rust-pgn-reader-python-binding 4.2.0__cp315-cp315-musllinux_1_2_armv7l.whl

rust_pgn_reader_python_binding/__init__.py

@@ -0,0 +1,5 @@
+from .rust_pgn_reader_python_binding import *
+
+__doc__ = rust_pgn_reader_python_binding.__doc__
+if hasattr(rust_pgn_reader_python_binding, "__all__"):
+    __all__ = rust_pgn_reader_python_binding.__all__

rust_pgn_reader_python_binding/__init__.pyi

@@ -0,0 +1,393 @@
+from typing import List, Optional, Tuple, Dict, Iterator, Union, overload
+import pyarrow
+import numpy as np
+import numpy.typing as npt
+from numpy.typing import NDArray
+
+class PyGameView:
+    """Zero-copy view into a single game's data within a ParsedGames result.
+
+    Board indexing note: Boards use square indexing (a1=0, h8=63).
+    To convert to rank/file:
+        rank = square // 8
+        file = square % 8
+    """
+
+    def __len__(self) -> int:
+        """Number of moves in this game."""
+        ...
+
+    @property
+    def num_positions(self) -> int:
+        """Number of positions recorded for this game."""
+        ...
+
+    # === Board state views ===
+
+    @property
+    def boards(self) -> NDArray[np.uint8]:
+        """Board positions, shape (num_positions, 8, 8)."""
+        ...
+
+    @property
+    def initial_board(self) -> NDArray[np.uint8]:
+        """Initial board position, shape (8, 8)."""
+        ...
+
+    @property
+    def final_board(self) -> NDArray[np.uint8]:
+        """Final board position, shape (8, 8)."""
+        ...
+
+    @property
+    def castling(self) -> NDArray[np.bool_]:
+        """Castling rights [K,Q,k,q], shape (num_positions, 4)."""
+        ...
+
+    @property
+    def en_passant(self) -> NDArray[np.int8]:
+        """En passant file (-1 if none), shape (num_positions,)."""
+        ...
+
+    @property
+    def halfmove_clock(self) -> NDArray[np.uint8]:
+        """Halfmove clock, shape (num_positions,)."""
+        ...
+
+    @property
+    def turn(self) -> NDArray[np.bool_]:
+        """Side to move (True=white), shape (num_positions,)."""
+        ...
+
+    # === Move views ===
+
+    @property
+    def from_squares(self) -> NDArray[np.uint8]:
+        """From squares, shape (num_moves,)."""
+        ...
+
+    @property
+    def to_squares(self) -> NDArray[np.uint8]:
+        """To squares, shape (num_moves,)."""
+        ...
+
+    @property
+    def promotions(self) -> NDArray[np.int8]:
+        """Promotions (-1=none, 2=N, 3=B, 4=R, 5=Q), shape (num_moves,)."""
+        ...
+
+    @property
+    def clocks(self) -> NDArray[np.float32]:
+        """Clock times in seconds (NaN if missing), shape (num_moves,)."""
+        ...
+
+    @property
+    def evals(self) -> NDArray[np.float32]:
+        """Engine evals (NaN if missing), shape (num_moves,)."""
+        ...
+
+    # === Per-game metadata ===
+
+    @property
+    def headers(self) -> Dict[str, str]:
+        """Raw PGN headers as dict."""
+        ...
+
+    @property
+    def outcome(self) -> Optional[str]:
+        """Game outcome from movetext: 'White', 'Black', 'Draw', 'Unknown', or None."""
+        ...
+
+    @property
+    def parse_error(self) -> Optional[str]:
+        """Parse error message if the game failed to parse, or None if valid."""
+        ...
+
+    @property
+    def is_checkmate(self) -> bool:
+        """Final position is checkmate."""
+        ...
+
+    @property
+    def is_stalemate(self) -> bool:
+        """Final position is stalemate."""
+        ...
+
+    @property
+    def is_insufficient(self) -> Tuple[bool, bool]:
+        """Insufficient material (white, black)."""
+        ...
+
+    @property
+    def legal_move_count(self) -> int:
+        """Legal move count in final position."""
+        ...
+
+    @property
+    def is_valid(self) -> bool:
+        """Whether game parsed successfully."""
+        ...
+
+    @property
+    def is_game_over(self) -> bool:
+        """Whether the game is over (checkmate, stalemate, or both sides insufficient)."""
+        ...
+
+    @property
+    def comments(self) -> List[Optional[str]]:
+        """Raw text comments per move (only populated when store_comments=True)."""
