ARROW-1167: [Python] Support chunking string columns in Table.from_pandas This resolves the error with converting the dataset in ARROW-1167, which only takes up 4.5 GB in memory but has a single column with over 2GB in binary data.
The unit test for this is not run in CI because of large memory allocation, but can be run with ``` py.test pyarrow --large_memory ``` cc @jeffknupp Author: Wes McKinney <wes.mckin...@twosigma.com> Closes #867 from wesm/ARROW-1167 and squashes the following commits: dae62326 [Wes McKinney] cpplint dcdec91a [Wes McKinney] Support ChunkedArray outputs of Array.from_pandas 150e9fc9 [Wes McKinney] Produced ChunkedArray when exceeding 2GB in a single BinaryArray column 707555f8 [Wes McKinney] Split up pandas_convert, make PandasObjectsToArrow return ChunkedArray to accommodate large string data Project: http://git-wip-us.apache.org/repos/asf/arrow/repo Commit: http://git-wip-us.apache.org/repos/asf/arrow/commit/2c5b412c Tree: http://git-wip-us.apache.org/repos/asf/arrow/tree/2c5b412c Diff: http://git-wip-us.apache.org/repos/asf/arrow/diff/2c5b412c Branch: refs/heads/master Commit: 2c5b412c2866b6561d35ba3399036c22b646d699 Parents: 6999dbd Author: Wes McKinney <wes.mckin...@twosigma.com> Authored: Wed Jul 19 08:16:25 2017 -0400 Committer: Wes McKinney <wes.mckin...@twosigma.com> Committed: Wed Jul 19 08:16:25 2017 -0400 ---------------------------------------------------------------------- cpp/src/arrow/builder.h | 6 + cpp/src/arrow/ipc/feather.cc | 3 +- cpp/src/arrow/python/CMakeLists.txt | 6 +- cpp/src/arrow/python/api.h | 3 +- cpp/src/arrow/python/arrow_to_pandas.cc | 1627 ++++++++++++++ cpp/src/arrow/python/arrow_to_pandas.h | 67 + cpp/src/arrow/python/pandas_convert.cc | 2609 ---------------------- cpp/src/arrow/python/pandas_convert.h | 77 - cpp/src/arrow/python/pandas_to_arrow.cc | 1099 +++++++++ cpp/src/arrow/python/pandas_to_arrow.h | 58 + cpp/src/arrow/python/python-test.cc | 2 +- cpp/src/arrow/table.h | 3 + python/pyarrow/array.pxi | 11 +- python/pyarrow/includes/libarrow.pxd | 3 +- python/pyarrow/parquet.py | 21 +- python/pyarrow/public-api.pxi | 15 + python/pyarrow/table.pxi | 38 +- python/pyarrow/tests/conftest.py | 5 +- python/pyarrow/tests/test_convert_pandas.py | 14 + 19 files changed, 2949 insertions(+), 2718 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/builder.h ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/builder.h b/cpp/src/arrow/builder.h index 6b54c9f..065e115 100644 --- a/cpp/src/arrow/builder.h +++ b/cpp/src/arrow/builder.h @@ -585,6 +585,9 @@ class ARROW_EXPORT BinaryBuilder : public ArrayBuilder { Status Resize(int64_t capacity) override; Status Finish(std::shared_ptr<Array>* out) override; + /// \return size of values buffer so far + int64_t value_data_length() const { return value_data_builder_.length(); } + /// Temporary access to a value. /// /// This pointer becomes invalid on the next modifying operation. @@ -632,6 +635,9 @@ class ARROW_EXPORT FixedSizeBinaryBuilder : public ArrayBuilder { Status Resize(int64_t capacity) override; Status Finish(std::shared_ptr<Array>* out) override; + /// \return size of values buffer so far + int64_t value_data_length() const { return byte_builder_.length(); } + protected: int32_t byte_width_; BufferBuilder byte_builder_; http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/ipc/feather.cc ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/ipc/feather.cc b/cpp/src/arrow/ipc/feather.cc index d5b938b..61b96e0 100644 --- a/cpp/src/arrow/ipc/feather.cc +++ b/cpp/src/arrow/ipc/feather.cc @@ -497,8 +497,7 @@ fbs::Type ToFlatbufferType(Type::type type) { return fbs::Type_MIN; } -static Status SanitizeUnsupportedTypes( - const Array& values, std::shared_ptr<Array>* out) { +static Status SanitizeUnsupportedTypes(const Array& values, std::shared_ptr<Array>* out) { if (values.type_id() == Type::NA) { // As long as R doesn't support NA, we write this as a StringColumn // to ensure stable roundtrips. http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/python/CMakeLists.txt ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/python/CMakeLists.txt b/cpp/src/arrow/python/CMakeLists.txt index d5e980b..0fdf81e 100644 --- a/cpp/src/arrow/python/CMakeLists.txt +++ b/cpp/src/arrow/python/CMakeLists.txt @@ -42,6 +42,7 @@ set(ARROW_PYTHON_TEST_LINK_LIBS ${ARROW_PYTHON_MIN_TEST_LIBS}) # ---------------------------------------------------------------------- set(ARROW_PYTHON_SRCS + arrow_to_pandas.cc builtin_convert.cc common.cc config.cc @@ -49,7 +50,7 @@ set(ARROW_PYTHON_SRCS init.cc io.cc numpy_convert.cc - pandas_convert.cc + pandas_to_arrow.cc pyarrow.cc ) @@ -81,6 +82,7 @@ endif() install(FILES api.h + arrow_to_pandas.h builtin_convert.h common.h config.h @@ -89,7 +91,7 @@ install(FILES io.h numpy_convert.h numpy_interop.h - pandas_convert.h + pandas_to_arrow.h platform.h pyarrow.h type_traits.h http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/python/api.h ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/python/api.h b/cpp/src/arrow/python/api.h index 895d1f4..7cb36ad 100644 --- a/cpp/src/arrow/python/api.h +++ b/cpp/src/arrow/python/api.h @@ -18,11 +18,12 @@ #ifndef ARROW_PYTHON_API_H #define ARROW_PYTHON_API_H +#include "arrow/python/arrow_to_pandas.h" #include "arrow/python/builtin_convert.h" #include "arrow/python/common.h" #include "arrow/python/helpers.h" #include "arrow/python/io.h" #include "arrow/python/numpy_convert.h" -#include "arrow/python/pandas_convert.h" +#include "arrow/python/pandas_to_arrow.h" #endif // ARROW_PYTHON_API_H http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/python/arrow_to_pandas.cc ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/python/arrow_to_pandas.cc b/cpp/src/arrow/python/arrow_to_pandas.cc new file mode 100644 index 0000000..d40609f --- /dev/null +++ b/cpp/src/arrow/python/arrow_to_pandas.cc @@ -0,0 +1,1627 @@ +// 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. + +// Functions for pandas conversion via NumPy + +#include "arrow/python/numpy_interop.h" + +#include "arrow/python/arrow_to_pandas.h" + +#include <algorithm> +#include <atomic> +#include <cmath> +#include <cstdint> +#include <memory> +#include <mutex> +#include <sstream> +#include <string> +#include <thread> +#include <unordered_map> +#include <vector> + +#include "arrow/array.h" +#include "arrow/status.h" +#include "arrow/table.h" +#include "arrow/type_fwd.h" +#include "arrow/type_traits.h" +#include "arrow/util/bit-util.h" +#include "arrow/util/decimal.h" +#include "arrow/util/logging.h" +#include "arrow/util/macros.h" +#include "arrow/visitor_inline.h" + +#include "arrow/python/builtin_convert.h" +#include "arrow/python/common.h" +#include "arrow/python/config.h" +#include "arrow/python/helpers.h" +#include "arrow/python/numpy-internal.h" +#include "arrow/python/numpy_convert.h" +#include "arrow/python/type_traits.h" +#include "arrow/python/util/datetime.h" + +namespace