Namespaces
namespace	detail
	Copyright 2023, XGBoost Contributors.

Data Structures
class	ArrayAdapter
	Adapter for dense array on host, in Python that's `numpy.ndarray`. More...

class	ArrayAdapterBatch

class	ArrowColumnarBatch

struct	ArrowSchemaImporter

struct	Cache
	Information about the cache including path and page offsets. More...

class	Column

struct	ColumnarMetaInfo

struct	COOTuple

class	CSCAdapter

class	CSCAdapterBatch

class	CSCArrayAdapter
	CSC adapter with support for array interface. More...

class	CSCArrayAdapterBatch

class	CSCPageSource

class	CSRAdapter

class	CSRAdapterBatch

class	CSRArrayAdapter
	Adapter for CSR array on host, in Python that's `scipy.sparse.csr_matrix`. More...

class	CSRArrayAdapterBatch

class	DataIterProxy

class	DataTableAdapter

class	DataTableAdapterBatch

class	DenseAdapter

class	DenseAdapterBatch

class	DMatrixProxy

class	EllpackPageSource

class	ExceHandler

class	FileAdapter
	FileAdapter wraps dmlc::parser to read files and provide access in a common interface. More...

class	FileAdapterBatch

class	FileIterator
	An iterator for implementing external memory support with file inputs. More...

class	GHistIndexRawFormat

class	GradientIndexPageSource

struct	IsValidFunctor

class	IterativeDMatrix
	DMatrix type for `QuantileDMatrix`, the naming `IterativeDMatix` is due to its construction process. More...

class	IteratorAdapter
	Data iterator that takes callback to return data, used in JVM package for accepting data iterator. More...

struct	LabelsCheck

class	PageSourceIncMixIn

class	PrimitiveColumn

class	RecordBatchesIterAdapter

class	SimpleBatchIteratorImpl

class	SimpleDMatrix

class	SingleBatchInternalIter

class	SortedCSCPageSource

class	SparsePageAdapterBatch

class	SparsePageDMatrix
	DMatrix used for external memory. More...

class	SparsePageFormat
	Format specification of various data formats like SparsePage. More...

struct	SparsePageFormatReg
	Registry entry for sparse page format. More...

class	SparsePageRawFormat

class	SparsePageSource

class	SparsePageSourceImpl
	Base class for all page sources. More...

class	TryLockGuard

struct	WeightsCheck

Typedefs
using	ArrowColumnarBatchVec = std::vector< std::unique_ptr< ArrowColumnarBatch > >

Enumerations
enum	ColumnDType : uint8_t { kUnknown , kInt8 , kUInt8 , kInt16 , kUInt16 , kInt32 , kUInt32 , kInt64 , kUInt64 , kFloat , kDouble }

Functions
None	_warn_unused_missing (DataType data, Optional[FloatCompatible] missing)

None	_check_data_shape (DataType data)

bool	_is_scipy_csr (DataType data)

bytes	_array_interface (np.ndarray data)

DataType	transform_scipy_sparse (DataType data, bool is_csr)

DispatchedDataBackendReturnType	_from_scipy_csr (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_scipy_csc (DataType data)

DispatchedDataBackendReturnType	_from_scipy_csc (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_scipy_coo (DataType data)

bool	_is_np_array_like (DataType data)

Tuple[np.ndarray, Optional[NumpyDType]]	_ensure_np_dtype (DataType data, Optional[NumpyDType] dtype)

np.ndarray	_maybe_np_slice (DataType data, Optional[NumpyDType] dtype)

DispatchedDataBackendReturnType	_from_numpy_array (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, DataSplitMode data_split_mode=DataSplitMode.ROW)

bool	_is_pandas_df (DataType data)

bool	_is_modin_df (DataType data)

None	_invalid_dataframe_dtype (DataType data)

