diff --git a/python/pyarrow/interchange/__init__.py b/python/pyarrow/interchange/__init__.py new file mode 100644 index 00000000000..13a83393a91 --- /dev/null +++ b/python/pyarrow/interchange/__init__.py @@ -0,0 +1,16 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. diff --git a/python/pyarrow/interchange/buffer.py b/python/pyarrow/interchange/buffer.py new file mode 100644 index 00000000000..9f30f2b99e3 --- /dev/null +++ b/python/pyarrow/interchange/buffer.py @@ -0,0 +1,103 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations +import enum + +import pyarrow as pa + + +class DlpackDeviceType(enum.IntEnum): + """Integer enum for device type codes matching DLPack.""" + + CPU = 1 + CUDA = 2 + CPU_PINNED = 3 + OPENCL = 4 + VULKAN = 7 + METAL = 8 + VPI = 9 + ROCM = 10 + + +class _PyArrowBuffer: + """ + Data in the buffer is guaranteed to be contiguous in memory. + Note that there is no dtype attribute present, a buffer can be thought of + as simply a block of memory. However, if the column that the buffer is + attached to has a dtype that's supported by DLPack and ``__dlpack__`` is + implemented, then that dtype information will be contained in the return + value from ``__dlpack__``. + This distinction is useful to support both data exchange via DLPack on a + buffer and (b) dtypes like variable-length strings which do not have a + fixed number of bytes per element. + """ + + def __init__(self, x: pa.Buffer, allow_copy: bool = True) -> None: + """ + Handle PyArrow Buffers. + """ + self._x = x + + @property + def bufsize(self) -> int: + """ + Buffer size in bytes. + """ + return self._x.size + + @property + def ptr(self) -> int: + """ + Pointer to start of the buffer as an integer. + """ + return self._x.address + + def __dlpack__(self): + """ + Produce DLPack capsule (see array API standard). + Raises: + - TypeError : if the buffer contains unsupported dtypes. + - NotImplementedError : if DLPack support is not implemented + Useful to have to connect to array libraries. Support optional because + it's not completely trivial to implement for a Python-only library. + """ + raise NotImplementedError("__dlpack__") + + def __dlpack_device__(self) -> tuple[DlpackDeviceType, int | None]: + """ + Device type and device ID for where the data in the buffer resides. + Uses device type codes matching DLPack. + Note: must be implemented even if ``__dlpack__`` is not. + """ + if self._x.is_cpu: + return (DlpackDeviceType.CPU, None) + else: + raise NotImplementedError("__dlpack_device__") + + def __repr__(self) -> str: + return ( + "PyArrowBuffer(" + + str( + { + "bufsize": self.bufsize, + "ptr": self.ptr, + "device": self.__dlpack_device__()[0].name, + } + ) + + ")" + ) diff --git a/python/pyarrow/interchange/column.py b/python/pyarrow/interchange/column.py new file mode 100644 index 00000000000..a5c9d4a2833 --- /dev/null +++ b/python/pyarrow/interchange/column.py @@ -0,0 +1,514 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations + +import enum +from typing import ( + Any, + Dict, + Iterable, + Optional, + Tuple, +) + +import sys +if sys.version_info >= (3, 8): + from typing import TypedDict +else: + from typing_extensions import TypedDict + +import pyarrow as pa +import pyarrow.compute as pc +from pyarrow.interchange.buffer import _PyArrowBuffer + + +class DtypeKind(enum.IntEnum): + """ + Integer enum for data types. + Attributes + ---------- + INT : int + Matches to signed integer data type. + UINT : int + Matches to unsigned integer data type. + FLOAT : int + Matches to floating point data type. + BOOL : int + Matches to boolean data type. + STRING : int + Matches to string data type (UTF-8 encoded). + DATETIME : int + Matches to datetime data type. + CATEGORICAL : int + Matches to categorical data type. + """ + + INT = 0 + UINT = 1 + FLOAT = 2 + BOOL = 20 + STRING = 21 # UTF-8 + DATETIME = 22 + CATEGORICAL = 23 + + +Dtype = Tuple[DtypeKind, int, str, str] # see Column.dtype + + +_PYARROW_KINDS = { + pa.int8(): (DtypeKind.INT, "c"), + pa.int16(): (DtypeKind.INT, "s"), + pa.int32(): (DtypeKind.INT, "i"), + pa.int64(): (DtypeKind.INT, "l"), + pa.uint8(): (DtypeKind.UINT, "C"), + pa.uint16(): (DtypeKind.UINT, "S"), + pa.uint32(): (DtypeKind.UINT, "I"), + pa.uint64(): (DtypeKind.UINT, "L"), + pa.float16(): (DtypeKind.FLOAT, "e"), + pa.float32(): (DtypeKind.FLOAT, "f"), + pa.float64(): (DtypeKind.FLOAT, "g"), + pa.bool_(): (DtypeKind.BOOL, "b"), + pa.string(): (DtypeKind.STRING, "u"), # utf-8 + pa.large_string(): (DtypeKind.STRING, "U"), +} + + +class ColumnNullType(enum.IntEnum): + """ + Integer enum for null type representation. + Attributes + ---------- + NON_NULLABLE : int + Non-nullable column. + USE_NAN : int + Use explicit float NaN value. + USE_SENTINEL : int + Sentinel value besides NaN. + USE_BITMASK : int + The bit is set/unset representing a null on a certain position. + USE_BYTEMASK : int + The byte is set/unset representing a null on a certain position. + """ + + NON_NULLABLE = 0 + USE_NAN = 1 + USE_SENTINEL = 2 + USE_BITMASK = 3 + USE_BYTEMASK = 4 + + +class ColumnBuffers(TypedDict): + # first element is a buffer containing the column data; + # second element is the data buffer's associated dtype + data: Tuple[_PyArrowBuffer, Dtype] + + # first element is a buffer containing mask values indicating missing data; + # second element is the mask value buffer's associated dtype. + # None if the null representation is not a bit or byte mask + validity: Optional[Tuple[_PyArrowBuffer, Dtype]] + + # first element is a buffer containing the offset values for + # variable-size binary data (e.g., variable-length strings); + # second element is the offsets buffer's associated dtype. + # None if the data buffer does not have an associated offsets buffer + offsets: Optional[Tuple[_PyArrowBuffer, Dtype]] + + +class CategoricalDescription(TypedDict): + # whether the ordering of dictionary indices is semantically meaningful + is_ordered: bool + # whether a dictionary-style mapping of categorical values to other objects + # exists + is_dictionary: bool + # Python-level only (e.g. ``{int: str}``). + # None if not a dictionary-style categorical. + # categories: Optional[Column] + + +class Endianness: + """Enum indicating the byte-order of a data-type.""" + + LITTLE = "<" + BIG = ">" + NATIVE = "=" + NA = "|" + + +class NoBufferPresent(Exception): + """Exception to signal that there is no requested buffer.""" + + +class _PyArrowColumn: + """ + A column object, with only the methods and properties required by the + interchange protocol defined. + A column can contain one or more chunks. Each chunk can contain up to three + buffers - a data buffer, a mask buffer (depending on null representation), + and an offsets buffer (if variable-size binary; e.g., variable-length + strings). + TBD: Arrow has a separate "null" dtype, and has no separate mask concept. + Instead, it seems to use "children" for both columns with a bit mask, + and for nested dtypes. Unclear whether this is elegant or confusing. + This design requires checking the null representation explicitly. + The Arrow design requires checking: + 1. the ARROW_FLAG_NULLABLE (for sentinel values) + 2. if a column has two children, combined with one of those children + having a null dtype. + Making the mask concept explicit seems useful. One null dtype would + not be enough to cover both bit and byte masks, so that would mean + even more checking if we did it the Arrow way. + TBD: there's also the "chunk" concept here, which is implicit in Arrow as + multiple buffers per array (= column here). Semantically it may make + sense to have both: chunks were meant for example for lazy evaluation + of data which doesn't fit in memory, while multiple buffers per column + could also come from doing a selection operation on a single + contiguous buffer. + Given these concepts, one would expect chunks to be all of the same + size (say a 10,000 row dataframe could have 10 chunks of 1,000 rows), + while multiple buffers could have data-dependent lengths. Not an issue + in pandas if one column is backed by a single NumPy array, but in + Arrow it seems possible. + Are multiple chunks *and* multiple buffers per column necessary for + the purposes of this interchange protocol, or must producers either + reuse the chunk concept for this or copy the data? + Note: this Column object can only be produced by ``__dataframe__``, so + doesn't need its own version or ``__column__`` protocol. + """ + + def __init__( + self, column: pa.Array | pa.ChunkedArray, allow_copy: bool = True + ) -> None: + """ + Handles PyArrow Arrays and ChunkedArrays. + """ + # Store the column as a private attribute + self._col = column + self._allow_copy = allow_copy + + @property + def size(self) -> int: + """ + Size of the column, in elements. + Corresponds to DataFrame.num_rows() if column is a single chunk; + equal to size of this current chunk otherwise. + Is a method rather than a property because it may cause a (potentially + expensive) computation for some dataframe implementations. + """ + return len(self._col) + + @property + def offset(self) -> int: + """ + Offset of first element. + May be > 0 if using chunks; for example for a column with N chunks of + equal size M (only the last chunk may be shorter), + ``offset = n * M``, ``n = 0 .. N-1``. + """ + if isinstance(self._col, pa.Array): + return self._col.offset + else: + # ChunkedArray gets copied with `combine_chunks` so the offset will + # always be 0 + return 0 + + @property + def dtype(self) -> Tuple[DtypeKind, int, str, str]: + """ + Dtype description as a tuple ``(kind, bit-width, format string, + endianness)``. + Bit-width : the number of bits as an integer + Format string : data type description format string in Apache Arrow C + Data Interface format. + Endianness : current only native endianness (``=``) is supported + Notes: + - Kind specifiers are aligned with