arrow { +namespace py { + +// ---------------------------------------------------------------------- +// Utility code + +template <typename T> +struct WrapBytes {}; + +template <> +struct WrapBytes<StringArray> { + static inline PyObject* Wrap(const uint8_t* data, int64_t length) { + return PyUnicode_FromStringAndSize(reinterpret_cast<const char*>(data), length); + } +}; + +template <> +struct WrapBytes<BinaryArray> { + static inline PyObject* Wrap(const uint8_t* data, int64_t length) { + return PyBytes_FromStringAndSize(reinterpret_cast<const char*>(data), length); + } +}; + +template <> +struct WrapBytes<FixedSizeBinaryArray> { + static inline PyObject* Wrap(const uint8_t* data, int64_t length) { + return PyBytes_FromStringAndSize(reinterpret_cast<const char*>(data), length); + } +}; + +static inline bool ListTypeSupported(const DataType& type) { + switch (type.id()) { + case Type::UINT8: + case Type::INT8: + case Type::UINT16: + case Type::INT16: + case Type::UINT32: + case Type::INT32: + case Type::INT64: + case Type::UINT64: + case Type::FLOAT: + case Type::DOUBLE: + case Type::STRING: + case Type::TIMESTAMP: + // The above types are all supported. + return true; + case Type::LIST: { + const ListType& list_type = static_cast<const ListType&>(type); + return ListTypeSupported(*list_type.value_type()); + } + default: + break; + } + return false; +} + +// ---------------------------------------------------------------------- +// pandas 0.x DataFrame conversion internals + +inline void set_numpy_metadata(int type, DataType* datatype, PyArray_Descr* out) { + if (type == NPY_DATETIME) { + auto date_dtype = reinterpret_cast<PyArray_DatetimeDTypeMetaData*>(out->c_metadata); + if (datatype->id() == Type::TIMESTAMP) { + auto timestamp_type = static_cast<TimestampType*>(datatype); + + switch (timestamp_type->unit()) { + case TimestampType::Unit::SECOND: + date_dtype->meta.base = NPY_FR_s; + break; + case TimestampType::Unit::MILLI: + date_dtype->meta.base = NPY_FR_ms; + break; + case TimestampType::Unit::MICRO: + date_dtype->meta.base = NPY_FR_us; + break; + case TimestampType::Unit::NANO: + date_dtype->meta.base = NPY_FR_ns; + break; + } + } else { + // datatype->type == Type::DATE64 + date_dtype->meta.base = NPY_FR_D; + } + } +} + +static inline PyArray_Descr* GetSafeNumPyDtype(int type) { + if (type == NPY_DATETIME) { + // It is not safe to mutate the result of DescrFromType + return PyArray_DescrNewFromType(type); + } else { + return PyArray_DescrFromType(type); + } +} +static inline PyObject* NewArray1DFromType( + DataType* arrow_type, int type, int64_t length, void* data) { + npy_intp dims[1] = {length}; + + PyArray_Descr* descr = GetSafeNumPyDtype(type); + if (descr == nullptr) { + // Error occurred, trust error state is set + return nullptr; + } + + set_numpy_metadata(type, arrow_type, descr); + return PyArray_NewFromDescr(&PyArray_Type, descr, 1, dims, nullptr, data, + NPY_ARRAY_OWNDATA | NPY_ARRAY_CARRAY | NPY_ARRAY_WRITEABLE, nullptr); +} + +class PandasBlock { + public: + enum type { + OBJECT, + UINT8, + INT8, + UINT16, + INT16, + UINT32, + INT32, + UINT64, + INT64, + FLOAT, + DOUBLE, + BOOL, + DATETIME, + DATETIME_WITH_TZ, + CATEGORICAL + }; + + PandasBlock(int64_t num_rows, int num_columns) + : num_rows_(num_rows), num_columns_(num_columns) {} + virtual ~PandasBlock() {} + + virtual Status Allocate() = 0; + virtual Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) = 0; + + PyObject* block_arr() const { return block_arr_.obj(); } + + virtual Status GetPyResult(PyObject** output) { + PyObject* result = PyDict_New(); + RETURN_IF_PYERROR(); + + PyDict_SetItemString(result, "block", block_arr_.obj()); + PyDict_SetItemString(result, "placement", placement_arr_.obj()); + + *output = result; + + return Status::OK(); + } + + protected: + Status AllocateNDArray(int npy_type, int ndim = 2) { + PyAcquireGIL lock; + + PyArray_Descr* descr = GetSafeNumPyDtype(npy_type); + + PyObject* block_arr; + if (ndim == 2) { + npy_intp block_dims[2] = {num_columns_, num_rows_}; + block_arr = PyArray_SimpleNewFromDescr(2, block_dims, descr); + } else { + npy_intp block_dims[1] = {num_rows_}; + block_arr = PyArray_SimpleNewFromDescr(1, block_dims, descr); + } + + if (block_arr == NULL) { + // TODO(wesm): propagating Python exception + return Status::OK(); + } + + PyArray_ENABLEFLAGS(reinterpret_cast<PyArrayObject*>(block_arr), NPY_ARRAY_OWNDATA); + + npy_intp placement_dims[1] = {num_columns_}; + PyObject* placement_arr = PyArray_SimpleNew(1, placement_dims, NPY_INT64); + if (placement_arr == NULL) { + // TODO(wesm): propagating Python exception + return Status::OK(); + } + + block_arr_.reset(block_arr); + placement_arr_.reset(placement_arr); + + block_data_ = reinterpret_cast<uint8_t*>( + PyArray_DATA(reinterpret_cast<PyArrayObject*>(block_arr))); + + placement_data_ = reinterpret_cast<int64_t*>( + PyArray_DATA(reinterpret_cast<PyArrayObject*>(placement_arr))); + + return Status::OK(); + } + + int64_t num_rows_; + int num_columns_; + + OwnedRef block_arr_; + uint8_t* block_data_; + + // ndarray<int32> + OwnedRef placement_arr_; + int64_t* placement_data_; + + private: + DISALLOW_COPY_AND_ASSIGN(PandasBlock); +}; + +template <typename T> +inline void ConvertIntegerWithNulls(const ChunkedArray& data, double* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const T*>(prim_arr->raw_values()); + // Upcast to double, set NaN as appropriate + + for (int i = 0; i < arr->length(); ++i) { + *out_values++ = prim_arr->IsNull(i) ? NAN : static_cast<double>(in_values[i]); + } + } +} + +template <typename T> +inline void ConvertIntegerNoNullsSameType(const ChunkedArray& data, T* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const T*>(prim_arr->raw_values()); + memcpy(out_values, in_values, sizeof(T) * arr->length()); + out_values += arr->length(); + } +} + +template <typename InType, typename OutType> +inline void ConvertIntegerNoNullsCast(const ChunkedArray& data, OutType* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const InType*>(prim_arr->raw_values()); + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values = in_values[i]; + } + } +} + +static Status ConvertBooleanWithNulls(const ChunkedArray& data, PyObject** out_values) { + PyAcquireGIL lock; + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto bool_arr = static_cast<BooleanArray*>(arr.get()); + + for (int64_t i = 0; i < arr->length(); ++i) { + if (bool_arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values++ = Py_None; + } else if (bool_arr->Value(i)) { + // True + Py_INCREF(Py_True); + *out_values++ = Py_True; + } else { + // False + Py_INCREF(Py_False); + *out_values++ = Py_False; + } + } + } + return Status::OK(); +} + +static void ConvertBooleanNoNulls(const ChunkedArray& data, uint8_t* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto bool_arr = static_cast<BooleanArray*>(arr.get()); + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values++ = static_cast<uint8_t>(bool_arr->Value(i)); + } + } +} + +template <typename Type> +inline Status ConvertBinaryLike(const ChunkedArray& data, PyObject** out_values) { + using ArrayType = typename TypeTraits<Type>::ArrayType; + PyAcquireGIL lock; + for (int c = 0; c < data.num_chunks(); c++) { + auto arr = static_cast<ArrayType*>(data.chunk(c).get()); + + const uint8_t* data_ptr; + int32_t length; + const bool has_nulls = data.null_count() > 0; + for (int64_t i = 0; i < arr->length(); ++i) { + if (has_nulls && arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values = Py_None; + } else { + data_ptr = arr->GetValue(i, &length); + *out_values = WrapBytes<ArrayType>::Wrap(data_ptr, length); + if (*out_values == nullptr) { + PyErr_Clear(); + std::stringstream ss; + ss << "Wrapping " + << std::string(reinterpret_cast<const char*>(data_ptr), length) << " failed"; + return Status::UnknownError(ss.str()); + } + } + ++out_values; + } + } + return Status::OK(); +} + +inline Status ConvertNulls(const ChunkedArray& data, PyObject** out_values) { + PyAcquireGIL lock; + for (int c = 0; c < data.num_chunks(); c++) { + std::shared_ptr<Array> arr = data.chunk(c); + + for (int64_t i = 0; i < arr->length(); ++i) { + // All values are null + Py_INCREF(Py_None); + *out_values = Py_None; + ++out_values; + } + } + return Status::OK(); +} + +inline Status ConvertFixedSizeBinary(const ChunkedArray& data, PyObject** out_values) { + PyAcquireGIL lock; + for (int c = 0; c < data.num_chunks(); c++) { + auto arr = static_cast<FixedSizeBinaryArray*>(data.chunk(c).get()); + + const uint8_t* data_ptr; + int32_t length = + std::dynamic_pointer_cast<FixedSizeBinaryType>(arr->type())->byte_width(); + const bool has_nulls = data.null_count() > 0; + for (int64_t i = 0; i < arr->length(); ++i) { + if (has_nulls && arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values = Py_None; + } else { + data_ptr = arr->GetValue(i); + *out_values = WrapBytes<FixedSizeBinaryArray>::Wrap(data_ptr, length); + if (*out_values == nullptr) { + PyErr_Clear(); + std::stringstream ss; + ss << "Wrapping " + << std::string(reinterpret_cast<const char*>(data_ptr), length) << " failed"; + return Status::UnknownError(ss.str()); + } + } + ++out_values; + } + } + return Status::OK(); +} + +inline Status ConvertStruct(const ChunkedArray& data, PyObject** out_values) { + PyAcquireGIL lock; + if (data.num_chunks() <= 0) { return Status::OK(); } + // ChunkedArray has at least one chunk + auto arr = static_cast<const StructArray*>(data.chunk(0).get()); + // Use it to cache the struct type and number of fields for all chunks + int32_t num_fields = arr->num_fields(); + auto array_type = arr->type(); + std::vector<OwnedRef> fields_data(num_fields); + OwnedRef dict_item; + for (int c = 0; c < data.num_chunks(); c++) { + auto arr = static_cast<const StructArray*>(data.chunk(c).get()); + // Convert the struct arrays first + for (int32_t i = 0; i < num_fields; i++) { + PyObject* numpy_array; + RETURN_NOT_OK( + ConvertArrayToPandas(arr->field(static_cast<int>(i)), nullptr, &numpy_array)); + fields_data[i].reset(numpy_array); + } + + // Construct a dictionary for each row + const bool has_nulls = data.null_count() > 0; + for (int64_t i = 0; i < arr->length(); ++i) { + if (has_nulls && arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values = Py_None; + } else { + // Build the new dict object for the row + dict_item.reset(PyDict_New()); + RETURN_IF_PYERROR(); + for (int32_t field_idx = 0; field_idx < num_fields; ++field_idx) { + OwnedRef field_value; + auto name = array_type->child(static_cast<int>(field_idx))->name(); + if (!arr->field(static_cast<int>(field_idx))->IsNull(i)) { + // Value exists in child array, obtain it + auto array = reinterpret_cast<PyArrayObject*>(fields_data[field_idx].obj()); + auto ptr = reinterpret_cast<const char*>(PyArray_GETPTR1(array, i)); + field_value.reset(PyArray_GETITEM(array, ptr)); + RETURN_IF_PYERROR(); + } else { + // Translate the Null to a None + Py_INCREF(Py_None); + field_value.reset(Py_None); + } + // PyDict_SetItemString does not steal the value reference + auto setitem_result = + PyDict_SetItemString(dict_item.obj(), name.c_str(), field_value.obj()); + RETURN_IF_PYERROR(); + DCHECK_EQ(setitem_result, 0); + } + *out_values = dict_item.obj(); + // Grant ownership to the resulting array + Py_INCREF(*out_values); + } + ++out_values; + } + } + return Status::OK(); +} + +template <typename ArrowType> +inline Status ConvertListsLike( + const std::shared_ptr<Column>& col, PyObject** out_values) { + const ChunkedArray& data = *col->data().get(); + auto list_type = std::static_pointer_cast<ListType>(col->type()); + + // Get column of underlying value arrays + std::vector<std::shared_ptr<Array>> value_arrays; + for (int c = 0; c < data.num_chunks(); c++) { + auto arr = std::static_pointer_cast<ListArray>(data.chunk(c)); + value_arrays.emplace_back(arr->values()); + } + auto flat_column = std::make_shared<Column>(list_type->value_field(), value_arrays); + // TODO(ARROW-489): Currently we don't have a Python reference for single columns. + // Storing a reference to the whole Array would be to expensive. + PyObject* numpy_array; + RETURN_NOT_OK(ConvertColumnToPandas(flat_column, nullptr, &numpy_array)); + + PyAcquireGIL lock; + + for (int c = 0; c < data.num_chunks(); c++) { + auto arr = std::static_pointer_cast<ListArray>(data.chunk(c)); + + const bool has_nulls = data.null_count() > 0; + for (int64_t i = 0; i < arr->length(); ++i) { + if (has_nulls && arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values = Py_None; + } else { + PyObject* start = PyLong_FromLong(arr->value_offset(i)); + PyObject* end = PyLong_FromLong(arr->value_offset(i + 1)); + PyObject* slice = PySlice_New(start, end, NULL); + *out_values = PyObject_GetItem(numpy_array, slice); + Py_DECREF(start); + Py_DECREF(end); + Py_DECREF(slice); + } + ++out_values; + } + } + + Py_XDECREF(numpy_array); + return Status::OK(); +} + +template <typename T> +inline void ConvertNumericNullable(const ChunkedArray& data, T na_value, T* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const T*>(prim_arr->raw_values()); + + const uint8_t* valid_bits = arr->null_bitmap_data(); + + if (arr->null_count() > 0) { + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values++ = BitUtil::BitNotSet(valid_bits, i) ? na_value : in_values[i]; + } + } else { + memcpy(out_values, in_values, sizeof(T) * arr->length()); + out_values += arr->length(); + } + } +} + +template <typename InType, typename OutType> +inline void ConvertNumericNullableCast( + const ChunkedArray& data, OutType na_value, OutType* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const InType*>(prim_arr->raw_values()); + + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values++ = arr->IsNull(i) ? na_value : static_cast<OutType>(in_values[i]); + } + } +} + +template <typename InType, int64_t