Tuple[Optional[FeatureNames], Optional[FeatureTypes]]	pandas_feature_info (DataFrame data, Optional[str] meta, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

bool	is_nullable_dtype (PandasDType dtype)

bool	is_pa_ext_dtype (Any dtype)

bool	is_pa_ext_categorical_dtype (Any dtype)

bool	is_pd_cat_dtype (PandasDType dtype)

bool	is_pd_sparse_dtype (PandasDType dtype)

DataFrame	pandas_cat_null (DataFrame data)

DataFrame	pandas_ext_num_types (DataFrame data)

Tuple[np.ndarray, Optional[FeatureNames], Optional[FeatureTypes]]	_transform_pandas_df (DataFrame data, bool enable_categorical, Optional[FeatureNames] feature_names=None, Optional[FeatureTypes] feature_types=None, Optional[str] meta=None, Optional[NumpyDType] meta_type=None)

DispatchedDataBackendReturnType	_from_pandas_df (DataFrame data, bool enable_categorical, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_pandas_series (DataType data)

None	_meta_from_pandas_series (DataType data, str name, Optional[NumpyDType] dtype, ctypes.c_void_p handle)

bool	_is_modin_series (DataType data)

DispatchedDataBackendReturnType	_from_pandas_series (DataType data, FloatCompatible missing, int nthread, bool enable_categorical, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_dt_df (DataType data)

Tuple[np.ndarray, Optional[FeatureNames], Optional[FeatureTypes]]	_transform_dt_df (DataType data, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, Optional[str] meta=None, Optional[NumpyDType] meta_type=None)

DispatchedDataBackendReturnType	_from_dt_df (DataType data, Optional[FloatCompatible] missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

bool	_is_arrow (DataType data)

Callable	record_batch_data_iter (Iterator data_iter)

DispatchedDataBackendReturnType	_from_arrow (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

bool	_is_cudf_df (DataType data)

bytes	_cudf_array_interfaces (DataType data, list cat_codes)

Tuple[ctypes.c_void_p, list, Optional[FeatureNames], Optional[FeatureTypes]]	_transform_cudf_df (DataType data, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

DispatchedDataBackendReturnType	_from_cudf_df (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

bool	_is_cudf_ser (DataType data)

bool	_is_cupy_array (DataType data)

CupyT	_transform_cupy_array (DataType data)

DispatchedDataBackendReturnType	_from_cupy_array (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_cupy_csr (DataType data)

bool	_is_cupy_csc (DataType data)

bool	_is_dlpack (DataType data)

bool	_transform_dlpack (DataType data)

DispatchedDataBackendReturnType	_from_dlpack (DataType data, FloatCompatible missing, int nthread, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_uri (DataType data)

DispatchedDataBackendReturnType	_from_uri (DataType data, Optional[FloatCompatible] missing, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, DataSplitMode data_split_mode=DataSplitMode.ROW)

bool	_is_list (DataType data)

DispatchedDataBackendReturnType	_from_list (Sequence data, FloatCompatible missing, int n_threads, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_tuple (DataType data)

DispatchedDataBackendReturnType	_from_tuple (Sequence data, FloatCompatible missing, int n_threads, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types)

bool	_is_iter (DataType data)

bool	_has_array_protocol (DataType data)

DataType	_convert_unknown_data (DataType data)

DispatchedDataBackendReturnType	dispatch_data_backend (DataType data, FloatCompatible missing, int threads, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical=False, DataSplitMode data_split_mode=DataSplitMode.ROW)

None	_validate_meta_shape (DataType data, str name)

None	_meta_from_numpy (np.ndarray data, str field, Optional[NumpyDType] dtype, ctypes.c_void_p handle)

None	_meta_from_list (Sequence data, str field, Optional[NumpyDType] dtype, ctypes.c_void_p handle)

None	_meta_from_tuple (Sequence data, str field, Optional[NumpyDType] dtype, ctypes.c_void_p handle)

None	_meta_from_cudf_df (DataType data, str field, ctypes.c_void_p handle)

None	_meta_from_cudf_series (DataType data, str field, ctypes.c_void_p handle)

None	_meta_from_cupy_array (DataType data, str field, ctypes.c_void_p handle)

