Trouble opening ndtiff dataset

[talonchandler]

I'm running into issues opening some, but not all, recently acquired ndtiff datasets.

Reproduce with:

python
>>> from iohub import read_micromanager
>>> reader = read_micromanager('/hpc/projects/compmicro/rawdata/mantis/2023_03_29_argolight/rings-LF-LS_1/rings-LF-LS_labelfree_1/')
Dataset opened
>>> reader.get_array(0)

fails with

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/talon.chandler/iohub/iohub/ndtiff.py", line 182, in get_array
    return np.asarray(self.get_zarr(position))
  File "/home/talon.chandler/iohub/iohub/ndtiff.py", line 166, in get_zarr
    return da.reshape(shape)
  File "/home/talon.chandler/.conda/envs/recorder-test/lib/python3.9/site-packages/dask/array/core.py", line 2185, in reshape
    return reshape(self, shape, merge_chunks=merge_chunks, limit=limit)
  File "/home/talon.chandler/.conda/envs/recorder-test/lib/python3.9/site-packages/dask/array/reshape.py", line 248, in reshape
    inchunks, outchunks = reshape_rechunk(x.shape, shape, x.chunks)
  File "/home/talon.chandler/.conda/envs/recorder-test/lib/python3.9/site-packages/dask/array/reshape.py", line 62, in reshape_rechunk
    raise NotImplementedError(_not_implemented_message)
NotImplementedError:
Dask's reshape only supports operations that merge or split existing dimensions
evenly. For example:

>>> x = da.ones((6, 5, 4), chunks=(3, 2, 2))
>>> x.reshape((3, 2, 5, 4))  # supported, splits 6 into 3 & 2
>>> x.reshape((30, 4))       # supported, merges 6 & 5 into 30
>>> x.reshape((4, 5, 6))     # unsupported, existing dimensions split unevenly

To work around this you may call reshape in multiple passes, or (if your data
is small enough) call ``compute`` first and handle reshaping in ``numpy``
directly.

[ziw-liu]

Are the ndtiff (or pycromanager) versions in the acquisition environment and the analysis environment the same?

[ieivanov]

This is fixed in ndtiff >= 1.12.1, see micro-manager/NDStorage#107

[talonchandler]

Tested and it's working. Thanks @ziw-liu and @ieivanov!