Define ElectrodesTable and FrequencyBandsTable types

[rly]

Fix #336

[bendichter]

This won't cause validation or IO errors for files that already exist? If this can be done without disrupting users I think it's a good idea

[oruebel]

I think on the PyNWB the validation and read of existing files can be handled. @bendichter does this change anything for MatNWB?

[bendichter]

@oruebel I don't know, possibly. I can check

[bendichter]

I'm having trouble getting the matnwb to work. They are several schema changes behind at this point.

[bendichter]

This break matnwb/tutorials/ecephys.m because the electrodes table is created manually as a DynamicTable, which needs to change to an Electrodes object. I'll fix that and see if there are any other issues

[bendichter]

I would prefer to have the convention that tables are called Tables, i.e. this would be an ElectrodesTable object, similar to how we name descendants of TimeSeries with Series

[bendichter]

@rly status update: this is working on the matnwb side

[rly]

I would prefer to have the convention that tables are called Tables, i.e. this would be an ElectrodesTable object, similar to how we name descendants of TimeSeries with Series

I would prefer this convention as well. However, we already have the Units table and the TimeIntervals table (which can have fixed names "epochs", "trials", "invalid_times"). The icephys extension will also add a few more DynamicTable types: IntracellularRecordings, SweepSequences, Runs, and Conditions.

[bendichter]

I was talking about a convention for the neurodata_type name, not the object name, so it's not an issue for "epochs", "trials", "invalid_times". I agree we have types that break this convention, but I think we should use it going forward. At least when I create extensions I am going to try to use this naming convention for children of DynamicTable. At any rate, this schema change looks good to me besides that.

[rly]

I renamed the types and suggested a renaming of the icephys extension types: oruebel/ndx-icephys-meta#27

If we go forward with this convention, I also suggest renaming Units to UnitsTable and making Units an alias to UnitsTable to maintain backwards compatibility. Same for TimeIntervals to TimeIntervalsTable.

[bendichter]

@rly that sounds scary. I'm worried about backwards compatibility, esp for matnwb