Refactor Indexer (#78)

This commit is contained in:
ykiko
2025-02-17 00:22:41 +08:00
committed by GitHub
parent 5a7915d40f
commit 26c66100b9
22 changed files with 1234 additions and 1042 deletions

View File

@@ -0,0 +1,211 @@
#pragma once
#include "Config.h"
#include "Database.h"
#include "Protocol.h"
#include "Async/Async.h"
#include "Compiler/Compilation.h"
#include "Support/JSON.h"
#include "Index/SymbolIndex.h"
namespace clice {
struct TranslationUnit;
struct HeaderIndex {
/// The index file path(not include suffix, e.g. `.sidx` and `.fidx`).
std::string path;
/// The hash of the symbol index.
llvm::XXH128_hash_t symbolHash;
/// The hash of the feature index.
llvm::XXH128_hash_t featureHash;
};
struct Context {
/// The include chain that introduces this context.
uint32_t include = -1;
/// The index information of this context.
uint32_t index = -1;
};
struct IncludeLocation {
/// The location of the include directive.
uint32_t line = -1;
/// The index of the file that includes this header.
uint32_t include = -1;
/// The file name of the header in the string pool. Beacuse
/// a header may be included by multiple files, so we use
/// a string pool to cache the file name to reduce the memory
/// usage.
uint32_t filename = -1;
};
struct Header {
/// The path of the header file.
std::string srcPath;
/// All indices of this header.
std::vector<HeaderIndex> indices;
/// All header contexts of this header.
llvm::DenseMap<TranslationUnit*, std::vector<Context>> contexts;
/// The active translation unit and the index of the context.
std::pair<TranslationUnit*, uint32_t> active = {nullptr, -1};
};
struct TranslationUnit {
/// The source file path.
std::string srcPath;
/// The index file path(not include suffix, e.g. `.sidx` and `.fidx`).
std::string indexPath;
/// All headers included by this translation unit.
llvm::DenseSet<Header*> headers;
/// The time when this translation unit is indexed. Used to determine
/// whether the index file is outdated.
std::chrono::milliseconds mtime;
/// All include locations introduced by this translation unit.
/// Note that if a file has guard macro or pragma once, we will
/// record it at most once.
std::vector<IncludeLocation> locations;
/// The version of the translation unit.
uint32_t version = 0;
};
namespace proto {
struct IncludeLocation {
/// The line number of the include directive.
uint32_t line;
/// The filename of the included header.
std::string filename;
};
struct HeaderContext {
/// The path of the source file.
std::string srcFile;
/// The path of the context file.
std::string contextFile;
/// The index of the context.
uint32_t index = -1;
/// The version of the context.
uint32_t version = 0;
};
using HeaderContextGroups = std::vector<std::vector<HeaderContext>>;
} // namespace proto
class IncludeGraph {
protected:
IncludeGraph(const config::IndexOptions& options) : options(options) {}
~IncludeGraph();
void load(const json::Value& json);
json::Value dump();
async::Task<> index(llvm::StringRef file, CompilationDatabase& database);
public:
/// Return all header context of the given file.
/// FIXME: The results are grouped by the index file. And a header actually
/// may have thousands of contexts, of course, users don't want to see all
/// of them. For each index file, we return the first 10 contexts. In the future
/// we may add a parameter to control the number of contexts or set filter.
proto::HeaderContextGroups contextAll(llvm::StringRef file);
/// Return current header context of the given file.
std::optional<proto::HeaderContext> contextCurrent(llvm::StringRef file);
/// Switch to the given header context.
void contextSwitch(const proto::HeaderContext& context);
/// Resolve the header context to the include chain.
std::vector<proto::IncludeLocation> contextResolve(const proto::HeaderContext& context);
private:
struct SymbolID {
uint64_t hash;
std::string name;
};
/// Return all indices of the given translation unit. If the file is empty,
/// return all indices of the IncludeGraph.
std::vector<std::string> indices(TranslationUnit* tu = nullptr);
/// Resolve the symbol at the given position.
async::Task<std::vector<SymbolID>> resolve(const proto::TextDocumentPositionParams& params);
using LookupCallback = llvm::unique_function<bool(llvm::StringRef path,
llvm::StringRef content,
const index::SymbolIndex::Symbol& symbol)>;
async::Task<> lookup(llvm::ArrayRef<SymbolID> targets,
llvm::ArrayRef<std::string> files,
LookupCallback callback);
public:
/// Lookup the reference information according to the given position.
async::Task<proto::ReferenceResult> lookup(const proto::ReferenceParams& params,
RelationKind kind);
/// According to the given file and offset, resolve the symbol at the offset.
async::Task<proto::HierarchyPrepareResult>
prepareHierarchy(const proto::HierarchyPrepareParams& params);
async::Task<proto::CallHierarchyIncomingCallsResult>
incomingCalls(const proto::HierarchyParams& params);
async::Task<proto::CallHierarchyOutgoingCallsResult>
outgoingCalls(const proto::HierarchyParams& params);
async::Task<proto::TypeHierarchyResult> typeHierarchy(const proto::HierarchyParams& params,
bool super);
private:
std::string getIndexPath(llvm::StringRef file);
/// Check whether the given file needs to be updated. If so,
/// return the translation unit. Otherwise, return nullptr.
async::Task<TranslationUnit*> check(llvm::StringRef file);
/// Add all possible header contexts for the tu from the AST info.
uint32_t addIncludeChain(std::vector<IncludeLocation>& locations,
llvm::DenseMap<clang::FileID, uint32_t>& files,
clang::SourceManager& SM,
clang::FileID fid);
void addContexts(ASTInfo& info,
TranslationUnit* tu,
llvm::DenseMap<clang::FileID, uint32_t>& files);
async::Task<> updateIndices(ASTInfo& info,
TranslationUnit* tu,
llvm::DenseMap<clang::FileID, uint32_t>& files);
private:
const config::IndexOptions& options;
llvm::StringMap<Header*> headers;
llvm::StringMap<TranslationUnit*> tus;
std::vector<std::string> pathPool;
llvm::StringMap<std::uint32_t> pathIndices;
SourceConverter SC;
};
} // namespace clice