SHIFT> +inline void ConvertDatetimeNanos(const ChunkedArray& data, int64_t* out_values) { + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const InType*>(prim_arr->raw_values()); + + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values++ = arr->IsNull(i) ? kPandasTimestampNull + : (static_cast<int64_t>(in_values[i]) * SHIFT); + } + } +} + +template <typename TYPE> +static Status ConvertTimes(const ChunkedArray& data, PyObject** out_values) { + using ArrayType = typename TypeTraits<TYPE>::ArrayType; + + PyAcquireGIL lock; + OwnedRef time_ref; + + PyDateTime_IMPORT; + + for (int c = 0; c < data.num_chunks(); c++) { + const auto& arr = static_cast<const ArrayType&>(*data.chunk(c)); + auto type = std::dynamic_pointer_cast<TYPE>(arr.type()); + DCHECK(type); + + const TimeUnit::type unit = type->unit(); + + for (int64_t i = 0; i < arr.length(); ++i) { + if (arr.IsNull(i)) { + Py_INCREF(Py_None); + *out_values++ = Py_None; + } else { + RETURN_NOT_OK(PyTime_from_int(arr.Value(i), unit, out_values++)); + RETURN_IF_PYERROR(); + } + } + } + + return Status::OK(); +} + +template <typename T> +Status ValidateDecimalPrecision(int precision) { + constexpr static const int maximum_precision = decimal::DecimalPrecision<T>::maximum; + if (!(precision > 0 && precision <= maximum_precision)) { + std::stringstream ss; + ss << "Invalid precision: " << precision << ". Minimum is 1, maximum is " + << maximum_precision; + return Status::Invalid(ss.str()); + } + return Status::OK(); +} + +template <typename T> +Status RawDecimalToString( + const uint8_t* bytes, int precision, int scale, std::string* result) { + DCHECK_NE(bytes, nullptr); + DCHECK_NE(result, nullptr); + RETURN_NOT_OK(ValidateDecimalPrecision<T>(precision)); + decimal::Decimal<T> decimal; + FromBytes(bytes, &decimal); + *result = ToString(decimal, precision, scale); + return Status::OK(); +} + +template Status RawDecimalToString<int32_t>( + const uint8_t*, int, int, std::string* result); +template Status RawDecimalToString<int64_t>( + const uint8_t*, int, int, std::string* result); + +Status RawDecimalToString(const uint8_t* bytes, int precision, int scale, + bool is_negative, std::string* result) { + DCHECK_NE(bytes, nullptr); + DCHECK_NE(result, nullptr); + RETURN_NOT_OK(ValidateDecimalPrecision<boost::multiprecision::int128_t>(precision)); + decimal::Decimal128 decimal; + FromBytes(bytes, is_negative, &decimal); + *result = ToString(decimal, precision, scale); + return Status::OK(); +} + +static Status ConvertDecimals(const ChunkedArray& data, PyObject** out_values) { + PyAcquireGIL lock; + OwnedRef decimal_ref; + OwnedRef Decimal_ref; + RETURN_NOT_OK(ImportModule("decimal", &decimal_ref)); + RETURN_NOT_OK(ImportFromModule(decimal_ref, "Decimal", &Decimal_ref)); + PyObject* Decimal = Decimal_ref.obj(); + + for (int c = 0; c < data.num_chunks(); c++) { + auto* arr(static_cast<arrow::DecimalArray*>(data.chunk(c).get())); + auto type(std::dynamic_pointer_cast<arrow::DecimalType>(arr->type())); + const int precision = type->precision(); + const int scale = type->scale(); + const int bit_width = type->bit_width(); + + for (int64_t i = 0; i < arr->length(); ++i) { + if (arr->IsNull(i)) { + Py_INCREF(Py_None); + *out_values++ = Py_None; + } else { + const uint8_t* raw_value = arr->GetValue(i); + std::string s; + switch (bit_width) { + case 32: + RETURN_NOT_OK(RawDecimalToString<int32_t>(raw_value, precision, scale, &s)); + break; + case 64: + RETURN_NOT_OK(RawDecimalToString<int64_t>(raw_value, precision, scale, &s)); + break; + case 128: + RETURN_NOT_OK( + RawDecimalToString(raw_value, precision, scale, arr->IsNegative(i), &s)); + break; + default: + break; + } + RETURN_NOT_OK(DecimalFromString(Decimal, s, out_values++)); + } + } + } + + return Status::OK(); +} + +#define CONVERTLISTSLIKE_CASE(ArrowType, ArrowEnum) \ + case Type::ArrowEnum: \ + RETURN_NOT_OK((ConvertListsLike<ArrowType>(col, out_buffer))); \ + break; + +class ObjectBlock : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status Allocate() override { return AllocateNDArray(NPY_OBJECT); } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + PyObject** out_buffer = + reinterpret_cast<PyObject**>(block_data_) + rel_placement * num_rows_; + + const ChunkedArray& data = *col->data().get(); + + if (type == Type::BOOL) { + RETURN_NOT_OK(ConvertBooleanWithNulls(data, out_buffer)); + } else if (type == Type::BINARY) { + RETURN_NOT_OK(ConvertBinaryLike<BinaryType>(data, out_buffer)); + } else if (type == Type::STRING) { + RETURN_NOT_OK(ConvertBinaryLike<StringType>(data, out_buffer)); + } else if (type == Type::FIXED_SIZE_BINARY) { + RETURN_NOT_OK(ConvertFixedSizeBinary(data, out_buffer)); + } else if (type == Type::TIME32) { + RETURN_NOT_OK(ConvertTimes<Time32Type>(data, out_buffer)); + } else if (type == Type::TIME64) { + RETURN_NOT_OK(ConvertTimes<Time64Type>(data, out_buffer)); + } else if (type == Type::DECIMAL) { + RETURN_NOT_OK(ConvertDecimals(data, out_buffer)); + } else if (type == Type::NA) { + RETURN_NOT_OK(ConvertNulls(data, out_buffer)); + } else if (type == Type::LIST) { + auto list_type = std::static_pointer_cast<ListType>(col->type()); + switch (list_type->value_type()->id()) { + CONVERTLISTSLIKE_CASE(UInt8Type, UINT8) + CONVERTLISTSLIKE_CASE(Int8Type, INT8) + CONVERTLISTSLIKE_CASE(UInt16Type, UINT16) + CONVERTLISTSLIKE_CASE(Int16Type, INT16) + CONVERTLISTSLIKE_CASE(UInt32Type, UINT32) + CONVERTLISTSLIKE_CASE(Int32Type, INT32) + CONVERTLISTSLIKE_CASE(UInt64Type, UINT64) + CONVERTLISTSLIKE_CASE(Int64Type, INT64) + CONVERTLISTSLIKE_CASE(TimestampType, TIMESTAMP) + CONVERTLISTSLIKE_CASE(FloatType, FLOAT) + CONVERTLISTSLIKE_CASE(DoubleType, DOUBLE) + CONVERTLISTSLIKE_CASE(StringType, STRING) + CONVERTLISTSLIKE_CASE(ListType, LIST) + default: { + std::stringstream ss; + ss << "Not implemented type for conversion from List to Pandas ObjectBlock: " + << list_type->value_type()->ToString(); + return Status::NotImplemented(ss.str()); + } + } + } else if (type == Type::STRUCT) { + RETURN_NOT_OK(ConvertStruct(data, out_buffer)); + } else { + std::stringstream ss; + ss << "Unsupported type for object array output: " << col->type()->ToString(); + return Status::NotImplemented(ss.str()); + } + + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +template <int ARROW_TYPE, typename C_TYPE> +class IntBlock : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status Allocate() override { + return AllocateNDArray(arrow_traits<ARROW_TYPE>::npy_type); + } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + C_TYPE* out_buffer = + reinterpret_cast<C_TYPE*>(block_data_) + rel_placement * num_rows_; + + const ChunkedArray& data = *col->data().get(); + + if (type != ARROW_TYPE) { + std::stringstream ss; + ss << "Cannot write Arrow data of type " << col->type()->ToString(); + ss << " to a Pandas int" << sizeof(C_TYPE) << " block."; + return