tiled.client.container.Container.write_sparse

Container.write_sparse(coords, data, shape, *, key=None, metadata=None, dims=None, specs=None)[source]

EXPERIMENTAL: Write a sparse array.

Parameters:

coordsarray-like
dataarray-like
shapetuple
keystr, optional: Key (name) for this new node. If None, the server will provide a unique key.
metadatadict, optional: User metadata. May be nested. Must contain only basic types (e.g. numbers, strings, lists, dicts) that are JSON-serializable.
dimsList[str], optional: A label for each dimension of the array.
specsList[Spec], optional: List of names that are used to label that the data and/or metadata conform to some named standard specification.

Examples

Write a sparse.COO array.

>>> import sparse
>>> from numpy import array
>>> coo = sparse.COO(coords=array([[2, 5]]), data=array([1.3, 7.5]), shape=(10,))
>>> c.write_sparse(coords=coo.coords, data=coo.data, shape=coo.shape)

This only supports a single chunk. For chunked upload, use lower-level methods.

# Define the overall shape and the dimensions of each chunk. >>> from tiled.structures.sparse import COOStructure >>> structure = COOStructure(shape=(10,), chunks=((5, 5),)) >>> data_source = DataSource(structure=structure, structure_family=StructureFamily.sparse) >>> x = c.new(“sparse”, [data_source]) # Upload the data in each chunk. # Coords are given with in the reference frame of each chunk. >>> x.write_block(coords=array([[2, 4]]), data=array([3.1, 2.8]), block=(0,)) >>> x.write_block(coords=array([[0, 1]]), data=array([6.7, 1.2]), block=(1,))