+        ...
+
+    @property
+    def legal_moves(self) -> List[List[Tuple[int, int, int]]]:
+        """Legal moves at each position (only populated when store_legal_moves=True).
+        Each entry is a list of (from_square, to_square, promotion) tuples."""
+        ...
+
+    # === Convenience methods ===
+
+    def move_uci(self, move_idx: int) -> str:
+        """Get UCI string for move at index."""
+        ...
+
+    def moves_uci(self) -> List[str]:
+        """Get all moves as UCI strings."""
+        ...
+
+    def __repr__(self) -> str: ...
+
+class ParsedGamesIter:
+    """Iterator over games in a ParsedGames result."""
+
+    def __iter__(self) -> "ParsedGamesIter": ...
+    def __next__(self) -> PyGameView: ...
+
+class PyChunkView:
+    """View into a single chunk's raw numpy arrays.
+
+    Access via ``parsed_games.chunks[i]``. Each chunk corresponds to one
+    parsing thread's output. Use this for advanced access patterns like
+    manual concatenation or custom batching.
+    """
+
+    @property
+    def num_games(self) -> int: ...
+    @property
+    def num_moves(self) -> int: ...
+    @property
+    def num_positions(self) -> int: ...
+    @property
+    def boards(self) -> NDArray[np.uint8]:
+        """Board positions, shape (N_positions, 8, 8), dtype uint8."""
+        ...
+    @property
+    def castling(self) -> NDArray[np.bool_]:
+        """Castling rights [K,Q,k,q], shape (N_positions, 4), dtype bool."""
+        ...
+    @property
+    def en_passant(self) -> NDArray[np.int8]: ...
+    @property
+    def halfmove_clock(self) -> NDArray[np.uint8]: ...
+    @property
+    def turn(self) -> NDArray[np.bool_]: ...
+    @property
+    def from_squares(self) -> NDArray[np.uint8]: ...
+    @property
+    def to_squares(self) -> NDArray[np.uint8]: ...
+    @property
+    def promotions(self) -> NDArray[np.int8]: ...
+    @property
+    def clocks(self) -> NDArray[np.float32]: ...
+    @property
+    def evals(self) -> NDArray[np.float32]: ...
+    @property
+    def move_offsets(self) -> NDArray[np.uint32]: ...
+    @property
+    def position_offsets(self) -> NDArray[np.uint32]: ...
+    @property
+    def is_checkmate(self) -> NDArray[np.bool_]: ...
+    @property
+    def is_stalemate(self) -> NDArray[np.bool_]: ...
+    @property
+    def is_insufficient(self) -> NDArray[np.bool_]: ...
+    @property
+    def legal_move_count(self) -> NDArray[np.uint16]: ...
+    @property
+    def valid(self) -> NDArray[np.bool_]: ...
+    @property
+    def headers(self) -> List[Dict[str, str]]: ...
+    @property
+    def outcome(self) -> List[Optional[str]]: ...
+    @property
+    def parse_errors(self) -> List[Optional[str]]: ...
+    @property
+    def comments(self) -> List[Optional[str]]: ...
+    @property
+    def legal_move_from_squares(self) -> NDArray[np.uint8]: ...
+    @property
+    def legal_move_to_squares(self) -> NDArray[np.uint8]: ...
+    @property
+    def legal_move_promotions(self) -> NDArray[np.int8]: ...
+    @property
+    def legal_move_offsets(self) -> NDArray[np.uint32]: ...
+    def __repr__(self) -> str: ...
+
+class ParsedGames:
+    """Chunked container for parsed chess games, optimized for ML training.
+
+    Internally stores data in multiple chunks (one per parsing thread) to
+    avoid the cost of merging. Per-game access is O(log(num_chunks)) via
+    binary search on precomputed boundaries.
+
+    Board layout:
+        Boards use square indexing: a1=0, b1=1, ..., h8=63
+        Piece encoding: 0=empty, 1-6=white PNBRQK, 7-12=black pnbrqk
+    """
+
+    # === Computed properties ===
+
+    @property
+    def num_games(self) -> int:
+        """Number of games in the result."""
+        ...
+
+    @property
+    def num_moves(self) -> int:
+        """Total number of moves across all games."""
+        ...
+
+    @property
+    def num_positions(self) -> int:
+        """Total number of board positions recorded."""
+        ...
+
+    @property
+    def num_chunks(self) -> int:
+        """Number of internal chunks."""
+        ...
+
