.download() fails on TiledDataset

[Cadair]

import dkist
from dkist.data.sample import VBI_AJQWW

ds = dkist.load_dataset(VBI_AJQWW)
ds.files.download()

INFO: Fetching updated search values for the DKIST client to /home/stuart/.local/share/dkist/api_search_values.json [dkist.net.attrs_values]
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[4], line 1
----> 1 ds.files.download()

File ~/Git/DKIST/dkist/dkist/io/file_manager.py:430, in FileManager.download(self, path, destination_endpoint, progress, wait, label)
    427 from dkist.net.helpers import _orchestrate_transfer_task
    429 if self._ndcube is None:
--> 430     raise ValueError(
    431         "This file manager has no associated Dataset object, so the data can not be downloaded."
    432     )
    434 inv = self._ndcube.meta["inventory"]
    435 path_inv = path_format_inventory(humanize_inventory(inv))

ValueError: This file manager has no associated Dataset object, so the data can not be downloaded.

[Cadair]

Also assigning .base_path silently fails as you need to iterate over all the sub-datasets.

[Cadair]

It looks like the FileManager object we dynamically construct for the TiledDataset object needs a reference to the parent cube, but the only thing that's used for is accessing ds.meta["inventory"] so the best fix would be replacing the reference to the cube with an inventory dict (passed by reference).

In the short term adding a fm._ndcube = namedtuple(meta={"inventory": inventory}) here:

dkist/dkist/dataset/tiled_dataset.py

Lines 297 to 307 in 041ee09

	return FileManager(
	StripedExternalArray(
	fileuris=fileuris,
	target=1,
	dtype=dtype,
	shape=shape,
	loader=AstropyFITSLoader,
	basepath=basepath,
	chunksize=chunksize
	)
	)

before the return would probably work.