DLPack where possible (hence the + jump to 20, leave enough room for future extension) + - Masks must be specified as boolean with either bit width 1 (for + bit masks) or 8 (for byte masks). + - Dtype width in bits was preferred over bytes + - Endianness isn't too useful, but included now in case in the + future we need to support non-native endianness + - Went with Apache Arrow format strings over NumPy format strings + because they're more complete from a dataframe perspective + - Format strings are mostly useful for datetime specification, and + for categoricals. + - For categoricals, the format string describes the type of the + categorical in the data buffer. In case of a separate encoding of + the categorical (e.g. an integer to string mapping), this can + be derived from ``self.describe_categorical``. + - Data types not included: complex, Arrow-style null, binary, + decimal, and nested (list, struct, map, union) dtypes. + """ + dtype = self._col.type + try: + bit_width = dtype.bit_width + except ValueError: # in case of a variable-length strings + bit_width = 8 + + if pa.types.is_timestamp(dtype): + kind = DtypeKind.DATETIME + ts = dtype.unit[0] + tz = dtype.tz if dtype.tz else "" + f_string = "ts{ts}:{tz}".format(ts=ts, tz=tz) + return kind, bit_width, f_string, Endianness.NATIVE + elif pa.types.is_dictionary(dtype): + kind = DtypeKind.CATEGORICAL + f_string = "L" + return kind, bit_width, f_string, Endianness.NATIVE + else: + return self._dtype_from_arrowdtype(dtype, bit_width) + + def _dtype_from_arrowdtype( + self, dtype, bit_width + ) -> Tuple[DtypeKind, int, str, str]: + """ + See `self.dtype` for details. + """ + # Note: 'c' (complex) not handled yet (not in array spec v1). + # 'b', 'B' (bytes), 'S', 'a', (old-style string) 'V' (void) + # not handled datetime and timedelta both map to datetime + # (is timedelta handled?) + + kind, f_string = _PYARROW_KINDS.get(dtype, (None, None)) + if kind is None: + raise ValueError( + f"Data type {dtype} not supported by interchange protocol") + + return kind, bit_width, f_string, Endianness.NATIVE + + @property + def describe_categorical(self) -> CategoricalDescription: + """ + If the dtype is categorical, there are two options: + - There are only values in the data buffer. + - There is a separate non-categorical Column encoding categorical + values. + Raises TypeError if the dtype is not categorical + Returns the dictionary with description on how to interpret the + data buffer: + - "is_ordered" : bool, whether the ordering of dictionary indices + is semantically meaningful. + - "is_dictionary" : bool, whether a mapping of + categorical values to other objects exists + - "categories" : Column representing the (implicit) mapping of + indices to category values (e.g. an array of + cat1, cat2, ...). None if not a dictionary-style + categorical. + TBD: are there any other in-memory representations that are needed? + """ + if isinstance(self._col, pa.ChunkedArray): + arr = self._col.combine_chunks() + else: + arr = self._col + + if not pa.types.is_dictionary(arr.type): + raise TypeError( + "describe_categorical only works on a column with " + "categorical dtype!" + ) + + return { + "is_ordered": self._col.type.ordered, + "is_dictionary": True, + "categories": _PyArrowColumn(arr.dictionary), + } + + @property + def describe_null(self) -> Tuple[ColumnNullType, Any]: + """ + Return the missing value (or "null") representation the column dtype + uses, as a tuple ``(kind, value)``. + Value : if kind is "sentinel value", the actual value. If kind is a bit + mask or a byte mask, the value (0 or 1) indicating a missing value. + None otherwise. + """ + # In case of no missing values, we need to set ColumnNullType to + # non nullable as in the current __dataframe__ protocol bit/byte masks + # can not be None + if self.null_count == 0: + return ColumnNullType.NON_NULLABLE, None + else: + return ColumnNullType.USE_BITMASK, 0 + + @property + def null_count(self) -> int: + """ + Number of null elements, if known. + Note: Arrow uses -1 to indicate "unknown", but None seems cleaner. + """ + return self._col.null_count + + @property + def metadata(self) -> Dict[str, Any]: + """ + The metadata for the column. See `DataFrame.metadata` for more details. + """ + pass + + def num_chunks(self) -> int: + """ + Return the number of chunks the column consists of. + """ + if isinstance(self._col, pa.Array): + n_chunks = 1 + else: + n_chunks = self._col.num_chunks + return n_chunks + + def get_chunks( + self, n_chunks: Optional[int] = None + ) -> Iterable[_PyArrowColumn]: + """ + Return an iterator yielding the chunks. + See `DataFrame.get_chunks` for details on ``n_chunks``. + """ + if n_chunks and n_chunks > 1: + chunk_size = self.size // n_chunks + if self.size % n_chunks != 0: + chunk_size += 1 + + if isinstance(self._col, pa.ChunkedArray): + array = self._col.combine_chunks() + else: + array = self._col + + i = 0 + for start in range(0, chunk_size * n_chunks, chunk_size): + yield _PyArrowColumn( + array.slice(start, chunk_size), self._allow_copy + ) + i += 1 + + elif isinstance(self._col, pa.ChunkedArray): + return [ + _PyArrowColumn(chunk, self._allow_copy) + for chunk in self._col.chunks + ] + else: + yield self + + def get_buffers(self) -> ColumnBuffers: + """ + Return a dictionary containing the underlying buffers. + The returned dictionary has the following contents: + - "data": a two-element tuple whose first element is a buffer + containing the data and whose second element is the data + buffer's associated dtype. + - "validity": a two-element tuple whose first element is a buffer + containing mask values indicating missing data and + whose second element is the mask value buffer's + associated dtype. None if the null representation is + not a bit or byte mask. + - "offsets": a two-element tuple whose first element is a buffer + containing the offset values for variable-size binary + data (e.g., variable-length strings) and whose second + element is the offsets buffer's associated dtype. None + if the data buffer does not have an associated offsets + buffer. + """ + buffers: ColumnBuffers = { + "data": self._get_data_buffer(), + "validity": None, + "offsets": None, + } + + try: + buffers["validity"] = self._get_validity_buffer() + except NoBufferPresent: + pass + + try: + buffers["offsets"] = self._get_offsets_buffer() + except NoBufferPresent: + pass + + return buffers + + def _get_data_buffer( + self, + ) -> Tuple[_PyArrowBuffer, Any]: # Any is for self.dtype tuple + """ + Return the buffer containing the data and the buffer's + associated dtype. + """ + if isinstance(self._col, pa.ChunkedArray): + array = self._col.combine_chunks() + else: + array = self._col + dtype = self.dtype + + # In case of boolean arrays, cast to uint8 array + # as bit packed buffers are not supported + if pa.types.is_boolean(array.type): + array = pc.cast(array, pa.uint8()) + dtype = _PyArrowColumn(array).dtype + # In case of dictionary arrays, use indices + # to define a buffer, codes are transferred through + # describe_categorical() + if pa.types.is_dictionary(array.type): + array = array.indices + dtype = _PyArrowColumn(array).dtype + + n = len(array.buffers()) + if n == 2: + return _PyArrowBuffer(array.buffers()[1]), dtype + elif n == 3: + return _PyArrowBuffer(array.buffers()[2]), dtype + + def _get_validity_buffer(self) -> Tuple[_PyArrowBuffer, Any]: + """ + Return the buffer containing the mask values indicating missing data + and the buffer's associated dtype. + Raises NoBufferPresent if null representation is not a bit or byte + mask. + """ + # Define the dtype of the returned buffer + dtype = (DtypeKind.BOOL, 1, "b", Endianness.NATIVE) + if isinstance(self._col, pa.ChunkedArray): + array = self._col.combine_chunks() + else: + array = self._col + buff = array.buffers()[0] + if buff: + return _PyArrowBuffer(buff), dtype + else: + raise NoBufferPresent( + "There are no missing values so " + "does not have a separate mask") + + def _get_offsets_buffer(self) -> Tuple[_PyArrowBuffer, Any]: + """ + Return the buffer containing the offset values for variable-size binary + data (e.g., variable-length strings) and the buffer's associated dtype. + Raises NoBufferPresent if the data buffer does not have an associated + offsets buffer. + """ + if isinstance(self._col, pa.ChunkedArray): + array = self._col.combine_chunks() + else: + array = self._col + n = len(array.buffers()) + if n == 2: + raise NoBufferPresent( + "This column has a fixed-length dtype so " + "it does not have an offsets buffer" + ) + elif n == 3: + # Define the dtype of the returned buffer + dtype = (DtypeKind.INT, 32, "i", Endianness.NATIVE) + return _PyArrowBuffer(array.buffers()[1]), dtype diff --git a/python/pyarrow/interchange/dataframe.py b/python/pyarrow/interchange/dataframe.py new file mode 100644 index 00000000000..278ee102ec9 --- /dev/null +++ b/python/pyarrow/interchange/dataframe.py @@ -0,0 +1,190 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations +from typing import ( + Any, + Iterable, + Optional, + Sequence, +) + +import pyarrow as pa + +from pyarrow.interchange.column import _PyArrowColumn + + +class _PyArrowDataFrame: + """ + A data frame class, with only the methods required by the interchange + protocol defined. + A "data frame" represents an ordered collection of named columns. + A column's "name" must be a unique string. + Columns may be accessed by name or by position. + This could be a public data frame class, or an object with the methods and + attributes defined on this DataFrame class could be returned from the + ``__dataframe__`` method of a public data frame class in a library adhering + to the dataframe interchange protocol specification. + """ + + def __init__( + self, df: pa.Table, nan_as_null: bool = False, allow_copy: bool = True + ) -> None: + """ + Constructor - an instance of this (private) class is returned from + `pa.Table.