You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "rok (via GitHub)" <gi...@apache.org> on 2023/04/06 10:14:33 UTC
[GitHub] [arrow] rok commented on a diff in pull request #34883: GH-34882: [Python] Binding for FixedShapeTensorType
rok commented on code in PR #34883:
URL: https://github.com/apache/arrow/pull/34883#discussion_r1159572171
##########
python/pyarrow/array.pxi:
##########
@@ -3076,6 +3076,111 @@ cdef class ExtensionArray(Array):
return Array._to_pandas(self.storage, options, **kwargs)
+class FixedShapeTensorArray(ExtensionArray):
+ """
+ Concrete class for fixed shape tensor extension arrays.
+
+ Examples
+ --------
+ Define the extension type for tensor array
+
+ >>> import pyarrow as pa
+ >>> tensor_type = pa.fixed_shape_tensor(pa.int32(), [2, 2])
+
+ Create an extension array
+
+ >>> arr = [[1, 2, 3, 4], [10, 20, 30, 40], [100, 200, 300, 400]]
+ >>> storage = pa.array(arr, pa.list_(pa.int32(), 4))
+ >>> pa.ExtensionArray.from_storage(tensor_type, storage)
+ <pyarrow.lib.FixedShapeTensorArray object at ...>
+ [
+ [
+ 1,
+ 2,
+ 3,
+ 4
+ ],
+ [
+ 10,
+ 20,
+ 30,
+ 40
+ ],
+ [
+ 100,
+ 200,
+ 300,
+ 400
+ ]
+ ]
+ """
+
+ def to_numpy_ndarray(self):
+ """
+ Convert fixed shape tensor extension array to a numpy array (with dim+1).
+ """
+ np_flat = np.asarray(self.storage.values)
+ numpy_tensor = np_flat.reshape((len(self),) + tuple(self.type.shape),
+ order='C')
Review Comment:
Does reshape incur a copy? Perhaps we should throw if `permutation[0] != 0` (as that would mean array can't be chunked)?
Also shouldn't we take permutation into account when reshaping?
##########
python/pyarrow/array.pxi:
##########
@@ -3076,6 +3076,111 @@ cdef class ExtensionArray(Array):
return Array._to_pandas(self.storage, options, **kwargs)
+class FixedShapeTensorArray(ExtensionArray):
+ """
+ Concrete class for fixed shape tensor extension arrays.
+
+ Examples
+ --------
+ Define the extension type for tensor array
+
+ >>> import pyarrow as pa
+ >>> tensor_type = pa.fixed_shape_tensor(pa.int32(), [2, 2])
+
+ Create an extension array
+
+ >>> arr = [[1, 2, 3, 4], [10, 20, 30, 40], [100, 200, 300, 400]]
+ >>> storage = pa.array(arr, pa.list_(pa.int32(), 4))
+ >>> pa.ExtensionArray.from_storage(tensor_type, storage)
+ <pyarrow.lib.FixedShapeTensorArray object at ...>
+ [
+ [
+ 1,
+ 2,
+ 3,
+ 4
+ ],
+ [
+ 10,
+ 20,
+ 30,
+ 40
+ ],
+ [
+ 100,
+ 200,
+ 300,
+ 400
+ ]
+ ]
+ """
+
+ def to_numpy_ndarray(self):
+ """
+ Convert fixed shape tensor extension array to a numpy array (with dim+1).
+ """
+ np_flat = np.asarray(self.storage.values)
+ numpy_tensor = np_flat.reshape((len(self),) + tuple(self.type.shape),
+ order='C')
+
+ return numpy_tensor
+
+ @staticmethod
+ def from_numpy_ndarray(obj):
+ """
+ Convert numpy tensors (ndarrays) to a fixed shape tensor extension array.
+ The first dimension of ndarray will become the length of the fixed
+ shape tensor array.
+
+ Numpy array needs to be C-contiguous in memory
+ (``obj.flags["C_CONTIGUOUS"]==True``).
+
+ Parameters
+ ----------
+ obj : ndarray
+
+ Examples
+ --------
+ >>> import pyarrow as pa
+ >>> import numpy as np
+ >>> arr = np.array(
+ ... [[[1, 2, 3], [4, 5, 6]], [[1, 2, 3], [4, 5, 6]]],
+ ... dtype=np.float32)
+ >>> pa.FixedShapeTensorArray.from_numpy_ndarray(arr)
+ <pyarrow.lib.FixedShapeTensorArray object at ...>
+ [
+ [
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6
+ ],
+ [
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6
+ ]
+ ]
+ """
+ if obj.flags["F_CONTIGUOUS"]:
+ raise ValueError('The data in the numpy array need to be in a single, '
+ 'C-style contiguous segment.')
Review Comment:
I don't remember of the top of my head how strides work in numpy, but could we extract permutation at this point and use it instead of forcing C-style? (see proposal for FromTensor for what I mean: https://github.com/apache/arrow/pull/34797/files#diff-47327ebe34666ff7683a620f0210f7d1a1388121ba787cc5ea7cc113f4f66dadR154-R165)
##########
python/pyarrow/types.pxi:
##########
@@ -4543,6 +4623,105 @@ def run_end_encoded(run_end_type, value_type):
return pyarrow_wrap_data_type(ree_type)
+def fixed_shape_tensor(DataType value_type, shape, dim_names=None, permutation=None):
+ """
+ Create instance of fixed shape tensor extension type with shape and optional
+ names of tensor dimensions and indices of the desired ordering.
Review Comment:
```suggestion
names of tensor dimensions and indices of the desired logical
ordering of dimensions.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org