tensorstore.TensorStore.__getitem__(self, transform: IndexTransform) → TensorStore

Computes a virtual view using an explicit index transform.

Example

>>> dataset = await ts.open(
...     {
...         'driver': 'zarr',
...         'kvstore': {
...             'driver': 'memory'
...         }
...     },
...     dtype=ts.uint32,
...     shape=[70, 80],
...     create=True)
>>> transform = ts.IndexTransform(
...     input_shape=[3],
...     output=[
...         ts.OutputIndexMap(index_array=[1, 2, 3]),
...         ts.OutputIndexMap(index_array=[5, 4, 3])
...     ])
>>> dataset[transform]
TensorStore({
  'context': {
    'cache_pool': {},
    'data_copy_concurrency': {},
    'memory_key_value_store': {},
  },
  'driver': 'zarr',
  'dtype': 'uint32',
  'kvstore': {'driver': 'memory'},
  'metadata': {
    'chunks': [70, 80],
    'compressor': {
      'blocksize': 0,
      'clevel': 5,
      'cname': 'lz4',
      'id': 'blosc',
      'shuffle': -1,
    },
    'dimension_separator': '.',
    'dtype': '<u4',
    'fill_value': None,
    'filters': None,
    'order': 'C',
    'shape': [70, 80],
    'zarr_format': 2,
  },
  'transform': {
    'input_exclusive_max': [3],
    'input_inclusive_min': [0],
    'output': [{'index_array': [1, 2, 3]}, {'index_array': [5, 4, 3]}],
  },
})
>>> await dataset[transform].write([1, 2, 3])
>>> await dataset[1:6, 1:6].read()
array([[0, 0, 0, 0, 1],
       [0, 0, 0, 2, 0],
       [0, 0, 3, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0]], dtype=uint32)
Parameters:
transform: IndexTransform

Index transform, transform.output_rank must equal self.rank.

Returns:

View of rank transform.input_rank and domain self.domain[transform].

This is the most general form of indexing, to which all other indexing methods reduce:

  • self[expr] is equivalent to self[ts.IndexTransform(self.domain)[expr]]

  • self.oindex[expr] is equivalent to self[ts.IndexTransform(self.domain).oindex[expr]]

  • self.vindex[expr] is equivalent to self[ts.IndexTransform(self.domain).vindex[expr]]

In most cases it is more convenient to use one of those other indexing forms instead.

See also