ArraySource

ArraySource(
    path: str = ““,
    format: str =”“,
    version: str =”“,
    key: str =”“,
    windows: tuple[tuple[int, int], …] = (),
    shape: tuple[int, …] = (),
    dtype: Any = None,
    origin_id: str =”“,
    extent: tuple[int, …] = (),
    filled: bool = False,
    value: Any = None,
)-> None

Where an array is stored, or the constant which fills it, and enough to load it without its patch.

Holds no data and opens nothing until load. A reader sets key; the I/O framework fills in the rest.

Parameters

Parameter	Description
path	The resource holding the array.
format	The name of the FiberIO which reads the resource.
version	The version of that format.
key	Which array in the resource: the logical patch of a multi-patch resource, or an absolute path (“/…”) to any stored array, such as a dense coordinate.
windows	A half-open `(start, stop)` sample range for each axis.
shape	The shape of the array the windows select.
dtype	The dtype of the loaded array.
origin_id	The id of the whole array, which a window’s id builds on. The framework gives a patch’s data array the patch’s `origin_id`; a caller which knows better, such as a hash of the contents, gives that. Empty means “by location”: path, format, version and key.
extent	The shape of the whole array, which says when the windows select all of it.
filled	Whether the array is `value` throughout rather than stored. Its own flag because a table cannot say so by the value: NaN stores as NULL.
value	The constant of a `filled` array. Such a source has no path and reads nothing; see `full`.

Examples

import numpy as np
import dascore as dc
from dascore.core.source import ArraySource
from dascore.utils.downloader import fetch

patch = dc.read(fetch("example_dasdae_event_1.h5"))[0]
source = patch._source
# Slicing reads nothing and gives another source.
sub = source[10:20]
assert sub.shape[0] == 10 and sub.data_id != source.data_id
assert sub.load().shape == sub.shape

# A constant source generates its array instead of reading one.
constant = ArraySource.full((2, 3), np.nan)
assert np.isnan(constant.load()).all()

Methods

Name	Description
describe	Return a source for the whole of an array of this shape and dtype.
detach	Return the provenance alone, for an array this no longer loads.
load	Read and return the array.
narrow	Return the source `indexer` selects, detached if not contiguous.
to_dict	Return a JSON-compatible dict which `from_dict` reads back.