None	_meta_from_dt (DataType data, str field, Optional[NumpyDType] dtype, ctypes.c_void_p handle)

None	dispatch_meta_backend (DMatrix matrix, DataType data, str name, Optional[NumpyDType] dtype=None)

TransformedData	_proxy_transform (DataType data, Optional[FeatureNames] feature_names, Optional[FeatureTypes] feature_types, bool enable_categorical)

None	dispatch_proxy_set_data (_ProxyDMatrix proxy, DataType data, Optional[list] cat_codes, bool allow_host)

	DMLC_REGISTRY_LINK_TAG (sparse_page_raw_format)

	DMLC_REGISTRY_LINK_TAG (gradient_index_format)

std::string	ValidateFileFormat (std::string const &uri)

	DMLC_REGISTRY_FILE_TAG (gradient_index_format)

	describe ("Raw GHistIndex binary data format.") .set_body([]()

bool	ReadHistogramCuts (common::HistogramCuts cuts, common::AlignedResourceReadStream fi)

std::size_t	WriteHistogramCuts (common::HistogramCuts const &cuts, common::AlignedFileWriteStream *fo)

void	GetCutsFromRef (Context const ctx, std::shared_ptr< DMatrix > ref, bst_feature_t n_features, BatchParam p, common::HistogramCuts p_cuts)
	Get quantile cuts from reference (Quantile)DMatrix.

void	GetCutsFromEllpack (EllpackPage const &page, common::HistogramCuts *cuts)
	Get quantile cuts from ellpack page.

std::shared_ptr< DMatrix >	CreateDMatrixFromProxy (Context const *ctx, std::shared_ptr< DMatrixProxy > proxy, float missing)
	Create a `SimpleDMatrix` instance from a `DMatrixProxy`.

DMatrixProxy *	MakeProxy (DMatrixHandle proxy)

template<bool get_value = true, typename Fn >
decltype(auto)	HostAdapterDispatch (DMatrixProxy const proxy, Fn fn, bool type_error=nullptr)
	Dispatch function call based on input type.

std::string	MakeId (std::string prefix, SparsePageDMatrix *ptr)

std::string	MakeCache (SparsePageDMatrix ptr, std::string format, std::string prefix, std::map< std::string, std::shared_ptr< Cache > > out)

	DMLC_REGISTRY_FILE_TAG (sparse_page_raw_format)

	describe ("Raw binary data format.") .set_body([]()

void	TryDeleteCacheFile (const std::string &file)

void	DevicePush (DMatrixProxy , float, SparsePage )

template<typename T >
SparsePageFormat< T > *	CreatePageFormat (const std::string &name)
	Create sparse page of format.

void	ValidateQueryGroup (std::vector< bst_group_t > const &group_ptr_)

	TEST (FileIterator, Basic)

	TEST (GradientIndex, ExternalMemoryBaseRowID)

	TEST (GradientIndex, FromCategoricalBasic)

	TEST (GradientIndex, FromCategoricalLarge)

	TEST (GradientIndex, PushBatch)

	TEST (GHistIndexPageRawFormat, IO)

	TEST (IterativeDMatrix, Ref)

	TEST (IterativeDMatrix, IsDense)

template<typename Page , typename Iter , typename Cuts >
void	TestRefDMatrix (Context const *ctx, Cuts &&get_cuts)

	TEST (ProxyDMatrix, HostData)

template<typename S >
void	TestSparsePageRawFormat ()

	TEST (SparsePageRawFormat, SparsePage)

	TEST (SparsePageRawFormat, CSCPage)

	TEST (SparsePageRawFormat, SortedCSCPage)

Variables
	DispatchedDataBackendReturnType

str	CAT_T = "c"

dict	_matrix_meta = {"base_margin", "label"}

dict	_pandas_dtype_mapper

dict	pandas_nullable_mapper

dict	pandas_pyarrow_mapper

tuple	_ENABLE_CAT_ERR

constexpr size_t	kAdapterUnknownSize = std::numeric_limits<size_t >::max()
	External data formats should implement an adapter as below.