+    # === Escape hatch: raw chunk access ===
+
+    @property
+    def chunks(self) -> List[PyChunkView]:
+        """Access raw per-chunk data.
+
+        Each chunk corresponds to one parsing thread's output. Use this
+        for advanced access patterns like manual concatenation.
+        """
+        ...
+
+    # === Sequence protocol ===
+
+    def __len__(self) -> int:
+        """Number of games in the result."""
+        ...
+
+    @overload
+    def __getitem__(self, idx: int) -> PyGameView: ...
+    @overload
+    def __getitem__(self, idx: slice) -> List[PyGameView]: ...
+    def __getitem__(
+        self, idx: Union[int, slice]
+    ) -> Union[PyGameView, List[PyGameView]]:
+        """Access game(s) by index or slice."""
+        ...
+
+    def __iter__(self) -> ParsedGamesIter:
+        """Iterate over all games."""
+        ...
+
+    def __repr__(self) -> str: ...
+
+    # === Mapping utilities ===
+
+    def position_to_game(self, position_indices: npt.ArrayLike) -> NDArray[np.int64]:
+        """Map global position indices to game indices.
+
+        Useful after shuffling/sampling positions to look up game metadata.
+
+        Args:
+            position_indices: Array of indices into the global position space.
+                Accepts any integer dtype; int64 is optimal (avoids conversion).
+
+        Returns:
+            Array of game indices (same shape as input)
+        """
+        ...
+
+    def move_to_game(self, move_indices: npt.ArrayLike) -> NDArray[np.int64]:
+        """Map global move indices to game indices.
+
+        Args:
+            move_indices: Array of indices into the global move space.
+                Accepts any integer dtype; int64 is optimal (avoids conversion).
+
+        Returns:
+            Array of game indices (same shape as input)
+        """
+        ...
+
+def parse_game(
+    pgn: str,
+    store_comments: bool = False,
+    store_legal_moves: bool = False,
+) -> ParsedGames:
+    """Parse a single PGN game string.
+
+    Convenience wrapper for parsing a single game. Returns a ParsedGames
+    container with one game.
+
+    Args:
+        pgn: PGN game string
+        store_comments: Whether to store raw text comments (default: False)
+        store_legal_moves: Whether to store legal moves at each position (default: False)
+
+    Returns:
+        ParsedGames object containing the parsed game
+    """
+    ...
+
+def parse_games(
+    pgn_chunked_array: pyarrow.ChunkedArray,
+    num_threads: Optional[int] = None,
+    chunk_multiplier: Optional[int] = None,
+    store_comments: bool = False,
+    store_legal_moves: bool = False,
+) -> ParsedGames:
+    """Parse chess games from a PyArrow ChunkedArray into flat NumPy arrays.
+
+    This API is optimized for ML training pipelines, returning flat NumPy arrays
+    that can be efficiently batched and processed.
+
+    Args:
+        pgn_chunked_array: PyArrow ChunkedArray containing PGN strings
+        num_threads: Number of threads for parallel parsing (default: all CPUs)
+        chunk_multiplier: Multiplier for number of chunks (default: 1)
+        store_comments: Whether to store raw text comments (default: False)
+        store_legal_moves: Whether to store legal moves at each position (default: False)
+
+    Returns:
+        ParsedGames object containing flat arrays and iteration support
+    """
+    ...
+
+def parse_games_from_strings(
+    pgns: List[str],
+    num_threads: Optional[int] = None,
+    store_comments: bool = False,
+    store_legal_moves: bool = False,
+) -> ParsedGames:
+    """Parse multiple PGN game strings in parallel.
+
+    Convenience wrapper for when you have a list of strings rather than an Arrow array.
+
+    Args:
+        pgns: List of PGN game strings
+        num_threads: Number of threads for parallel parsing (default: all CPUs)
+        store_comments: Whether to store raw text comments (default: False)
+        store_legal_moves: Whether to store legal moves at each position (default: False)
+
+    Returns:
+        ParsedGames object containing flat arrays and iteration support
+    """
+    ...

