in validation, add check for if a varlen datasets is empty

[bendichter]

Motivation

fix #788

@oruebel here's my attempt

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (216aa34) 88.46% compared to head (337162a) 88.50%.

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #789      +/-   ##
==========================================
+ Coverage   88.46%   88.50%   +0.04%     
==========================================
  Files          45       45              
  Lines        9579     9582       +3     
  Branches     2723     2724       +1     
==========================================
+ Hits         8474     8481       +7     
+ Misses        781      778       -3     
+ Partials      324      323       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[oruebel]

I think this fix looks appropriate. Now we need a test.

[oruebel]

Ok, I have a test that will trigger the issue. I'll just add it to this PR.

[oruebel]

@rly can you please review this PR. I made a bunch of changes on the PR so I should not be the approver.

The main change is the following line, to make sure the validator works if we encounter an empty dataset of strings (e.g, an empty keywords dataset in NWB)

hdmf/src/hdmf/validate/validator.py

Lines 144 to 146 in d2f916e

	# Empty string array
	if data.dtype.metadata["vlen"] == str:
	return "utf"

The other changes are mainly to add inline comments to make the get_type function easier to parse and adding of unit tests.