. std::unordered_map usr_to_symbol; // Insert the contents of |update| into |db|. void ApplyIndexUpdate(IndexUpdate* update); void RemoveUsrs(const std::vector& to_remove); void ImportOrUpdate(const std::vector& updates); void ImportOrUpdate(const std::vector& updates); void ImportOrUpdate(const std::vector& updates); void ImportOrUpdate(const std::vector& updates); }; // TODO: For supporting vscode, lets' // - have our normal daemon system // - have frontend --language-server which accepts JSON RPC language server in stdin and emits language server // JSON in stdout. vscode extension will run the executable this way. it will connect to daemon as normal. // this means that vscode instance can be killed without actually killing core indexer process. // $ indexer --language-server // - maybe? have simple front end which lets user run // $ indexer --action references --location foo.cc:20:5 // // // https://github.com/Microsoft/vscode-languageserver-node/blob/master/client/src/main.ts