Status::NotImplemented(ss.str()); + } + + ConvertIntegerNoNullsSameType<C_TYPE>(data, out_buffer); + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +using UInt8Block = IntBlock<Type::UINT8, uint8_t>; +using Int8Block = IntBlock<Type::INT8, int8_t>; +using UInt16Block = IntBlock<Type::UINT16, uint16_t>; +using Int16Block = IntBlock<Type::INT16, int16_t>; +using UInt32Block = IntBlock<Type::UINT32, uint32_t>; +using Int32Block = IntBlock<Type::INT32, int32_t>; +using UInt64Block = IntBlock<Type::UINT64, uint64_t>; +using Int64Block = IntBlock<Type::INT64, int64_t>; + +class Float32Block : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status Allocate() override { return AllocateNDArray(NPY_FLOAT32); } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + if (type != Type::FLOAT) { + std::stringstream ss; + ss << "Cannot write Arrow data of type " << col->type()->ToString(); + ss << " to a Pandas float32 block."; + return Status::NotImplemented(ss.str()); + } + + float* out_buffer = reinterpret_cast<float*>(block_data_) + rel_placement * num_rows_; + + ConvertNumericNullable<float>(*col->data().get(), NAN, out_buffer); + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +class Float64Block : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status Allocate() override { return AllocateNDArray(NPY_FLOAT64); } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + double* out_buffer = + reinterpret_cast<double*>(block_data_) + rel_placement * num_rows_; + + const ChunkedArray& data = *col->data().get(); + +#define INTEGER_CASE(IN_TYPE) \ + ConvertIntegerWithNulls<IN_TYPE>(data, out_buffer); \ + break; + + switch (type) { + case Type::UINT8: + INTEGER_CASE(uint8_t); + case Type::INT8: + INTEGER_CASE(int8_t); + case Type::UINT16: + INTEGER_CASE(uint16_t); + case Type::INT16: + INTEGER_CASE(int16_t); + case Type::UINT32: + INTEGER_CASE(uint32_t); + case Type::INT32: + INTEGER_CASE(int32_t); + case Type::UINT64: + INTEGER_CASE(uint64_t); + case Type::INT64: + INTEGER_CASE(int64_t); + case Type::FLOAT: + ConvertNumericNullableCast<float, double>(data, NAN, out_buffer); + break; + case Type::DOUBLE: + ConvertNumericNullable<double>(data, NAN, out_buffer); + break; + default: + std::stringstream ss; + ss << "Cannot write Arrow data of type " << col->type()->ToString(); + ss << " to a Pandas float64 block."; + return Status::NotImplemented(ss.str()); + } + +#undef INTEGER_CASE + + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +class BoolBlock : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status Allocate() override { return AllocateNDArray(NPY_BOOL); } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + if (type != Type::BOOL) { + std::stringstream ss; + ss << "Cannot write Arrow data of type " << col->type()->ToString(); + ss << " to a Pandas boolean block."; + return Status::NotImplemented(ss.str()); + } + + uint8_t* out_buffer = + reinterpret_cast<uint8_t*>(block_data_) + rel_placement * num_rows_; + + ConvertBooleanNoNulls(*col->data().get(), out_buffer); + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +class DatetimeBlock : public PandasBlock { + public: + using PandasBlock::PandasBlock; + Status AllocateDatetime(int ndim) { + RETURN_NOT_OK(AllocateNDArray(NPY_DATETIME, ndim)); + + PyAcquireGIL lock; + auto date_dtype = reinterpret_cast<PyArray_DatetimeDTypeMetaData*>( + PyArray_DESCR(reinterpret_cast<PyArrayObject*>(block_arr_.obj()))->c_metadata); + date_dtype->meta.base = NPY_FR_ns; + return Status::OK(); + } + + Status Allocate() override { return AllocateDatetime(2); } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + Type::type type = col->type()->id(); + + int64_t* out_buffer = + reinterpret_cast<int64_t*>(block_data_) + rel_placement * num_rows_; + + const ChunkedArray& data = *col.get()->data(); + + if (type == Type::DATE32) { + // Convert from days since epoch to datetime64[ns] + ConvertDatetimeNanos<int32_t, kNanosecondsInDay>(data, out_buffer); + } else if (type == Type::DATE64) { + // Date64Type is millisecond timestamp stored as int64_t + // TODO(wesm): Do we want to make sure to zero out the milliseconds? + ConvertDatetimeNanos<int64_t, 1000000L>(data, out_buffer); + } else if (type == Type::TIMESTAMP) { + auto ts_type = static_cast<TimestampType*>(col->type().get()); + + if (ts_type->unit() == TimeUnit::NANO) { + ConvertNumericNullable<int64_t>(data, kPandasTimestampNull, out_buffer); + } else if (ts_type->unit() == TimeUnit::MICRO) { + ConvertDatetimeNanos<int64_t, 1000L>(data, out_buffer); + } else if (ts_type->unit() == TimeUnit::MILLI) { + ConvertDatetimeNanos<int64_t, 1000000L>(data, out_buffer); + } else if (ts_type->unit() == TimeUnit::SECOND) { + ConvertDatetimeNanos<int64_t, 1000000000L>(data, out_buffer); + } else { + return Status::NotImplemented("Unsupported time unit"); + } + } else { + std::stringstream ss; + ss << "Cannot write Arrow data of type " << col->type()->ToString(); + ss << " to a Pandas datetime block."; + return Status::NotImplemented(ss.str()); + } + + placement_data_[rel_placement] = abs_placement; + return Status::OK(); + } +}; + +class DatetimeTZBlock : public DatetimeBlock { + public: + DatetimeTZBlock(const std::string& timezone, int64_t num_rows) + : DatetimeBlock(num_rows, 1), timezone_(timezone) {} + + // Like Categorical, the internal ndarray is 1-dimensional + Status Allocate() override { return AllocateDatetime(1); } + + Status GetPyResult(PyObject** output) override { + PyObject* result = PyDict_New(); + RETURN_IF_PYERROR(); + + PyObject* py_tz = PyUnicode_FromStringAndSize( + timezone_.c_str(), static_cast<Py_ssize_t>(timezone_.size())); + RETURN_IF_PYERROR(); + + PyDict_SetItemString(result, "block", block_arr_.obj()); + PyDict_SetItemString(result, "timezone", py_tz); + PyDict_SetItemString(result, "placement", placement_arr_.obj()); + + *output = result; + + return Status::OK(); + } + + private: + std::string timezone_; +}; + +template <int ARROW_INDEX_TYPE> +class CategoricalBlock : public PandasBlock { + public: + explicit CategoricalBlock(int64_t num_rows) : PandasBlock(num_rows, 1) {} + Status Allocate() override { + constexpr int npy_type = arrow_traits<ARROW_INDEX_TYPE>::npy_type; + + if (!(npy_type == NPY_INT8 || npy_type == NPY_INT16 || npy_type == NPY_INT32 || + npy_type == NPY_INT64)) { + return Status::Invalid("Category indices must be signed integers"); + } + return AllocateNDArray(npy_type, 1); + } + + Status Write(const std::shared_ptr<Column>& col, int64_t abs_placement, + int64_t rel_placement) override { + using T = typename arrow_traits<ARROW_INDEX_TYPE>::T; + + T* out_values = reinterpret_cast<T*>(block_data_) + rel_placement * num_rows_; + + const ChunkedArray& data = *col->data().get(); + + for (int c = 0; c < data.