Add MVSC and GCC support in preparation for VCPKG

.github/actions/generate-dataset/action.yml

@@ -21,9 +21,11 @@ runs:
     # 2️⃣  Remove any .venv that might have been created earlier
     #     with Python 3.13, so we always recreate it with 3.12.
     # ─────────────────────────────────────────────────────────────
-    - name: Remove stale virtual-env
+    - name: Recreate virtual-env with the correct Python
       shell: bash
-      run: rm -rf "$GITHUB_WORKSPACE/.venv"
+      run: |
+        rm -rf "$GITHUB_WORKSPACE/.venv"
+        python3 -m venv "$GITHUB_WORKSPACE/.venv"
 
     # ─────────────────────────────────────────────────────────────
     # 3️⃣  Generate the synthetic data
@@ -36,5 +38,6 @@ runs:
     # 4️⃣  Generate the sentence embeddings
     # ─────────────────────────────────────────────────────────────
     - name: Generate embeddings
+      if: runner.os != 'Windows'
       shell: bash
       run: make generate-embeddings

.github/workflows/benchmark.yaml

@@ -10,9 +10,9 @@ run-name: >-
 
 on:
   push:
-    branches: [ '*' ]
+    branches: [ main, dev ]
   pull_request:
-    branches: [ '*' ]
+    branches: [ main, dev ]
 
 concurrency:
   group: Benchmarker CI-${{ github.ref }}
@@ -28,7 +28,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        platform: [ ubuntu-latest, macos-latest ]
+        platform: [ ubuntu-latest ]
 
     defaults:
       run:

.github/workflows/examples.yml

@@ -10,6 +10,7 @@ run-name: >-
 
 on:
   push:
+    branches: [ main, dev ]
   pull_request:
     branches: [ main, dev ]
 

.github/workflows/flatbuffers-ci.yml

@@ -12,8 +12,10 @@ run-name: >-
 # Trigger on every push & PR, on all branches
 # ─────────────────────────────────────────────────────────────
 on:
-  push:     # no branches filter ⇒ every branch
-  pull_request:  # no branches filter ⇒ every target branch
+  push:
+    branches: [ main, dev ]
+  pull_request:
+    branches: [ main, dev ]
 concurrency:
   group: flatbuffers-${{ github.ref }}
   cancel-in-progress: true

.github/workflows/fsst.yaml

@@ -13,8 +13,9 @@ run-name: >-
 # ──────────────────────────────────────────────────────────────────────────────
 on:
   push:
+    branches: [ main, dev ]
   pull_request:
-    branches: [ "main", "dev" ]
+    branches: [ main, dev ]
 
 # Cancel in-flight runs on the same branch/PR so we do not waste minutes
 concurrency:

.github/workflows/header-check.yml

@@ -11,6 +11,7 @@ run-name: >-
 
 on:
   push:
+    branches: [ main, dev ]
   pull_request:
     branches: [ main, dev ]
 

.github/workflows/python.yml

@@ -13,6 +13,7 @@ run-name: >-
 # ────────────────────────────────────────────────────────
 on:
   push:
+    branches: [ main, dev ]
   pull_request:
     branches: [ main, dev ]
   workflow_dispatch:

.github/workflows/rust.yml

@@ -14,8 +14,9 @@ run-name: >-
 
 on:
   push:
+    branches: [ main, dev ]
   pull_request:
-    branches: [ "main", "dev" ]
+    branches: [ main, dev ]
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}

.gitignore

@@ -126,3 +126,8 @@ rust/target/
 skbuild-*/
 .idea/
 .venv/
+
+# Windows build artifacts
+build_win.bat
+cmake_output.txt
+build_win/