rust_pgn_reader_python_binding-4.2.0.dist-info/METADATA

@@ -0,0 +1,99 @@
+Metadata-Version: 2.4
+Name: rust_pgn_reader_python_binding
+Version: 4.2.0
+Classifier: Programming Language :: Rust
+Classifier: Programming Language :: Python :: Implementation :: CPython
+Classifier: Programming Language :: Python :: Implementation :: PyPy
+Requires-Dist: numpy
+Requires-Dist: pyarrow
+Summary: A fast chess PGN & parquet parser
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
+Project-URL: Homepage, https://github.com/vladkvit/rust_pgn_reader_python_binding
+Project-URL: Repository, https://github.com/vladkvit/rust_pgn_reader_python_binding.git
+
+# rust_pgn_reader_python_binding
+## Fast PGN parsing bindings for Python
+This project adds Python bindings to [rust-pgn-reader](https://github.com/niklasf/rust-pgn-reader). In addition, it also parses and extracts [%clk ..] and [%eval ..] tags from comments.
+
+## Installing
+`pip install rust_pgn_reader_python_binding`
+
+## API
+
+Three entry points are available:
+
+- `parse_game(pgn)` - Parse a single PGN string
+- `parse_games(chunked_array)` - Parse games from a PyArrow ChunkedArray (multithreaded)
+- `parse_games_from_strings(pgns)` - Parse a list of PGN strings (multithreaded)
+
+All return a `ParsedGames` container with flat NumPy arrays, supporting:
+- Indexing (`result[i]`), slicing (`result[1:3]`), and iteration (`for game in result`)
+- Per-game views (`PyGameView`) with zero-copy array slices
+- Position-to-game and move-to-game mapping for ML workflows
+- Optional comment storage (`store_comments=True`)
+- Optional legal move storage (`store_legal_moves=True`)
+
+## Benchmarks
+Below are some benchmarks on Lichess's 2013-07 chess games (293,459 games) on a 7800X3D.
+
+| Parser                                                                     | File format | Time   |
+|----------------------------------------------------------------------------|-------------|--------|
+| **rust_pgn_reader_python_binding, parse_games (multithreaded)**            | parquet     | 0.14s  |
+| rust_pgn_reader_python_binding, parse_games (singlethreaded)               | parquet     | 1.3s   |
+| rust_pgn_reader_python_binding, parse_games_from_strings (multithreaded)   | PGN         | 0.41s  |
+| [chess-library](https://github.com/Disservin/chess-library)                | PGN         | 2s     |
+| [rust-pgn-reader](https://github.com/niklasf/rust-pgn-reader/tree/master)  | PGN         | 1s     |
+| [python-chess](https://github.com/niklasf/python-chess)                    | PGN         | 3+ min |
+
+To replicate, download `2013-07-train-00000-of-00001.parquet` and then run:
+
+`python src/bench_parse_games.py` (recommended — multithreaded parse_games via Arrow)
+
+`python src/bench_parse_games_singlethreaded.py` (singlethreaded parse_games via Arrow)
+
+`python src/bench_parse_pgn.py` (multithreaded .pgn parsing)
+
+`python src/bench_data_access.py 2013-07-train-00000-of-00001.parquet` (parsing + data access + memory)
+
+## Building
+`maturin develop`
+
+`maturin develop --release`
+
+For a more thorough tutorial, follow https://lukesalamone.github.io/posts/how-to-create-rust-python-bindings/
+
+## Profiling
+`samply record --rate 10000 python src\bench_parse_games.py`
+`py-spy record -s -F -f speedscope --output profile.speedscope -- python ./src/bench_parse_games.py`
+
+Linux/WSL-only:
+`py-spy record -s -F -n -f speedscope --output profile.speedscope -- python ./src/bench_parse_games.py`
+
+## Testing
+`cargo test`
+
+`python -m unittest src/test.py`
+
+## Type stubs
+We have a manual `rust_pgn_reader_python_binding.pyi` python type stub.
+To keep it in sync with the Rust bindings, run:
+
+`python tools/check_stubs.py`
+
+This regenerates the stub via `maturin generate-stubs` (PyO3
+`experimental-inspect`, behind the `stubgen` Cargo feature) and diffs the
+symbol surface (module functions, classes, methods/properties, but not
+annotations) against the hand-written file. CI runs the same check in the
+`stubs` job.
+
+
+## Further performance squeezing:
+- Interleaved parquet read + parse
+- Opening cache. For positions that occur plenty of times, a `(zobrist(pos), san_token)` → `Move`
+  cache could skip `san_resolver` + legality on cache hits.
+- crude move-count estimator from movetext byte length for buffer preallocation (bytes-per-ply ≈ 6-6.5, over-estimate
+  is relatively cheap, under-estimate forces realloc copies). We currently assume 70 moves per game. Since we have long chunks (not per-game),
+  long game moves eat up the slack from short games. Wont' help on Lichess games (we're already doing quite well). But if going with a 
+  different / unknown corpus, 70 will be off (different game lengths, comments). But comments will also throw off our byte-based move count estimator.
+

