MATLAB table -> NWB DynamicTable converter

[bendichter]

We have a function util.table2nwb that takes in a MATLAB table and outputs an matnwb DynamicTable that can be directly added to an NWBFile object. This is a start, but is not fully functional enough to be used for all common table needs, like writing the units table. Here is a list of features that are needed:

• [x] take as input a MATLAB table object and output an NWB table
• [ ] create an indexed column for columns that have varying length arrays (util.create_indexed_column might be useful here)
• [ ] handle DynamicTableRegions (and indexed DynamicTableRegions)
• [ ] wrapper functions for default tables (electrodes, units, trials, invalid_times, etc.) that handle default columns correctly
• [ ] tests
  • [ ] test for round trip in MATLAB
  • [ ] test for read in pynwb
• [ ] documentation
  • [ ] function comments
  • [ ] incorporating into tutorials

How to do each of these is demonstrated here:

https://github.com/NeurodataWithoutBorders/matnwb/blob/fa4f8bdbd6f22cadca6f2d24901309d623a10947/tutorials/ecephys.m#L143-L225

[lawrence-mbf]

Do we eventually want this to be integrated into read/write?

[bendichter]

I was imagining a one-line convert command before write, but I think integrating it in directly could be nice too. What would that look like?

[lawrence-mbf]

There used to be a way to manually extend matNWB classes. I don't think that's around anymore but you would essentially make a subclass to DynamicTable as a container for the converted table, and have it run util.table2nwb() right before exporting normally. That might be farther down the line though.

[bendichter]

OK, sounds good. I like the idea of having the ability to extend matNWB classes. Having to_table() as a method, for instance, would be really nice.

[lawrence-mbf]

One of the inputs of the DynamicTable GetRow/AddRow #267 supports tables as well. In fact, adding a full table to an DynamicTable will convert the table to its equivalent DynamicTable type. Indexing is also supported here.