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data.chunk(c); + const auto& dict_arr = static_cast<const DictionaryArray&>(*arr); + const auto& indices = static_cast<const PrimitiveArray&>(*dict_arr.indices()); + auto in_values = reinterpret_cast<const T*>(indices.raw_values()); + + // Null is -1 in CategoricalBlock + for (int i = 0; i < arr->length(); ++i) { + *out_values++ = indices.IsNull(i) ? -1 : in_values[i]; + } + } + + placement_data_[rel_placement] = abs_placement; + + auto dict_type = static_cast<const DictionaryType*>(col->type().get()); + + PyObject* dict; + RETURN_NOT_OK(ConvertArrayToPandas(dict_type->dictionary(), nullptr, &dict)); + dictionary_.reset(dict); + + return Status::OK(); + } + + Status GetPyResult(PyObject** output) override { + PyObject* result = PyDict_New(); + RETURN_IF_PYERROR(); + + PyDict_SetItemString(result, "block", block_arr_.obj()); + PyDict_SetItemString(result, "dictionary", dictionary_.obj()); + PyDict_SetItemString(result, "placement", placement_arr_.obj()); + + *output = result; + + return Status::OK(); + } + + protected: + OwnedRef dictionary_; +}; + +Status MakeBlock(PandasBlock::type type, int64_t num_rows, int num_columns, + std::shared_ptr<PandasBlock>* block) { +#define BLOCK_CASE(NAME, TYPE) \ + case PandasBlock::NAME: \ + *block = std::make_shared<TYPE>(num_rows, num_columns); \ + break; + + switch (type) { + BLOCK_CASE(OBJECT, ObjectBlock); + BLOCK_CASE(UINT8, UInt8Block); + BLOCK_CASE(INT8, Int8Block); + BLOCK_CASE(UINT16, UInt16Block); + BLOCK_CASE(INT16, Int16Block); + BLOCK_CASE(UINT32, UInt32Block); + BLOCK_CASE(INT32, Int32Block); + BLOCK_CASE(UINT64, UInt64Block); + BLOCK_CASE(INT64, Int64Block); + BLOCK_CASE(FLOAT, Float32Block); + BLOCK_CASE(DOUBLE, Float64Block); + BLOCK_CASE(BOOL, BoolBlock); + BLOCK_CASE(DATETIME, DatetimeBlock); + default: + return Status::NotImplemented("Unsupported block type"); + } + +#undef BLOCK_CASE + + return (*block)->Allocate(); +} + +static inline Status MakeCategoricalBlock(const std::shared_ptr<DataType>& type, + int64_t num_rows, std::shared_ptr<PandasBlock>* block) { + // All categoricals become a block with a single column + auto dict_type = static_cast<const DictionaryType*>(type.get()); + switch (dict_type->index_type()->id()) { + case Type::INT8: + *block = std::make_shared<CategoricalBlock<Type::INT8>>(num_rows); + break; + case Type::INT16: + *block = std::make_shared<CategoricalBlock<Type::INT16>>(num_rows); + break; + case Type::INT32: + *block = std::make_shared<CategoricalBlock<Type::INT32>>(num_rows); + break; + case Type::INT64: + *block = std::make_shared<CategoricalBlock<Type::INT64>>(num_rows); + break; + default: { + std::stringstream ss; + ss << "Categorical index type not implemented: " + << dict_type->index_type()->ToString(); + return Status::NotImplemented(ss.str()); + } + } + return (*block)->Allocate(); +} + +using BlockMap = std::unordered_map<int, std::shared_ptr<PandasBlock>>; + +// Construct the exact pandas 0.x "BlockManager" memory layout +// +// * For each column determine the correct output pandas type +// * Allocate 2D blocks (ncols x nrows) for each distinct data type in output +// * Allocate block placement arrays +// * Write Arrow columns out into each slice of memory; populate block +// * placement arrays as we go +class DataFrameBlockCreator { + public: + explicit DataFrameBlockCreator(const std::shared_ptr<Table>& table) : table_(table) {} + + Status Convert(int nthreads, PyObject** output) { + column_types_.resize(table_->num_columns()); + column_block_placement_.resize(table_->num_columns()); + type_counts_.clear(); + blocks_.clear(); + + RETURN_NOT_OK(CreateBlocks()); + RETURN_NOT_OK(WriteTableToBlocks(nthreads)); + + return GetResultList(output); + } + + Status CreateBlocks() { + for (int i = 0; i < table_->num_columns(); ++i) { + std::shared_ptr<Column> col = table_->column(i); + PandasBlock::type output_type; + + Type::type column_type = col->type()->id(); + switch (column_type) { + case Type::BOOL: + output_type = col->null_count() > 0 ? PandasBlock::OBJECT : PandasBlock::BOOL; + break; + case Type::UINT8: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::UINT8; + break; + case Type::INT8: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::INT8; + break; + case Type::UINT16: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::UINT16; + break; + case Type::INT16: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::INT16; + break; + case Type::UINT32: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::UINT32; + break; + case Type::INT32: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::INT32; + break; + case Type::INT64: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::INT64; + break; + case Type::UINT64: + output_type = col->null_count() > 0 ? PandasBlock::DOUBLE : PandasBlock::UINT64; + break; + case Type::FLOAT: + output_type = PandasBlock::FLOAT; + break; + case Type::DOUBLE: + output_type = PandasBlock::DOUBLE; + break; + case Type::NA: + case Type::STRING: + case Type::BINARY: + case Type::FIXED_SIZE_BINARY: + case Type::STRUCT: + case Type::TIME32: + case Type::TIME64: + case Type::DECIMAL: + output_type = PandasBlock::OBJECT; + break; + case Type::DATE32: + output_type = PandasBlock::DATETIME; + break; + case Type::DATE64: + output_type = PandasBlock::DATETIME; + break; + case Type::TIMESTAMP: { + const auto& ts_type = static_cast<const TimestampType&>(*col->type()); + if (ts_type.timezone() != "") { + output_type = PandasBlock::DATETIME_WITH_TZ; + } else { + output_type = PandasBlock::DATETIME; + } + } break; + case Type::LIST: { + auto list_type = std::static_pointer_cast<ListType>(col->type()); + if (!ListTypeSupported(*list_type->value_type())) { + std::stringstream ss; + ss << "Not implemented type for list in DataFrameBlock: " + << list_type->value_type()->ToString(); + return Status::NotImplemented(ss.str()); + } + output_type = PandasBlock::OBJECT; + } break; + case Type::DICTIONARY: + output_type = PandasBlock::CATEGORICAL; + break; + default: + std::stringstream ss; + ss << "No known equivalent Pandas block for Arrow data of type "; + ss << col->type()->ToString() << " is known."; + return Status::NotImplemented(ss.str()); + } + + int block_placement = 0; + std::shared_ptr<PandasBlock> block; + if (output_type == PandasBlock::CATEGORICAL) { + RETURN_NOT_OK(MakeCategoricalBlock(col->type(), table_->num_rows(), &block)); + categorical_blocks_[i] = block; + } else if (output_type == PandasBlock::DATETIME_WITH_TZ) { + const auto& ts_type = static_cast<const TimestampType&>(*col->type()); + block = std::make_shared<DatetimeTZBlock>(ts_type.timezone(), table_->num_rows()); + RETURN_NOT_OK(block->Allocate()); + datetimetz_blocks_[i] = block; + } else { + auto it = type_counts_.find(output_type); + if (it != type_counts_.end()) { + block_placement = it->second; + // Increment count + it->second += 1; + } else { + // Add key to map + type_counts_[output_type] = 1; + } + } + + column_types_[i] = output_type; + column_block_placement_[i] = block_placement; + } + + // Create normal non-categorical blocks + for (const auto& it : type_counts_) { + PandasBlock::type type = static_cast<PandasBlock::type>(it.first); + std::shared_ptr<PandasBlock> block; + RETURN_NOT_OK(MakeBlock(type, table_->num_rows(), it.second, &block)); + blocks_[type] = block; + } + return Status::OK(); + } + + Status WriteTableToBlocks(int nthreads) { + auto WriteColumn = [this](int i) { + std::shared_ptr<Column> col = this->table_->column(i); + PandasBlock::type output_type = this->column_types_[i]; + + int rel_placement = this->column_block_placement_[i]; + + std::shared_ptr<PandasBlock> block; + if (output_type == PandasBlock::CATEGORICAL) { + auto it = this->categorical_blocks_.find(i); + if (it == this->blocks_.end()) { + return Status::KeyError("No categorical block allocated"); + } + block = it->second; + } else if (output_type == PandasBlock::DATETIME_WITH_TZ) { + auto it = this->datetimetz_blocks_.find(i); + if (it == this->datetimetz_blocks_.end()) { + return Status::KeyError("No datetimetz block allocated"); + } + block = it->second; + } else { + auto it = this->blocks_.find(output_type); + if (it == this->blocks_.end()) { return Status::KeyError("No block allocated"); } + block = it->second; + } + return block->Write(col, i, rel_placement); + }; + + nthreads = std::min<int>(nthreads, table_->num_columns()); + + if (nthreads == 1) { + for (int i = 0; i < table_->num_columns(); ++i) { + RETURN_NOT_OK(WriteColumn(i)); + } + } else { + std::vector<std::thread> thread_pool; + thread_pool.reserve(nthreads); + std::atomic<int> task_counter(0); + + std::mutex error_mtx; + bool error_occurred = false; + Status error; + + for (int thread_id = 0; thread_id < nthreads; ++thread_id) { + thread_pool.emplace_back( + [this, &error, &error_occurred, &error_mtx, &task_counter, &WriteColumn]() { + int column_num; + while (!error_occurred) { + column_num = task_counter.fetch_add(1); + if (column_num >= this->table_->num_columns()) { break; } + Status s = WriteColumn(column_num); + if (!s.ok()) { + std::lock_guard<std::mutex> lock(error_mtx); + error_occurred = true; + error = s; + break; + } + } + }); + } + for (auto&& thread : thread_pool) { + thread.join(); + } + + if (error_occurred) { return error; } + } + return Status::OK(); + } + + Status AppendBlocks(const BlockMap& blocks, PyObject* list) { + for (const auto& it : blocks) { + PyObject* item; + RETURN_NOT_OK(it.second->GetPyResult(&item)); + if (PyList_Append(list, item) < 0) { RETURN_IF_PYERROR(); } + + // ARROW-1017; PyList_Append increments object refcount + Py_DECREF(item); + } + return Status::OK(); + } + + Status GetResultList(PyObject** out) { + PyAcquireGIL lock; + + PyObject* result = PyList_New(0); + RETURN_IF_PYERROR(); + + RETURN_NOT_OK(AppendBlocks(blocks_, result)); + RETURN_NOT_OK(AppendBlocks(categorical_blocks_, result)); + RETURN_NOT_OK(AppendBlocks(datetimetz_blocks_, result)); + + *out = result; + return Status::OK(); + } + + private: + std::shared_ptr<Table> table_; + + // column num -> block type id + std::vector<PandasBlock::type> column_types_; + + // column num -> relative placement within internal block + std::vector<int> column_block_placement_; + + // block type -> type count + std::unordered_map<int, int> type_counts_; + + // block type -> block + BlockMap blocks_; + + // column number -> categorical block + BlockMap categorical_blocks_; + + // column number -> datetimetz block + BlockMap datetimetz_blocks_; +}; + +class ArrowDeserializer { + public: + ArrowDeserializer(const std::shared_ptr<Column>& col, PyObject* py_ref) + : col_(col), data_(*col->data().get()), py_ref_(py_ref) {} + + Status AllocateOutput(int type) { + PyAcquireGIL lock; + + result_ = NewArray1DFromType(col_->type().get(), type, col_->length(), nullptr); + arr_ = reinterpret_cast<PyArrayObject*>(result_); + return Status::OK(); + } + + template <int TYPE> + Status ConvertValuesZeroCopy(int npy_type, std::shared_ptr<Array> arr) { + typedef typename arrow_traits<TYPE>::T T; + + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const T*>(prim_arr->raw_values()); + + // Zero-Copy. We can pass the data pointer directly to NumPy. + void* data = const_cast<T*>(in_values); + + PyAcquireGIL lock; + + // Zero-Copy. We can pass the data pointer directly to NumPy. + result_ = NewArray1DFromType(col_->type().get(), npy_type, col_->length(), data); + arr_ = reinterpret_cast<PyArrayObject*>(result_); + + if (arr_ == NULL) { + // Error occurred, trust that error set + return Status::OK(); + } + + if (PyArray_SetBaseObject(arr_, py_ref_) == -1) { + // Error occurred, trust that SetBaseObject set the error state + return Status::OK(); + } else { + // PyArray_SetBaseObject steals our reference to py_ref_ + Py_INCREF(py_ref_); + } + + // Arrow data is immutable. + PyArray_CLEARFLAGS(arr_, NPY_ARRAY_WRITEABLE); + + // Arrow data is owned by another + PyArray_CLEARFLAGS(arr_, NPY_ARRAY_OWNDATA); + + return Status::OK(); + } + + // ---------------------------------------------------------------------- + // Allocate new array and deserialize. Can do a zero copy conversion for some + // types + + template <typename Type> + typename std::enable_if<std::is_base_of<FloatingPoint, Type>::value, Status>::type + Visit(const Type& type) { + constexpr int TYPE = Type::type_id; + using traits = arrow_traits<TYPE>; + + typedef typename traits::T T; + int npy_type = traits::npy_type; + + if (data_.num_chunks() == 1 && data_.null_count() == 0 && py_ref_ != nullptr) { + return ConvertValuesZeroCopy<TYPE>(npy_type, data_.chunk(0)); + } + + RETURN_NOT_OK(AllocateOutput(npy_type)); + auto out_values = reinterpret_cast<T*>(PyArray_DATA(arr_)); + ConvertNumericNullable<T>(data_, traits::na_value, out_values); + + return Status::OK(); + } + + template <typename Type> + typename std::enable_if<std::is_base_of<DateType, Type>::value || + std::is_base_of<TimestampType, Type>::value, + Status>::type + Visit(const Type& type) { + constexpr int TYPE = Type::type_id; + using traits = arrow_traits<TYPE>; + + typedef typename traits::T T; + + RETURN_NOT_OK(AllocateOutput(traits::npy_type)); + auto out_values = reinterpret_cast<T*>(PyArray_DATA(arr_)); + + constexpr T na_value = traits::na_value; + constexpr int64_t kShift = traits::npy_shift; + + for (int c = 0; c < data_.num_chunks(); c++) { + const std::shared_ptr<Array> arr = data_.chunk(c); + auto prim_arr = static_cast<PrimitiveArray*>(arr.get()); + auto in_values = reinterpret_cast<const T*>(prim_arr->raw_values()); + + for (int64_t i = 0; i < arr->length(); ++i) { + *out_values++ = arr->IsNull(i) ? na_value : in_values[i] / kShift; + } + } + return Status::OK(); + } + + template <typename Type> + typename std::enable_if<std::is_base_of<TimeType, Type>::value, Status>::type Visit( + const Type& type) { + return Status::NotImplemented("Don't know how to serialize Arrow time type to NumPy"); + } + + // Integer specialization + template <typename Type> + typename std::enable_if<std::is_base_of<Integer, Type>::value, Status>::type Visit( + const Type& type) { + constexpr