View File

@@ -2,179 +2,41 @@
#include "Config.h"
#include "Database.h"
#include "Protocol.h"
#include "Async/Async.h"
#include "AST/RelationKind.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/StringSet.h"
#include "IncludeGraph.h"
namespace clice {
class ASTInfo;
class IncludeGraph;
class Indexer {
class Indexer : public IncludeGraph {
public:
Indexer(const config::IndexOptions& options, CompilationDatabase& database) :
options(options), database(database) {}
Indexer(CompilationDatabase& database, const config::IndexOptions& options);
~Indexer();
/// Add a file to wait for indexing.
void add(std::string file);
struct TranslationUnit;
/// Remove a file from indexing.
void remove(std::string file);
struct HeaderIndex {
/// The index file path(not include suffix, e.g. `.sidx` and `.fidx`).
std::string path;
void indexAll();
/// The hash of the symbol index.
llvm::XXH128_hash_t symbolHash;
void save();
/// The hash of the feature index.
llvm::XXH128_hash_t featureHash;
};
struct Context {
/// The include chain that introduces this context.
uint32_t include = -1;
/// The index information of this context.
uint32_t index = -1;
};
struct IncludeLocation {
/// The location of the include directive.
uint32_t line = -1;
/// The index of the file that includes this header.
uint32_t include = -1;
/// The file name of the header in the string pool. Beacuse
/// a header may be included by multiple files, so we use
/// a string pool to cache the file name to reduce the memory
/// usage.
uint32_t filename = -1;
};
struct Header {
/// The path of the header file.
std::string srcPath;
/// All indices of this header.
std::vector<HeaderIndex> indices;
/// All header contexts of this header.
llvm::DenseMap<TranslationUnit*, std::vector<Context>> contexts;
};
struct TranslationUnit {
/// The source file path.
std::string srcPath;
/// The index file path(not include suffix, e.g. `.sidx` and `.fidx`).
std::string indexPath;
/// All headers included by this translation unit.
llvm::DenseSet<Header*> headers;
/// The time when this translation unit is indexed. Used to determine
/// whether the index file is outdated.
std::chrono::milliseconds mtime;
/// All include locations introduced by this translation unit.
/// Note that if a file has guard macro or pragma once, we will
/// record it at most once.
std::vector<IncludeLocation> locations;
};
using Self = Indexer;
/// Check whether the given file needs to be updated and return the translation unit.
/// If not need to update, return nullptr.
async::Task<TranslationUnit*> check(this Self& self, llvm::StringRef file);
uint32_t addIncludeChain(std::vector<Indexer::IncludeLocation>& locations,
llvm::DenseMap<clang::FileID, uint32_t>& files,
clang::SourceManager& SM,
clang::FileID fid);
/// Add all possible header contexts for the AST info.
void addContexts(this Self& self,
ASTInfo& info,
TranslationUnit* tu,
llvm::DenseMap<clang::FileID, uint32_t>& files);
/// Index the given AST, write the index information to disk.
async::Task<> updateIndices(this Self& self,
ASTInfo& info,
TranslationUnit* tu,
llvm::DenseMap<clang::FileID, uint32_t>& files);
async::Task<> index(this Self& self, llvm::StringRef file);
/// Index the given file(for opened file).
async::Task<> index(llvm::StringRef file, ASTInfo& info);
async::Task<> indexAll();
/// Generate the index file path based on time and file name.
std::string getIndexPath(llvm::StringRef file);
/// Dump the index information to JSON.
json::Value dumpToJSON();
async::Task<proto::SemanticTokens> semanticTokens(llvm::StringRef file);
/// Dump all index information of the given file for test.
void dumpForTest(llvm::StringRef file);
void lookupHeaderContexts(llvm::StringRef file);
/// Save the index information to disk.
void saveToDisk();
/// Load the index information from disk.
void loadFromDisk();
void load();
private:
struct SymbolID {
uint64_t id;
std::string name;
};
async::Task<std::unique_ptr<llvm::MemoryBuffer>> read(llvm::StringRef path);
async::Task<> lookup(llvm::ArrayRef<SymbolID> ids,
RelationKind kind,
llvm::StringRef srcPath,
llvm::StringRef content,
std::string indexPath,
std::vector<proto::Location>& result);
public:
async::Task<std::vector<proto::Location>>
lookup(const proto::TextDocumentPositionParams& params, RelationKind kind);
async::Task<proto::CallHierarchyIncomingCallsResult>
incomingCalls(const proto::CallHierarchyIncomingCallsParams& params);
async::Task<proto::CallHierarchyOutgoingCallsResult>
outgoingCalls(const proto::CallHierarchyOutgoingCallsParams& params);
async::Task<proto::TypeHierarchySupertypesResult>
supertypes(const proto::TypeHierarchySupertypesParams& params);
async::Task<proto::TypeHierarchySubtypesResult>
subtypes(const proto::TypeHierarchySubtypesParams& params);
async::Task<> index(std::string file);
private:
const config::IndexOptions& options;
CompilationDatabase& database;
llvm::StringMap<Header*> headers;
llvm::StringMap<TranslationUnit*> tus;
const config::IndexOptions& options;
bool locked = false;
llvm::StringMap<async::Task<>> tasks;
llvm::StringSet<> pending;
std::vector<std::string> pathPool;
llvm::StringMap<std::uint32_t> pathIndices;
std::size_t concurrency = std::thread::hardware_concurrency();
};
} // namespace clice