__dataframe__`. + """ + self._df = df + # ``nan_as_null`` is a keyword intended for the consumer to tell the + # producer to overwrite null values in the data with ``NaN`` (or + # ``NaT``). This currently has no effect; once support for nullable + # extension dtypes is added, this value should be propagated to + # columns. + self._nan_as_null = nan_as_null + self._allow_copy = allow_copy + + def __dataframe__( + self, nan_as_null: bool = False, allow_copy: bool = True + ) -> _PyArrowDataFrame: + """ + Construct a new exchange object, potentially changing the parameters. + ``nan_as_null`` is a keyword intended for the consumer to tell the + producer to overwrite null values in the data with ``NaN``. + It is intended for cases where the consumer does not support the bit + mask or byte mask that is the producer's native representation. + ``allow_copy`` is a keyword that defines whether or not the library is + allowed to make a copy of the data. For example, copying data would be + necessary if a library supports strided buffers, given that this + protocol specifies contiguous buffers. + """ + return _PyArrowDataFrame(self._df, nan_as_null, allow_copy) + + @property + def metadata(self) -> dict[str, Any]: + """ + The metadata for the data frame, as a dictionary with string keys. The + contents of `metadata` may be anything, they are meant for a library + to store information that it needs to, e.g., roundtrip losslessly or + for two implementations to share data that is not (yet) part of the + interchange protocol specification. For avoiding collisions with other + entries, please add name the keys with the name of the library + followed by a period and the desired name, e.g, ``pandas.indexcol``. + """ + # The metadata for the data frame, as a dictionary with string keys. + # Add schema metadata here (pandas metadata or custom metadata) + if self._df.schema.metadata: + schema_metadata = {k.decode('utf8'): v.decode('utf8') + for k, v in self._df.schema.metadata.items()} + return schema_metadata + else: + return {} + + def num_columns(self) -> int: + """ + Return the number of columns in the DataFrame. + """ + return self._df.num_columns + + def num_rows(self) -> int: + """ + Return the number of rows in the DataFrame, if available. + """ + return self._df.num_rows + + def num_chunks(self) -> int: + """ + Return the number of chunks the DataFrame consists of. + """ + return self._df.column(0).num_chunks + + def column_names(self) -> Iterable[str]: + """ + Return an iterator yielding the column names. + """ + return self._df.column_names + + def get_column(self, i: int) -> _PyArrowColumn: + """ + Return the column at the indicated position. + """ + return _PyArrowColumn(self._df.column(i), + allow_copy=self._allow_copy) + + def get_column_by_name(self, name: str) -> _PyArrowColumn: + """ + Return the column whose name is the indicated name. + """ + return _PyArrowColumn(self._df.column(name), + allow_copy=self._allow_copy) + + def get_columns(self) -> Iterable[_PyArrowColumn]: + """ + Return an iterator yielding the columns. + """ + return [ + _PyArrowColumn(col, allow_copy=self._allow_copy) + for col in self._df.columns + ] + + def select_columns(self, indices: Sequence[int]) -> _PyArrowDataFrame: + """ + Create a new DataFrame by selecting a subset of columns by index. + """ + return _PyArrowDataFrame( + self._df.select(list(indices)), self._nan_as_null, self._allow_copy + ) + + def select_columns_by_name( + self, names: Sequence[str] + ) -> _PyArrowDataFrame: + """ + Create a new DataFrame by selecting a subset of columns by name. + """ + return _PyArrowDataFrame( + self._df.select(list(names)), self._nan_as_null, self._allow_copy + ) + + def get_chunks( + self, n_chunks: Optional[int] = None + ) -> Iterable[_PyArrowDataFrame]: + """ + Return an iterator yielding the chunks. + By default (None), yields the chunks that the data is stored as by the + producer. If given, ``n_chunks`` must be a multiple of + ``self.num_chunks()``, meaning the producer must subdivide each chunk + before yielding it. + Note that the producer must ensure that all columns are chunked the + same way. + """ + if n_chunks and n_chunks > 1: + chunk_size = self.num_rows() // n_chunks + if self.num_rows() % n_chunks != 0: + chunk_size += 1 + batches = self._df.to_batches(max_chunksize=chunk_size) + # In case when the size of the chunk is such that the resulting + # list is one less chunk then n_chunks -> append an empty chunk + if len(batches) == n_chunks - 1: + batches.append(pa.record_batch([[]], schema=self._df.schema)) + else: + batches = self._df.to_batches() + + iterator_tables = [_PyArrowDataFrame( + pa.Table.from_batches([batch]), self._nan_as_null, self._allow_copy + ) + for batch in batches + ] + return iterator_tables diff --git a/python/pyarrow/interchange/from_dataframe.py b/python/pyarrow/interchange/from_dataframe.py new file mode 100644 index 00000000000..a7746371129 --- /dev/null +++ b/python/pyarrow/interchange/from_dataframe.py @@ -0,0 +1,245 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from __future__ import annotations + +from typing import ( + Any, +) + +from pyarrow.interchange.buffer import _PyArrowBuffer +from pyarrow.interchange.column import ( + _PyArrowColumn, + DtypeKind, +) +from pyarrow.interchange.dataframe import _PyArrowDataFrame + +import numpy as np +import pyarrow as pa + + +# A typing protocol could be added later to let Mypy validate code using +# `from_dataframe` better. +DataFrameObject = Any +ColumnObject = Any +BufferObject = Any + + +def from_dataframe(df: DataFrameObject, allow_copy=True) -> pa.Table: + """ + Build a ``pa.Table`` from any DataFrame supporting the interchange + protocol. + + Parameters + ---------- + df : DataFrameObject + Object supporting the interchange protocol, i.e. `__dataframe__` + method. + allow_copy : bool, default: True + Whether to allow copying the memory to perform the conversion + (if false then zero-copy approach is requested). + Returns + ------- + pa.Table + """ + if isinstance(df, pa.Table): + return df + + if not hasattr(df, "__dataframe__"): + raise ValueError("`df` does not support __dataframe__") + + return _from_dataframe(df.__dataframe__(allow_copy=allow_copy)) + + +def _from_dataframe(df: DataFrameObject, allow_copy=True): + """ + Build a ``pa.Table`` from the DataFrame interchange object. + Parameters + ---------- + df : DataFrameObject + Object supporting the interchange protocol, i.e. `__dataframe__` + method. + allow_copy : bool, default: True + Whether to allow copying the memory to perform the conversion + (if false then zero-copy approach is requested). + Returns + ------- + pa.Table + """ + pass + + +def protocol_df_chunk_to_pyarrow(df: DataFrameObject) -> pa.Table: + """ + Convert interchange protocol chunk to ``pd.DataFrame``. + Parameters + ---------- + df : DataFrameObject + Returns + ------- + pa.Table + """ + # We need a dict of columns here, with each column being a NumPy array + # (at least for now, deal with non-NumPy dtypes later). + columns: dict[str, Any] = {} + buffers = [] # hold on to buffers, keeps memory alive + for name in df.column_names(): + if not isinstance(name, str): + raise ValueError(f"Column {name} is not a string") + if name in columns: + raise ValueError(f"Column {name} is not unique") + col = df.get_column_by_name(name) + dtype = col.dtype[0] + if dtype in ( + DtypeKind.INT, + DtypeKind.UINT, + DtypeKind.FLOAT, + DtypeKind.BOOL, + ): + columns[name], buf = primitive_column_to_ndarray(col) + elif dtype == DtypeKind.CATEGORICAL: + columns[name], buf = categorical_column_to_dictionary(col) + elif dtype == DtypeKind.STRING: + columns[name], buf = string_column_to_ndarray(col) + elif dtype == DtypeKind.DATETIME: + columns[name], buf = datetime_column_to_ndarray(col) + else: + raise NotImplementedError(f"Data type {dtype} not handled yet") + + buffers.append(buf) + + pass + + +def primitive_column_to_array(col: ColumnObject) -> tuple[pa.Array, Any]: + """ + Convert a column holding one of the primitive dtypes to a PyArrow array. + A primitive type is one of: int, uint, float, bool. + Parameters + ---------- + col : ColumnObject + Returns + ------- + tuple + Tuple of pa.Array holding the data and the memory owner object + that keeps the memory alive. + """ + pass + + +def categorical_column_to_dictionary( + col: ColumnObject +) -> tuple[pa.Array, Any]: + """ + Convert a column holding categorical data to a pandas Series. + Parameters + ---------- + col : ColumnObject + Returns + ------- + tuple + Tuple of pa.Array holding the data and the memory owner object + that keeps the memory alive. + """ + pass + + +def string_column_to_array(col: ColumnObject) -> tuple[pa.Array, Any]: + """ + Convert a column holding string data to a NumPy array. + Parameters + ---------- + col : ColumnObject + Returns + ------- + tuple + Tuple of pa.Array holding the data and the memory owner object + that keeps the memory alive. + """ + pass + + +def parse_datetime_format_str(format_str, data): + """Parse datetime `format_str` to interpret the `data`.""" + pass + + +def datetime_column_to_array(col: ColumnObject) -> tuple[pa.Array, Any]: + """ + Convert a column holding DateTime data to a NumPy array. + Parameters + ---------- + col : ColumnObject + Returns + ------- + tuple + Tuple of pa.Array holding the data and the memory owner object + that keeps the memory alive. + """ + pass + + +def buffer_to_array( + buffer: BufferObject, + dtype: tuple[DtypeKind, int, str, str], + offset: int = 0, + length: int | None = None, +) -> pa.Array: + """ + Build a NumPy array from the passed buffer. + Parameters + ---------- + buffer : BufferObject + Buffer to build a PyArrow array from. + dtype : tuple + Data type of the buffer conforming protocol dtypes format. + offset : int, default: 0 + Number of elements to offset from the start of the buffer. + length : int, optional + If the buffer is a bit-mask, specifies a number of bits to read + from the buffer. Has no effect otherwise. + Returns + ------- + pa.Array + + Notes + ----- + The returned array doesn't own the memory. The caller of this function + is responsible for keeping the memory owner object alive as long as + the returned NumPy array is being used. + """ + pass + + +def bitmask_to_bool_array( + bitmask: np.ndarray, mask_length: int, first_byte_offset: int = 0 +) -> pa.Array: + """ + Convert bit-mask to a boolean NumPy array. + Parameters + ---------- + bitmask : np.ndarray[uint8] + NumPy array of uint8 dtype representing the bitmask. + mask_length : int + Number of elements in the mask to interpret. + first_byte_offset : int, default: 0 + Number of elements to offset from the start of the first byte. + Returns + ------- + pa.Array[bool] + """ + pass diff --git a/python/pyarrow/table.pxi b/python/pyarrow/table.pxi index 5c58ae61f19..94d8b6b6487 100644 --- a/python/pyarrow/table.pxi +++ b/python/pyarrow/table.pxi @@ -2809,6 +2809,36 @@ cdef class Table(_PandasConvertible): return self.column(key) + # ---------------------------------------------------------------------- + def __dataframe__(self, nan_as_null: bool = False, allow_copy: bool = True): + """ + Return the dataframe interchange object implementing the interchange protocol. + Parameters + ---------- + nan_as_null : bool, default False + Whether to tell the DataFrame to overwrite null values in the data + with ``NaN`` (or ``NaT``). + allow_copy : bool, default True + Whether to allow memory copying when exporting. If set to False + it would cause non-zero-copy exports to fail. + Returns + ------- + DataFrame interchange object + The object which consuming library can use to ingress the dataframe. + Notes + ----- + Details on the interchange protocol: + https://data-apis.org/dataframe-protocol/latest/index.html + `nan_as_null` currently has no effect; once support for nullable extension + dtypes is added, this value should be propagated to columns. + """ + + from pyarrow.interchange.dataframe import _PyArrowDataFrame + + return _PyArrowDataFrame(self, nan_as_null, allow_copy) + + # ---------------------------------------------------------------------- + def slice(self, offset=0, length=None): """ Compute zero-copy slice of this Table. diff --git a/python/pyarrow/tests/interchange/__init__.py b/python/pyarrow/tests/interchange/__init__.py new file mode 100644 index 00000000000..13a83393a91 --- /dev/null +++ b/python/pyarrow/tests/interchange/__init__.py @@ -0,0 +1,16 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. diff --git a/python/pyarrow/tests/interchange/test_extra.py b/python/pyarrow/tests/interchange/test_extra.py new file mode 100644 index 00000000000..4181b117be6 --- /dev/null +++ b/python/pyarrow/tests/interchange/test_extra.py @@ -0,0 +1,87 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from datetime import datetime as dt +import pyarrow as pa +from pyarrow.vendored.version import Version +import pytest + +from pyarrow.interchange.column import ( + _PyArrowColumn, + ColumnNullType, + DtypeKind, +) + +try: + import pandas as pd + import pandas.testing as tm + from pandas.core.interchange.from_dataframe import from_dataframe +except ImportError: + pass + + +def test_datetime(): + df = pd.DataFrame({"A": [dt(2007, 7, 13), None]}) + table = pa.table(df) + col = table.__dataframe__().get_column_by_name("A") + + assert col.size == 2 + assert col.null_count == 1 + assert col.dtype[0] == DtypeKind.DATETIME + assert col.describe_null == (ColumnNullType.USE_BITMASK, 0) + + +@pytest.mark.parametrize( + ["test_data", "kind"], + [ + (["foo", "bar"], 21), + ([1.5, 2.5, 3.5], 2), + ([1, 2, 3, 4], 0), + ], +) +def test_array_to_pyarrowcolumn(test_data, kind): + arr = pa.array(test_data) + arr_column = _PyArrowColumn(arr) + + assert arr_column._col == arr + assert arr_column.size == len(test_data) + assert arr_column.dtype[0] == kind + assert arr_column.num_chunks() == 1 + assert arr_column.null_count == 0 + assert arr_column.get_buffers()["validity"] is None + assert len(list(arr_column.get_chunks())) == 1 + + for chunk in arr_column.get_chunks(): + assert chunk == arr_column + + +@pytest.mark.pandas +def test_offset_of_sliced_array(): + if Version(pd.__version__) < Version("1.5.0"): + pytest.skip("__dataframe__ added to pandas in 1.5.0") + + arr = pa.array([1, 2, 3, 4]) + arr_sliced = arr.slice(2, 2) + + table = pa.table([arr], names=["arr"]) + table_sliced = pa.table([arr_sliced], names=["arr_sliced"]) + + df = from_dataframe(table) + df_sliced = from_dataframe(table_sliced) + + tm.assert_series_equal(df["arr"][2:4], df_sliced["arr_sliced"], + check_index=False, check_names=False) diff --git a/python/pyarrow/tests/interchange/test_interchange_spec.py b/python/pyarrow/tests/interchange/test_interchange_spec.py new file mode 100644 index 00000000000..425e8f7f95d --- /dev/null +++ b/python/pyarrow/tests/interchange/test_interchange_spec.py @@ -0,0 +1,178 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +import ctypes + +import pyarrow as pa +import pytest + + +@pytest.mark.parametrize( + "test_data", + [ + {"a": ["foo", "bar"], "b": ["baz", "qux"]}, + {"a": [1.5, 2.5, 3.5], "b": [9.2, 10.5, 11.8]}, + {"A": [1, 2, 3, 4], "B": [1, 2, 3, 4]}, + ], +) +def test_only_one_dtype(test_data): + columns = list(test_data.keys()) + table = pa.table(test_data) + df = table.