Detailed Description

Data dispatching for DMatrix.

Function Documentation

◆ _cudf_array_interfaces()

bytes xgboost.data._cudf_array_interfaces	(	DataType	data,
		list	cat_codes
	)

protected

Extract CuDF __cuda_array_interface__.  This is special as it returns a new list
of data and a list of array interfaces.  The data is list of categorical codes that
caller can safely ignore, but have to keep their reference alive until usage of
array interface is finished.

◆ _from_cupy_array()

DispatchedDataBackendReturnType xgboost.data._from_cupy_array	(	DataType	data,
		FloatCompatible	missing,
		int	nthread,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types
	)

protected

Initialize DMatrix from cupy ndarray.

◆ _from_numpy_array()

DispatchedDataBackendReturnType xgboost.data._from_numpy_array	(	DataType	data,
		FloatCompatible	missing,
		int	nthread,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types,
		DataSplitMode	data_split_mode = `DataSplitMode.ROW`
	)

protected

Initialize data from a 2-D numpy matrix.

◆ _from_scipy_csc()

DispatchedDataBackendReturnType xgboost.data._from_scipy_csc	(	DataType	data,
		FloatCompatible	missing,
		int	nthread,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types
	)

protected

Initialize data from a CSC matrix.

◆ _from_scipy_csr()

DispatchedDataBackendReturnType xgboost.data._from_scipy_csr	(	DataType	data,
		FloatCompatible	missing,
		int	nthread,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types
	)

protected

Initialize data from a CSR matrix.

◆ _maybe_np_slice()

np.ndarray xgboost.data._maybe_np_slice	(	DataType	data,
		Optional[NumpyDType]	dtype
	)

protected

Handle numpy slice.  This can be removed if we use __array_interface__.

◆ _meta_from_pandas_series()

None xgboost.data._meta_from_pandas_series	(	DataType	data,
		str	name,
		Optional[NumpyDType]	dtype,
		ctypes.c_void_p	handle
	)

protected

Help transform pandas series for meta data like labels

◆ _transform_dt_df()

Tuple[np.ndarray, Optional[FeatureNames], Optional[FeatureTypes]] xgboost.data._transform_dt_df	(	DataType	data,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types,
		Optional[str]	meta = `None`,
		Optional[NumpyDType]	meta_type = `None`
	)

protected

Validate feature names and types if data table

◆ CreatePageFormat()

template<typename T >

SparsePageFormat< T > * xgboost::data::CreatePageFormat ( const std::string & name )

inline

Create sparse page of format.

Returns: The created format functors.

◆ dispatch_data_backend()

DispatchedDataBackendReturnType xgboost.data.dispatch_data_backend	(	DataType	data,
		FloatCompatible	missing,
		int	threads,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types,
		bool	enable_categorical = `False`,
		DataSplitMode	data_split_mode = `DataSplitMode.ROW`
	)

Dispatch data for DMatrix.

◆ dispatch_meta_backend()

None xgboost.data.dispatch_meta_backend	(	DMatrix	matrix,
		DataType	data,
		str	name,
		Optional[NumpyDType]	dtype = `None`
	)

Dispatch for meta info.

◆ dispatch_proxy_set_data()

None xgboost.data.dispatch_proxy_set_data	(	_ProxyDMatrix	proxy,
		DataType	data,
		Optional[list]	cat_codes,
		bool	allow_host
	)

Dispatch for QuantileDMatrix.

◆ GetCutsFromRef()

void xgboost::data::GetCutsFromRef	(	Context const *	ctx,
		std::shared_ptr< DMatrix >	ref,
		bst_feature_t	n_features,
		BatchParam	p,
		common::HistogramCuts *	p_cuts
	)

Get quantile cuts from reference (Quantile)DMatrix.

Parameters

ctx	The context of the new DMatrix.
ref	The reference DMatrix.
n_features	Number of features, used for validation only.
p	Batch parameter for the new DMatrix.
p_cuts	Output quantile cuts.

