Fixing regions props node attribute init

[JoOkuma]

No description provided.

[yfukai]

Hi, I'm sorry for my absence for a while @JoOkuma, I believe you're working on #191, but am I right?

[JoOkuma]

Hi @yfukai , yes.
I've found a few issues with the dtypes when they are array or other complex types, and there are null values.
It's not directly related to this, but arises from some of the regionprops attributes.

[yfukai]

Thanks @JoOkuma for this! I believe that arises from the dtype inference, and can be solved by #262 . Maybe you could give it a try.

[JoOkuma]

@yfukai, #262 fixed this. The tests are passing locally for me.
We can merge this after #262 is merged.

[yfukai]

Thanks! I feel sometimes we'd like to filter the nodes by the intensities of specific channels. Would it make sense to flatten the vector-like props? regionprops_table automatically does this, and I was wondering if we could use this instead.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 86.45833% with 26 lines in your changes missing coverage. Please review.
✅ Project coverage is 87.76%. Comparing base (e4bb183) to head (3a17bdb).

Files with missing lines	Patch %	Lines
src/tracksdata/utils/_dtypes.py	71.83%	11 Missing and 9 partials ⚠️
src/tracksdata/graph/_sql_graph.py	95.50%	3 Missing and 1 partial ⚠️
src/tracksdata/nodes/_regionprops.py	88.23%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #259      +/-   ##
==========================================
- Coverage   87.92%   87.76%   -0.16%     
==========================================
  Files          56       56              
  Lines        4471     4611     +140     
  Branches      789      809      +20     
==========================================
+ Hits         3931     4047     +116     
- Misses        344      361      +17     
- Partials      196      203       +7     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JoOkuma]

@yfukai

Would it make sense to flatten the vector-like props? regionprops_table automatically does this, and I was wondering if we could use this instead.

I've been using both patterns when analysing the attributes. It's often easier to get inertia_tensor rather than getting all inertia_tensor_*_*.

For node_attrs, we have unpack=True, but it doesn't address filtering.

We could add an option to RegionPropsNodes to unpack before adding to the attributes.
What do you think?

[JoOkuma]

@yfukai I opened an issue for this #269