__dataframe__() + + column_size = len(test_data[columns[0]]) + for column in columns: + null_count = df.get_column_by_name(column).null_count + assert null_count == 0 + assert isinstance(null_count, int) + assert df.get_column_by_name(column).size == column_size + assert df.get_column_by_name(column).offset == 0 + + +def test_mixed_dtypes(): + table = pa.table( + { + "a": [1, 2, 3], # dtype kind INT = 0 + "b": [3, 4, 5], # dtype kind INT = 0 + "c": [1.5, 2.5, 3.5], # dtype kind FLOAT = 2 + "d": [9, 10, 11], # dtype kind INT = 0 + "e": [True, False, True], # dtype kind BOOLEAN = 20 + "f": ["a", "", "c"], # dtype kind STRING = 21 + } + ) + df = table.__dataframe__() + # for meanings of dtype[0] see the spec; we cannot import the + # spec here as this file is expected to be vendored *anywhere*; + # values for dtype[0] are explained above + columns = {"a": 0, "b": 0, "c": 2, "d": 0, "e": 20, "f": 21} + + for column, kind in columns.items(): + col = df.get_column_by_name(column) + assert col.null_count == 0 + assert isinstance(col.null_count, int) + assert col.size == 3 + assert col.offset == 0 + + assert col.dtype[0] == kind + + assert df.get_column_by_name("c").dtype[1] == 64 + + +def test_na_float(): + table = pa.table({"a": [1.0, None, 2.0]}) + df = table.__dataframe__() + col = df.get_column_by_name("a") + assert col.null_count == 1 + assert isinstance(col.null_count, int) + + +def test_noncategorical(): + table = pa.table({"a": [1, 2, 3]}) + df = table.__dataframe__() + col = df.get_column_by_name("a") + with pytest.raises(TypeError, match=".*categorical.*"): + col.describe_categorical + + +def test_categorical(): + import pyarrow as pa + arr = ["Mon", "Tue", "Mon", "Wed", "Mon", "Thu", "Fri", "Sat", "Sun"] + table = pa.table( + {"weekday": pa.array(arr).dictionary_encode()} + ) + + col = table.__dataframe__().get_column_by_name("weekday") + categorical = col.describe_categorical + assert isinstance(categorical["is_ordered"], bool) + assert isinstance(categorical["is_dictionary"], bool) + + +def test_dataframe(): + n = pa.chunked_array([[2, 2, 4], [4, 5, 100]]) + a = pa.chunked_array([["Flamingo", "Parrot", "Cow"], + ["Horse", "Brittle stars", "Centipede"]]) + table = pa.table([n, a], names=['n_legs', 'animals']) + df = table.__dataframe__() + + assert df.num_columns() == 2 + assert df.num_rows() == 6 + assert df.num_chunks() == 2 + assert list(df.column_names()) == ['n_legs', 'animals'] + assert list(df.select_columns((1,)).column_names()) == list( + df.select_columns_by_name(("animals",)).column_names() + ) + + +@pytest.mark.parametrize(["size", "n_chunks"], [(10, 3), (12, 3), (12, 5)]) +def test_df_get_chunks(size, n_chunks): + table = pa.table({"x": list(range(size))}) + df = table.__dataframe__() + chunks = list(df.get_chunks(n_chunks)) + assert len(chunks) == n_chunks + assert sum(chunk.num_rows() for chunk in chunks) == size + + +@pytest.mark.parametrize(["size", "n_chunks"], [(10, 3), (12, 3), (12, 5)]) +def test_column_get_chunks(size, n_chunks): + table = pa.table({"x": list(range(size))}) + df = table.__dataframe__() + chunks = list(df.get_column(0).get_chunks(n_chunks)) + assert len(chunks) == n_chunks + assert sum(chunk.size for chunk in chunks) == size + + +def test_get_columns(): + table = pa.table({"a": [0, 1], "b": [2.5, 3.5]}) + df = table.__dataframe__() + for col in df.get_columns(): + assert col.size == 2 + assert col.num_chunks() == 1 + # for meanings of dtype[0] see the spec; we cannot import the + # spec here as this file is expected to be vendored *anywhere* + assert df.get_column(0).dtype[0] == 0 # INT + assert df.get_column(1).dtype[0] == 2 # FLOAT + + +def test_buffer(): + arr = [0, 1, -1] + table = pa.table({"a": arr}) + df = table.__dataframe__() + col = df.get_column(0) + buf = col.get_buffers() + + dataBuf, dataDtype = buf["data"] + + assert dataBuf.bufsize > 0 + assert dataBuf.ptr != 0 + device, _ = dataBuf.__dlpack_device__() + + # for meanings of dtype[0] see the spec; we cannot import the spec + # here as this file is expected to be vendored *anywhere* + assert dataDtype[0] == 0 # INT + + if device == 1: # CPU-only as we're going to directly read memory here + bitwidth = dataDtype[1] + ctype = { + 8: ctypes.c_int8, + 16: ctypes.c_int16, + 32: ctypes.c_int32, + 64: ctypes.c_int64, + }[bitwidth] + + for idx, truth in enumerate(arr): + val = ctype.from_address(dataBuf.ptr + idx * (bitwidth // 8)).value + assert val == truth, f"Buffer at index {idx} mismatch"