View File

@@ -1,6 +1,5 @@
#pragma once
#include "Config.h"
#include "Indexer.h"
#include "Protocol.h"
@@ -85,31 +84,34 @@ private:
/// Language Features
/// ============================================================================
async::Task<> onGotoDeclaration(json::Value id, const proto::DeclarationParams& params);
async::Task<> onGotoDefinition(json::Value id, const proto::DefinitionParams& params);
async::Task<> onGotoTypeDefinition(json::Value id, const proto::TypeDefinitionParams& params);
async::Task<> onGotoImplementation(json::Value id, const proto::ImplementationParams& params);
async::Task<> onFindReferences(json::Value id, const proto::ReferenceParams& params);
async::Task<> onPrepareCallHierarchy(json::Value id,
const proto::CallHierarchyPrepareParams& params);
async::Task<> onIncomingCall(json::Value id,
const proto::CallHierarchyIncomingCallsParams& params);
async::Task<> onOutgoingCall(json::Value id,
const proto::CallHierarchyOutgoingCallsParams& params);
async::Task<> onPrepareTypeHierarchy(json::Value id,
const proto::TypeHierarchyPrepareParams& params);
async::Task<> onSupertypes(json::Value id, const proto::TypeHierarchySupertypesParams& params);
async::Task<> onSubtypes(json::Value id, const proto::TypeHierarchySubtypesParams& params);
// async::Task<> onGotoDeclaration(json::Value id, const proto::DeclarationParams& params);
//
// async::Task<> onGotoDefinition(json::Value id, const proto::DefinitionParams& params);
//
// async::Task<> onGotoTypeDefinition(json::Value id, const proto::TypeDefinitionParams&
// params);
//
// async::Task<> onGotoImplementation(json::Value id, const proto::ImplementationParams&
// params);
//
// async::Task<> onFindReferences(json::Value id, const proto::ReferenceParams& params);
//
// async::Task<> onPrepareCallHierarchy(json::Value id,
// const proto::CallHierarchyPrepareParams& params);
//
// async::Task<> onIncomingCall(json::Value id,
// const proto::CallHierarchyIncomingCallsParams& params);
//
// async::Task<> onOutgoingCall(json::Value id,
// const proto::CallHierarchyOutgoingCallsParams& params);
//
// async::Task<> onPrepareTypeHierarchy(json::Value id,
// const proto::TypeHierarchyPrepareParams& params);
//
// async::Task<> onSupertypes(json::Value id, const proto::TypeHierarchySupertypesParams&
// params);
//
// async::Task<> onSubtypes(json::Value id, const proto::TypeHierarchySubtypesParams& params);
async::Task<> onDocumentHighlight(json::Value id, const proto::DocumentHighlightParams& params);