◆ HostAdapterDispatch()

template<bool get_value = true, typename Fn >

decltype(auto) xgboost::data::HostAdapterDispatch	(	DMatrixProxy const *	proxy,
		Fn	fn,
		bool *	type_error = `nullptr`
	)

Dispatch function call based on input type.

Template Parameters

get_value	Whether the funciton Fn accept an adapter batch or the adapter itself.
Fn	The type of the function to be dispatched.

Parameters

proxy	The proxy object holding the reference to the input.
fn	The function to be dispatched.
type_error[out]	Set to ture if it's not null and the input data is not recognized by the host.

Returns: The return value of the function being dispatched.

◆ is_nullable_dtype()

bool xgboost.data.is_nullable_dtype ( PandasDType dtype )

Whether dtype is a pandas nullable type.

◆ is_pa_ext_categorical_dtype()

bool xgboost.data.is_pa_ext_categorical_dtype ( Any dtype )

Check whether dtype is a dictionary type.

◆ is_pa_ext_dtype()

bool xgboost.data.is_pa_ext_dtype ( Any dtype )

Return whether dtype is a pyarrow extension type for pandas

◆ is_pd_cat_dtype()

bool xgboost.data.is_pd_cat_dtype ( PandasDType dtype )

Wrapper for testing pandas category type.

◆ is_pd_sparse_dtype()

bool xgboost.data.is_pd_sparse_dtype ( PandasDType dtype )

Wrapper for testing pandas sparse type.

◆ pandas_cat_null()

DataFrame xgboost.data.pandas_cat_null ( DataFrame data )

Handle categorical dtype and nullable extension types from pandas.

◆ pandas_ext_num_types()

DataFrame xgboost.data.pandas_ext_num_types ( DataFrame data )

Experimental suppport for handling pyarrow extension numeric types.

◆ pandas_feature_info()

Tuple[Optional[FeatureNames], Optional[FeatureTypes]] xgboost.data.pandas_feature_info	(	DataFrame	data,
		Optional[str]	meta,
		Optional[FeatureNames]	feature_names,
		Optional[FeatureTypes]	feature_types,
		bool	enable_categorical
	)

Handle feature info for pandas dataframe.

◆ record_batch_data_iter()

Callable xgboost.data.record_batch_data_iter ( Iterator data_iter )

Data iterator used to ingest Arrow columnar record batches. We are not using
class DataIter because it is only intended for building Device DMatrix and external
memory DMatrix.

◆ transform_scipy_sparse()

DataType xgboost.data.transform_scipy_sparse	(	DataType	data,
		bool	is_csr
	)

Ensure correct data alignment and data type for scipy sparse inputs. Input should
be either csr or csc matrix.

Variable Documentation

◆ _ENABLE_CAT_ERR

tuple xgboost.data._ENABLE_CAT_ERR

protected

Initial value:

=  (
    "When categorical type is supplied, The experimental DMatrix parameter"
    "`enable_categorical` must be set to `True`."
)

◆ _pandas_dtype_mapper

dict xgboost.data._pandas_dtype_mapper

protected

Initial value:

=  {
    "int8": "int",
    "int16": "int",
    "int32": "int",
    "int64": "int",
    "uint8": "int",
    "uint16": "int",
    "uint32": "int",
    "uint64": "int",
    "float16": "float",
    "float32": "float",
    "float64": "float",
    "bool": "i",
}

◆ DispatchedDataBackendReturnType

xgboost.data.DispatchedDataBackendReturnType

Initial value:

=  Tuple[
    ctypes.c_void_p, Optional[FeatureNames], Optional[FeatureTypes]
]

◆ kAdapterUnknownSize

constexpr size_t xgboost::data.kAdapterUnknownSize = std::numeric_limits<size_t >::max()

constexpr

External data formats should implement an adapter as below.

The adapter provides a uniform access to data outside xgboost, allowing construction of DMatrix objects from a range of sources without duplicating code.

