MCPcopy Create free account
hub / github.com/apache/arrow / NdarraysToSparseCSFTensor

Function NdarraysToSparseCSFTensor

python/pyarrow/src/arrow/python/numpy_convert.cc:474–526  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

472}
473
474Status NdarraysToSparseCSFTensor(MemoryPool* pool, PyObject* data_ao, PyObject* indptr_ao,
475 PyObject* indices_ao, const std::vector<int64_t>& shape,
476 const std::vector<int64_t>& axis_order,
477 const std::vector<std::string>& dim_names,
478 std::shared_ptr<SparseCSFTensor>* out) {
479 if (!PyArray_Check(data_ao)) {
480 return Status::TypeError("Did not pass ndarray object for data");
481 }
482 const int ndim = static_cast<const int>(shape.size());
483 PyArrayObject* ndarray_data = reinterpret_cast<PyArrayObject*>(data_ao);
484 std::shared_ptr<Buffer> data = std::make_shared<NumPyBuffer>(data_ao);
485 ARROW_ASSIGN_OR_RAISE(
486 auto type_data,
487 GetTensorType(reinterpret_cast<PyObject*>(PyArray_DESCR(ndarray_data))));
488
489 std::vector<std::shared_ptr<Tensor>> indptr(ndim - 1);
490 std::vector<std::shared_ptr<Tensor>> indices(ndim);
491
492 for (int i = 0; i < ndim - 1; ++i) {
493#ifdef Py_GIL_DISABLED
494 PyObject* item = PySequence_ITEM(indptr_ao, i);
495 RETURN_IF_PYERROR();
496 OwnedRef item_ref(item);
497#else
498 PyObject* item = PySequence_Fast_GET_ITEM(indptr_ao, i);
499#endif
500 if (!PyArray_Check(item)) {
501 return Status::TypeError("Did not pass ndarray object for indptr");
502 }
503 RETURN_NOT_OK(NdarrayToTensor(pool, item, {}, &indptr[i]));
504 ARROW_CHECK_EQ(indptr[i]->type_id(), Type::INT64); // Should be ensured by caller
505 }
506
507 for (int i = 0; i < ndim; ++i) {
508#ifdef Py_GIL_DISABLED
509 PyObject* item = PySequence_ITEM(indices_ao, i);
510 RETURN_IF_PYERROR();
511 OwnedRef item_ref(item);
512#else
513 PyObject* item = PySequence_Fast_GET_ITEM(indices_ao, i);
514#endif
515 if (!PyArray_Check(item)) {
516 return Status::TypeError("Did not pass ndarray object for indices");
517 }
518 RETURN_NOT_OK(NdarrayToTensor(pool, item, {}, &indices[i]));
519 ARROW_CHECK_EQ(indices[i]->type_id(), Type::INT64); // Should be ensured by caller
520 }
521
522 auto sparse_index = std::make_shared<SparseCSFIndex>(indptr, indices, axis_order);
523 *out = std::make_shared<SparseTensorImpl<SparseCSFIndex>>(sparse_index, type_data, data,
524 shape, dim_names);
525 return Status::OK();
526}
527
528Status NdarraysToSparseCSRMatrix(MemoryPool* pool, PyObject* data_ao, PyObject* indptr_ao,
529 PyObject* indices_ao, const std::vector<int64_t>& shape,

Callers

nothing calls this directly

Calls 5

NdarrayToTensorFunction · 0.85
TypeErrorFunction · 0.50
OKFunction · 0.50
sizeMethod · 0.45
type_idMethod · 0.45

Tested by

no test coverage detected