share function works for a single user like:

share(g, block, 'anita', 1, 1);

but

1. there is no way to make an object(s) public (safety_level = 1) or group-accessible (safety_level = 2). This feature is supported by the API with the same principle.
2. would be cool if multiple users can be specified (supported by the API) - minor for the moment

Due to lacking log4j.properties file, HTTPComponents throws a warning on start-up complaining about lack of appenders.

Would be really useful if there is a way to see the currently logged-in user information, as well as some connection information somehow.

Now the get_list function returns only object IDs. would be cool if this could be extended by other object attributes, parent (and references to children?). Otherwise now it's almost impossible to query any object without remembering its exact ID.

We found that it's hard to explore the structure of a, say, block (loaded in cascade mode) by NEO_IDs. The query:

b.segment_set

ans =

segment_34: [1x1 struct]
segment_35: [1x1 struct]
segment_36: [1x1 struct]

does give only list of NEO_IDs, which is sort of meaningless. Maybe names of structs inside "...set" containers should be assigned as ID or smth like that.

1. 'safety_level' field: a bit strange seems that it is assigned for metadata objects (section, property, value) but is absent for other objects (NEO). If there is no specific reason for that I'd make it consistent.
2. Value['attributes'] - 'current_state' is no longer exists.

Would be nice to have bulk update implemented somehow, e.g. not to loop over a list of objects on the Client side but just make an update in one request. at the API level it could be done by sending a POST to for example

http://141.84.42.103:8010/electrophysiology/analogsignal/?id__in=[30,40,50]&bulk_update=1

providing object ids as a list + having bulk_update=1 parameter to confirm update.

To get to the signals from a loaded block one has to type:

b.segment_set.segment_12345.analogsignal_set.analogsignalarray_12345

or to get to some of array parameter like data even more:

b.segment_set.segment_12345.analogsignalarray_set.analogsignalarray_12345.signal.data

Maybe one should think of some aliases for NEO objects, like

• "as", "irsa" etc for object NEO_IDs, so to have "as_12345"
• have "events" instead of "event_set" (just 3 letters less, but still)

Any other ideas?

These aliases could be made configurable, say, in requirements.json.

Implement "shortcuts" for typical operations, to make actions easier. For example, make a function to create a property with several values in one line. Or maybe the same for object creation - make it possible in one line (?).

POST to create / update an object is followed by a GET request to get an object. However the response to the initial POST already contains the created / updated object. So probably additional GET is not needed and must be removed. (is there any other reason for it?)

get_updated fails for data-related objects.

irr_events.name = 'stimulus event array';
irr_events = get_updated(g, irr_events);
??? Error using ==> update at 22
[GNODE] Cannot perform update without id. Please specify

Error in ==> get_updated at 14
updated_id = update(session, obj){1};

Besides that, the get_updated function is not documented.

when I create a signal the t_start and sampling_rate are not assigned to the object that I get back:

I = make_dummy(g, 'analogsignal');
I.name = 'I';
I.sampling_rate = struct('units', 'Hz', 'data', data.SamplingFrequencyHz);
I.t_start = struct('units', 'ms', 'data', 0);
I.signal = struct('units', 'mV', 'data', data.I);
I.sampling_rate

ans =

units: 'Hz'
 data: 20000

I = get_created(g, I);
I.sampling_rate

ans =

units: 'Hz'
  url: ''

When there is no network, the application doesn't print any warnings, and silently downloads empty structures, which is a bit confusing. Would be cool to change that somehow.

sr_sec = make_dummy(g, 'section');
sr_sec.name = 'spontaneous recording';
sr_sec = get_created(g, sr_sec);
??? Reference to non-existent element of a cell array.

Error in ==> get_created at 15
new_id = create(session, obj){1};

2 main changes:

• data in HDF5 files, now in all data-related objects instead of array inside JSON there is a link to the datafile with an appropriate dataslice
• objects are accessible back in time: a 'at_time' parameter is supported for GET requests. for example, http://141.84.42.103:8010/electrophysiology/block/8?at_time=2012-10-10 12:59:32

• h5read should be substituted with hdf5read
• Nested namespaces are not yet supported, should be flattened

remove 'definition' and 'description' for property / section from required fields.