Bug in current Docker tutorials - paths no longer match code expectations

[serbinsh]

Working through setting up ISOFIT on an SMCE resource for training purposes and I noticed that within the existing Dockerized version of the code, there are some new inconsistencies with existing Jupyter notebooks and new locations of the ISOFIT data/examples. For example, running

data_prep.ipynb

And I think this because the data is now located in:

/examples/isotuts/NEON/neon_subset/

instead of the expected

examples/isotuts/NEON/data/

A test to try and fix this, I modified the path definitions as follows

# Path to the input NEON data
indata = os.path.join(home, 'neon_subset') 
subset_dir = os.path.join(home, 'data', 'subsets')

From the original

# Path to the input NEON data
indata = os.path.join(home, 'data') 
subset_dir = os.path.join(indata, 'subsets')

Fixes the above issue, though then I ran into a new bug

trying to sort it out because I can confirm the steps before work

[serbinsh]

[serbinsh]

It has to do with array dimensions of the file and maybe a mismatch with the hdr files? Perhaps they were subset already and the hdr wasnt updated?

[serbinsh]

If i use

plt.imshow(rdn[0:63,0:67,:] / np.max(rdn[0:63,0:67,:],axis=(0,1)))

It will create the plot but there are still issues with subsetting for the output figures

[pgbrodrick]

@serbinsh , thanks for the flag - this tutorial is not something I've run with docker before. It's primary intent was to provide some information as to how the subsets file was generated.

The data for this notebook, as noted in the documentation up to, is at a download link (https://avng.jpl.nasa.gov/pub/PBrodrick/isofit/tutorials/neon_full_data.zip), not in an isofit data directory. The user would need to download the data and put it into the directories noted in the first cell to make this work. It's not really intended as an automated process...I can add some information to the top to note more strongly that users should just skip to the next notebook.

As to the broadcasting issues, I think you just need to include the broadcast dimensions, e.g.:

plt.imshow(rdn[0:63,0:67,:] / np.max(rdn[0:63,0:67,:],axis=(0,1)[np.newaxis,np.newaxis,:]))

If I'm reading the error correctly.

[serbinsh]

OK - I think I need a power lesson on which examples were intended to be run via jupyter and which were meant for jupyter in a dockerized version. I perhaps mis-understood this from the start!