9c9e6b0bcb
## Summary Introduces a two-layer state model that cleanly separates disk-based project state from per-open-file editing state, and redistributes responsibilities across server components so each has a single, clear role. ## New types **Workspace** — all persistent, project-wide shared state: - CompilationDatabase, PathPool, DependencyGraph, CompileGraph - path_to_module mapping, PCH cache, PCM cache, PCM paths - ProjectIndex, MergedIndex shards - CliceConfig - Methods: on_file_saved(), on_file_closed(), load/save/cleanup_cache(), build_module_map(), fill_pcm_deps(), cancel_all() **Session** — volatile per-open-file editing state: - text, version, generation, ast_dirty - pch_ref (references Workspace.pch_cache), ast_deps, header_context - file_index (OpenFileIndex for unsaved buffer) - path_id member for self-identification ## Component responsibilities after refactor | Component | Role | Owns state? | |-----------|------|-------------| | **Workspace** | Disk truth + shared caches | Yes (all project state) | | **Session** | One open file editing state | Yes (per-file only) | | **Compiler** | Compilation pipeline, worker communication | No (references only) | | **Indexer** | Index queries + background indexing scheduling | Scheduling state only | | **MasterServer** | LSP protocol dispatch + lifecycle coordination | sessions map | ## What moved where **Into Workspace** (from Compiler/MasterServer): - PCH/PCM cache management (load_cache, save_cache, cleanup_cache) - Module map building (build_module_map, fill_pcm_deps) - File lifecycle hooks (on_file_saved, on_file_closed) - cancel_all, OpenFileIndex/MergedIndexShard type definitions **Into Session** (from Compiler documents map): - Document text, version, generation, ast_dirty - PCH reference, dependency snapshot, header context **Into Indexer** (from MasterServer): - Background indexing queue, scheduling state, idle timer - schedule(), enqueue(), run_background_indexing() **Into syntax/completion.h** (from Compiler): - detect_completion_context() — pure text parsing - complete_module_import() — prefix match on module names - complete_include_path() — directory listing against search paths **Inlined into MasterServer** (from Compiler): - didOpen/didChange/didClose/didSave handlers - switchContext/currentContext - publish_diagnostics/clear_diagnostics **Deleted from Compiler** (9 methods): - open_document, apply_changes, close_document, on_save - switch_context, get_active_context, invalidate_host_contexts - on_file_closed, on_file_saved, complete_include, complete_import ## Tests - 481 tests pass (465 existing + 16 new completion tests) - New: tests/unit/syntax/completion_tests.cpp ## Diff stats 15 files changed, +1857, -1555 <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit ## Release Notes * **New Features** * Enhanced completion support for include paths and module imports with improved context detection. * Added background indexing system for automatic project symbol indexing. * **Bug Fixes** * Improved reliability of document change tracking and compilation state management. * Better handling of header file compilation contexts. * **Tests** * Added unit tests for completion context detection and module/include path completion. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
clice
clice is a next-generation language server designed for modern C++. Through excellent asynchronous task scheduling and intelligent caching, it achieves a lower memory footprint and faster response times.
Beyond performance, clice provides instantiation-aware template processing, supports switching header contexts between different source files (including non-self-contained headers), and offers comprehensive support for C++20 modules, from code completion to go-to-definition. Our goal is to provide C++ developers with a truly fast, precise, and intelligent development companion.
Important
Support for header contexts and C++20 modules are core features currently under active development. They will be progressively refined in upcoming releases. Stay tuned!
Getting started
Download the latest clice binary from the releases page and install the vscode extension. Then, add the following configuration to your .vscode/settings.json file:
{
// Optional: Set this to an empty string to turn off the clangd.
"clangd.path": "",
// Point this to the clice binary you downloaded.
"clice.executable": "/path/to/your/clice/executable",
}
Note
As an early version, please do not use it in a production environment. Crashes are expected, and we welcome you to submit issues.
Documentation
To learn more about building, installing, and configuring clice, or to dive deep into its features and architecture, please visit our official documentation at clice.io.