rust_pgn_reader_python_binding-4.2.0.dist-info/RECORD

@@ -0,0 +1,9 @@
+rust_pgn_reader_python_binding/__init__.py,sha256=lhSdJaKZEMrZVdzPOP73MTNFcg6hOIbqlkQ070gWYAQ,203
+rust_pgn_reader_python_binding/__init__.pyi,sha256=reLZezmDzhgf918Nz3BQAgiv8PI1brq0UI6uzVdlsJk,11542
+rust_pgn_reader_python_binding/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+rust_pgn_reader_python_binding/rust_pgn_reader_python_binding.cpython-315-arm-linux-musleabihf.so,sha256=OegcECjiDIBZkc_hodHt2CypLWm3ycoFfRH_5argmo0,16626733
+rust_pgn_reader_python_binding-4.2.0.dist-info/METADATA,sha256=Ba9f6gB_q4GCpn9AUNAmvNDEC1F3hRFHFa9zWyfBEao,4689
+rust_pgn_reader_python_binding-4.2.0.dist-info/WHEEL,sha256=7vKPdusLGuyoFe-7QFlLo7up40KfuBKCpQAnzbdKWtc,108
+rust_pgn_reader_python_binding-4.2.0.dist-info/sboms/rust_pgn_reader_python_binding.cyclonedx.json,sha256=2szTLTDLkhk3OUAP_65bHm94Wl9WEXwSRfczDWHKhqQ,102729
+rust_pgn_reader_python_binding.libs/libgcc_s-262c4f60.so.1,sha256=fA1fr7_M_sZmVSS8yaamft5FzaZdrq_bTxQOPu00OFE,2811941
+rust_pgn_reader_python_binding-4.2.0.dist-info/RECORD,,

rust_pgn_reader_python_binding-4.2.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: maturin (1.16.0)
+Root-Is-Purelib: false
+Tag: cp315-cp315-musllinux_1_2_armv7l