CLAUDE.md

@@ -0,0 +1,121 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## What is FastLanes
+
+FastLanes is a C++20 columnar compression storage format — "Like Parquet, but with 40% better compression and 40× faster decoding." Zero external dependencies, SIMD-friendly without explicit SIMD instructions. Bindings exist for Python (`python/`), Rust (`rust/`), C (`src/c_api/`), and CUDA (`cuda/`).
+
+## Build Commands
+
+FastLanes uses CMake 3.22+ with Ninja. On Linux/macOS it requires Clang >= 13. On Windows it uses MSVC (set up via `vcvarsall.bat`).
+
+### Configure and build (Release with tests)
+```bash
+cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DFLS_BUILD_TESTING=ON
+cmake --build build --parallel
+```
+
+### Run all tests
+```bash
+cd build && ctest -j4 --output-on-failure --timeout 300 -E QuickFuzz
+```
+
+### Run a single test by filter
+```bash
+build/test/src/dataset_tests/dataset_tests.exe --gtest_filter=FastLanesReaderTester.issue_000
+```
+
+### Run a single test target
+```bash
+cmake --build build --target unit_test && ctest -R unit_test --output-on-failure
+```
+
+### Key CMake options
+| Option | Default | Purpose |
+|--------|---------|---------|
+| `FLS_BUILD_TESTING` | OFF | Build tests (fetches GoogleTest v1.15.2) |
+| `FLS_BUILD_SHARED_LIBS` | OFF | Build as shared library (DLL) instead of static |
+| `FLS_BUILD_BENCHMARKING` | OFF | Build benchmarks |
+| `FLS_BUILD_PYTHON` | OFF | Build Python bindings |
+| `FLS_BUILD_CUDA` | OFF | Build CUDA reader |
+| `FLS_ENABLE_CLANG_TIDY` | OFF | Enable clang-tidy on all targets |
+
+### Windows-specific (MSVC)
+
+Invoke builds via a `.bat` that calls `vcvarsall.bat` first. Example pattern:
+```bat
+call "C:\Program Files\Microsoft Visual Studio\...\vcvarsall.bat" arm64
+cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DFLS_BUILD_TESTING=ON
+cmake --build build --parallel
+```
+
+Test data can be cached across builds by setting `FASTLANES_DATA_DIR` environment variable to an existing data directory (e.g., `build_release/_deps/data-src`).
+
+### Format code
+The project uses `.clang-format` (LLVM base, tabs, 120-column limit). Run clang-format on changed files before committing.
+
+## Architecture
+
+### Public API
+
+The main entry point is `fastlanes::Connection` (in `src/include/fls/connection.hpp`):
+```cpp
+auto conn = fastlanes::connect();
+conn->read_csv("input/");       // ingest CSV
+conn->to_fls("output/");        // write FastLanes format
+auto reader = conn->read_fls("data.fls");  // read back
+auto table = reader->materialize();
+```
+
+`TableReader` provides rowgroup-level random access. `RowgroupReader` decompresses individual rowgroups. The reader stack: `TableReader` → `RowgroupReader` → `RowgroupView` → `ColumnView` → `SegmentView`.
+
+### Library structure
+
+All source is under `src/`. Each subdirectory builds an OBJECT library that gets linked into the single `FastLanes` library target. Key components:
+
+- **`cor/`** — Core: architecture detection, CPU features, layout (`Buf`), compression/decompression engines
+- **`expression/`** — Expression-based encoding: physical expressions, operators (RLE, FSST, ALP, dict, delta, etc.), interpreter
+- **`encoder/`** — High-level encoding pipeline, materializer (decompression)
+- **`wizard/`** — Schema discovery: analyzes data and selects optimal encoding per column
+- **`reader/`** — File reading: segments, column views, rowgroup views, table reader
+- **`table/`** — In-memory table representation: `Rowgroup`, `Table`, `Vector`, typed columns
+- **`footer/`** — FlatBuffers-generated metadata descriptors (table, rowgroup, column, segment)
+- **`alp/`** — ALP (Adaptive Lossless Floating-Point) compression codec
+- **`primitive/`** — Low-level primitives: bitpacking, patching, FSST string compression
+
+### DLL / Shared library support (Windows)
+
+The `FLS_API` macro in `src/include/fls/api/api.hpp` controls symbol visibility:
+- `FLS_STATIC` defined → `FLS_API` is empty (static build)
+- `FLS_BUILD_DLL` defined → `FLS_API` is `__declspec(dllexport)` (building the DLL)
+- Neither defined → `FLS_API` is `__declspec(dllimport)` (consuming the DLL)
+
+When `FLS_BUILD_SHARED_LIBS=ON`, `FLS_BUILD_DLL` is set directory-scoped via `add_compile_definitions` in `src/CMakeLists.txt` so all object libraries under `src/` get it. Test targets (under `test/`) don't get it, so `FLS_API` correctly resolves to `dllimport` for them.
+
+Any public function or class that test code (or external consumers) calls across the DLL boundary must be marked `FLS_API`. For template functions, the explicit instantiations in the .cpp must also carry `FLS_API`.
+
+Note: `WINDOWS_EXPORT_ALL_SYMBOLS` does NOT work for this project — the symbol count exceeds the 65535 .def file limit.
+
+**MSVC dllexport gotchas:** MSVC eagerly instantiates all special member functions for `__declspec(dllexport)` classes. This causes two problems:
+
+1. **Non-copyable members** (e.g., `vector<unique_ptr<T>>`): MSVC tries to generate copy ctor/assign and fails. Fix: explicitly `= delete` copy operations on the class.
+
+2. **Incomplete types in unique_ptr**: MSVC tries to generate the destructor inline, which needs the complete type. Fix: either include the complete type's header, or declare the destructor in the header and define it `= default` in the .cpp where the type is complete.
+
+### Type aliases
+
+Defined in `src/include/fls/common/alias.hpp`:
+- `n_t` = `uint64_t` (counts), `idx_t` = `uint32_t` (indices), `bw_t` = `uint8_t` (bit width)
+- `up<T>` = `unique_ptr<T>`, `sp<T>` = `shared_ptr<T>`
+
+### Test structure
+
+Tests live in `test/src/` with six suites: `dataset_tests`, `expression_tests`, `fls_reader_tests`, `primitive_tests`, `quick_fuzz_tests`, `unit_tests`. All use GoogleTest. On MSVC, a `msvc_heap_guard` object library handles SEH guard-page exceptions that would otherwise cause spurious test failures.
+
+## Code Style
+
+- `.clang-tidy` is strict: `WarningsAsErrors: '*'` — all warnings are errors
+- Types: `CamelCase`. Functions: `aNy_CasE`. Members: `lower_case` (private: `m_` prefix). Constants: `UPPER_CASE`. Typedefs: `lower_case` with `_t` suffix
+- Tabs for indentation, 120-column limit
+- PRs target `dev` branch