The adapter object is an iterator that returns batches of data. Each batch contains a number of "lines". A line represents a set of elements from a sparse input matrix, normally a row in the case of a CSR matrix or a column for a CSC matrix. Typically in sparse matrix formats we can efficiently access subsets of elements at a time, but cannot efficiently lookups elements by random access, hence the "line" abstraction, allowing the sparse matrix to return subsets of elements efficiently. Individual elements are described by a COO tuple (row index, column index, value).

This abstraction allows us to read through different sparse matrix formats using the same interface. In particular we can write a DMatrix constructor that uses the same code to construct itself from a CSR matrix, CSC matrix, dense matrix, CSV, LIBSVM file, or potentially other formats. To see why this is necessary, imagine we have 5 external matrix formats and 5 internal DMatrix types where each DMatrix needs a custom constructor for each possible input. The number of constructors is 5*5=25. Using an abstraction over the input data types the number of constructors is reduced to 5, as each DMatrix is oblivious to the external data format. Adding a new input source is simply a case of implementing an adapter.

Most of the below adapters do not need more than one batch as the data originates from an in memory source. The file adapter does require batches to avoid loading the entire file in memory.

An important detail is empty row/column handling. Files loaded from disk do not provide meta information about the number of rows/columns to expect, this needs to be inferred during construction. Other sparse formats may specify a number of rows/columns, but we can encounter entirely sparse rows or columns, leading to disagreement between the inferred number and the meta-info provided. To resolve this, adapters have methods specifying the number of rows/columns expected, these methods may return zero where these values must be inferred from data. A constructed DMatrix should agree with the input source on numbers of rows/columns, appending empty rows if necessary.

An adapter can return this value for number of rows or columns indicating that this value is currently unknown and should be inferred while passing over the data.

◆ pandas_nullable_mapper

dict xgboost.data.pandas_nullable_mapper

Initial value:

=  {
    "Int8": "int",
    "Int16": "int",
    "Int32": "int",
    "Int64": "int",
    "UInt8": "i",
    "UInt16": "i",
    "UInt32": "i",
    "UInt64": "i",
    "Float32": "float",
    "Float64": "float",
    "boolean": "i",
}

◆ pandas_pyarrow_mapper

dict xgboost.data.pandas_pyarrow_mapper

Initial value:

=  {
    "int8[pyarrow]": "i",
    "int16[pyarrow]": "i",
    "int32[pyarrow]": "i",
    "int64[pyarrow]": "i",
    "uint8[pyarrow]": "i",
    "uint16[pyarrow]": "i",
    "uint32[pyarrow]": "i",
    "uint64[pyarrow]": "i",
    "float[pyarrow]": "float",
    "float32[pyarrow]": "float",
    "double[pyarrow]": "float",
    "float64[pyarrow]": "float",
    "bool[pyarrow]": "i",
}

Namespaces

Data Structures

Typedefs

Enumerations

Functions

Variables

Detailed Description

Function Documentation

◆ _cudf_array_interfaces()

◆ _from_cupy_array()

◆ _from_numpy_array()

◆ _from_scipy_csc()

◆ _from_scipy_csr()

◆ _maybe_np_slice()

◆ _meta_from_pandas_series()

◆ _transform_dt_df()

◆ CreatePageFormat()

◆ dispatch_data_backend()

◆ dispatch_meta_backend()

◆ dispatch_proxy_set_data()

◆ GetCutsFromRef()

◆ HostAdapterDispatch()

◆ is_nullable_dtype()

◆ is_pa_ext_categorical_dtype()

◆ is_pa_ext_dtype()

◆ is_pd_cat_dtype()

◆ is_pd_sparse_dtype()

◆ pandas_cat_null()

◆ pandas_ext_num_types()

◆ pandas_feature_info()

◆ record_batch_data_iter()

◆ transform_scipy_sparse()

Variable Documentation

◆ _ENABLE_CAT_ERR

◆ _pandas_dtype_mapper

◆ DispatchedDataBackendReturnType

◆ kAdapterUnknownSize

◆ pandas_nullable_mapper

◆ pandas_pyarrow_mapper