int TYPE = Type::type_id; + using traits = arrow_traits<TYPE>; + + typedef typename traits::T T; + + if (data_.num_chunks() == 1 && data_.null_count() == 0 && py_ref_ != nullptr) { + return ConvertValuesZeroCopy<TYPE>(traits::npy_type, data_.chunk(0)); + } + + if (data_.null_count() > 0) { + RETURN_NOT_OK(AllocateOutput(NPY_FLOAT64)); + auto out_values = reinterpret_cast<double*>(PyArray_DATA(arr_)); + ConvertIntegerWithNulls<T>(data_, out_values); + } else { + RETURN_NOT_OK(AllocateOutput(traits::npy_type)); + auto out_values = reinterpret_cast<T*>(PyArray_DATA(arr_)); + ConvertIntegerNoNullsSameType<T>(data_, out_values); + } + + return Status::OK(); + } + + template <typename FUNCTOR> + inline Status VisitObjects(FUNCTOR func) { + RETURN_NOT_OK(AllocateOutput(NPY_OBJECT)); + auto out_values = reinterpret_cast<PyObject**>(PyArray_DATA(arr_)); + return func(data_, out_values); + } + + // UTF8 strings + template <typename Type> + typename std::enable_if<std::is_base_of<BinaryType, Type>::value, Status>::type Visit( + const Type& type) { + return VisitObjects(ConvertBinaryLike<Type>); + } + + Status Visit(const NullType& type) { return VisitObjects(ConvertNulls); } + + // Fixed length binary strings + Status Visit(const FixedSizeBinaryType& type) { + return VisitObjects(ConvertFixedSizeBinary); + } + + Status Visit(const DecimalType& type) { return VisitObjects(ConvertDecimals); } + + Status Visit(const Time32Type& type) { return VisitObjects(ConvertTimes<Time32Type>); } + + Status Visit(const Time64Type& type) { return VisitObjects(ConvertTimes<Time64Type>); } + + Status Visit(const StructType& type) { return VisitObjects(ConvertStruct); } + + // Boolean specialization + Status Visit(const BooleanType& type) { + if (data_.null_count() > 0) { + return VisitObjects(ConvertBooleanWithNulls); + } else { + RETURN_NOT_OK(AllocateOutput(arrow_traits<Type::BOOL>::npy_type)); + auto out_values = reinterpret_cast<uint8_t*>(PyArray_DATA(arr_)); + ConvertBooleanNoNulls(data_, out_values); + } + return Status::OK(); + } + + Status Visit(const ListType& type) { +#define CONVERTVALUES_LISTSLIKE_CASE(ArrowType, ArrowEnum) \ + case Type::ArrowEnum: \ + return ConvertListsLike<ArrowType>(col_, out_values); + + RETURN_NOT_OK(AllocateOutput(NPY_OBJECT)); + auto out_values = reinterpret_cast<PyObject**>(PyArray_DATA(arr_)); + auto list_type = std::static_pointer_cast<ListType>(col_->type()); + switch (list_type->value_type()->id()) { + CONVERTVALUES_LISTSLIKE_CASE(UInt8Type, UINT8) + CONVERTVALUES_LISTSLIKE_CASE(Int8Type, INT8) + CONVERTVALUES_LISTSLIKE_CASE(UInt16Type, UINT16) + CONVERTVALUES_LISTSLIKE_CASE(Int16Type, INT16) + CONVERTVALUES_LISTSLIKE_CASE(UInt32Type, UINT32) + CONVERTVALUES_LISTSLIKE_CASE(Int32Type, INT32) + CONVERTVALUES_LISTSLIKE_CASE(UInt64Type, UINT64) + CONVERTVALUES_LISTSLIKE_CASE(Int64Type, INT64) + CONVERTVALUES_LISTSLIKE_CASE(TimestampType, TIMESTAMP) + CONVERTVALUES_LISTSLIKE_CASE(FloatType, FLOAT) + CONVERTVALUES_LISTSLIKE_CASE(DoubleType, DOUBLE) + CONVERTVALUES_LISTSLIKE_CASE(StringType, STRING) + CONVERTVALUES_LISTSLIKE_CASE(DecimalType, DECIMAL) + CONVERTVALUES_LISTSLIKE_CASE(ListType, LIST) + default: { + std::stringstream ss; + ss << "Not implemented type for lists: " << list_type->value_type()->ToString(); + return Status::NotImplemented(ss.str()); + } + } +#undef CONVERTVALUES_LISTSLIKE_CASE + } + + Status Visit(const DictionaryType& type) { + std::shared_ptr<PandasBlock> block; + RETURN_NOT_OK(MakeCategoricalBlock(col_->type(), col_->length(), &block)); + RETURN_NOT_OK(block->Write(col_, 0, 0)); + + auto dict_type = static_cast<const DictionaryType*>(col_->type().get()); + + PyAcquireGIL lock; + result_ = PyDict_New(); + RETURN_IF_PYERROR(); + + PyObject* dictionary; + + // Release GIL before calling ConvertArrayToPandas, will be reacquired + // there if needed + lock.release(); + RETURN_NOT_OK(ConvertArrayToPandas(dict_type->dictionary(), nullptr, &dictionary)); + lock.acquire(); + + PyDict_SetItemString(result_, "indices", block->block_arr()); + PyDict_SetItemString(result_, "dictionary", dictionary); + + return Status::OK(); + } + + Status Visit(const UnionType& type) { return Status::NotImplemented("union type"); } + + Status Convert(PyObject** out) { + RETURN_NOT_OK(VisitTypeInline(*col_->type(), this)); + *out = result_; + return Status::OK(); + } + + private: + std::shared_ptr<Column> col_; + const ChunkedArray& data_; + PyObject* py_ref_; + PyArrayObject* arr_; + PyObject* result_; +}; + +Status ConvertArrayToPandas( + const std::shared_ptr<Array>& arr, PyObject* py_ref, PyObject** out) { + static std::string dummy_name = "dummy"; + auto field = std::make_shared<Field>(dummy_name, arr->type()); + auto col = std::make_shared<Column>(field, arr); + return ConvertColumnToPandas(col, py_ref, out); +} + +Status ConvertColumnToPandas( + const std::shared_ptr<Column>& col, PyObject* py_ref, PyObject** out) { + ArrowDeserializer converter(col, py_ref); + return converter.Convert(out); +} + +Status ConvertTableToPandas( + const std::shared_ptr<Table>& table, int nthreads, PyObject** out) { + DataFrameBlockCreator helper(table); + return helper.Convert(nthreads, out); +} + +} // namespace py +} // namespace arrow http://git-wip-us.apache.org/repos/asf/arrow/blob/2c5b412c/cpp/src/arrow/python/arrow_to_pandas.h ---------------------------------------------------------------------- diff --git a/cpp/src/arrow/python/arrow_to_pandas.h b/cpp/src/arrow/python/arrow_to_pandas.h new file mode 100644 index 0000000..c606dcb --- /dev/null +++ b/cpp/src/arrow/python/arrow_to_pandas.h @@ -0,0 +1,67 @@ +// 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. + +// Functions for converting between pandas's NumPy-based data representation +// and Arrow data structures + +#ifndef ARROW_PYTHON_ADAPTERS_PANDAS_H +#define ARROW_PYTHON_ADAPTERS_PANDAS_H + +#include "arrow/python/platform.h" + +#include <memory> +#include <string> + +#include "arrow/util/visibility.h" + +namespace arrow { + +class Array; +class Column; +class DataType; +class MemoryPool; +class Status; +class Table; + +namespace py { + +ARROW_EXPORT +Status ConvertArrayToPandas( + const std::shared_ptr<Array>& arr, PyObject* py_ref, PyObject** out); + +ARROW_EXPORT +Status ConvertColumnToPandas( + const std::shared_ptr<Column>& col, PyObject* py_ref, PyObject** out); + +struct PandasOptions { + bool strings_to_categorical; +}; + +// Convert a whole table as efficiently as possible to a pandas.DataFrame. +// +// The returned Python object is a list of tuples consisting of the exact 2D +// BlockManager structure of the pandas.DataFrame used as of pandas 0.19.x. +// +// tuple item: (indices: ndarray[int32], block: ndarray[TYPE, ndim=2]) +ARROW_EXPORT +Status ConvertTableToPandas( + const std::shared_ptr<Table>& table, int nthreads, PyObject** out); + +} // namespace py +} // namespace arrow + +#endif // ARROW_PYTHON_ADAPTERS_PANDAS_H
