It's not clear from this that these configurations made it back into the larger object. It's also not clear how you would use this configuration to modify an in-memory NWBFile object. It would also be great to write this file and then show that the written file followed the chunking and compression parameters we chose. This could be done using HDFView, or some other app or API

It's not clear from this that these configurations made it back into the larger object.

This was obscured by the odd injection of all the notes; with those moved, I put the two printouts side by side and added some spacing

It would also be great to write this file and then show that the written file followed the chunking and compression parameters we chose.

The point the demo here makes is just that they get saved in the backend_configuration object itself - I can add something to prove that it actually writes the file in that manner (for chunks and compression anyway, can't prove buffering that way)

Though I will probably just demonstrate it with h5py since mentioning/showcasing HDFView could lead to annoyance because if a user wanted to try it they'd need to sign up for HDF5 - also Neurosift doesn't display this information, though I'm sure you could just request that - could also use PyNWB HTML repr after hdmf-dev/hdmf#1100 is merged, but this section is also currently for python script / ipython usage, not specific to a jupyter notebook

Added a section to the FAQ for how to check that customizations (or the default for that matter) actually configure the item correctly in the resulting file

OK great, could you please add a section on writing the data and confirming (somehow) that the data was written with the proper settings?

I did, it's the last section of the FAQ. Uses the file written using the preceeding section

OK that section of the FAQ looks great, but I still feel like it would be nice to have another snippet here showing how to apply the configs to an in-memory NWBFile object and write to disk

I don't understand what you mean here - what are you imagining and how is it different from the preceding section?

Maybe not that important and this is what you meant but you can pass chunks=None to zarr and get all the data in one file:

import zarr

# Create a Zarr group
zarr_folder = zarr.open_group('test_zarr.zarr', mode='w')

# Create an array and add it to the Zarr group
array_data = np.random.rand(20, 20, 20)
zarr_array = zarr_folder.create_dataset('array', data=array_data, chunks=None)
zarr_array.info

I did not check myself (trusted @bendichter on that one)

Our config literally just passes the values on to downstream calls so I assume this would work here, in which case we can remove this line as a minor point (even though chunks=None is the option, and even though the array says 1/1 chunks, that's all 'equivalent' to contiguous layout in a sense)