[Bug]: Discrepancy between accepted dimensions

KyzarNexus commented 1 year ago

What happened?

There appears to be a discrepancy between the data dimensions accepted for '+types+core\OpticalSeries.m' & the schema definition in 'nwb-schema\2.6.0\core\nwb.image.yaml'. For the nwb.image.yaml OpticalSeries definition, the accepted dimensions for an RGB image are [frame, x, y, 3] (ln159-162). However, the valid RGB shape allowable for OpticalSeries.m is [3, Inf, Inf, Inf] (ln 60). That being said, the error resulting from this can be circumvented by permuting the data before adding it to an OpticalSeries, but is there a reason why the acceptable dimensions are reversed here?

Steps to Reproduce

nwb = NwbFile();

% Dims: (#Frames, Xpixel, Ypixel, numRGB) According to 2.6.0 OpticalSeries Schema
sampleData = uint8(randi([0 255],10,200,100,3));
% sampleData = permute(sampleData,[4, 3, 2, 1]); % Permutation for compatability with types+\core+\OpticalSeries.m ; Uncomment to resolve error. 
timestamps = 1:10;

stimulus = types.core.OpticalSeries(...
           'description', 'Images presented to the subject. Data is stored in the dimensions [nrFrames,nrPixelsX,nrPixelsY,rgbCount].', ...
           'data', sampleData, ...
           'timestamps', timestamps, ...
           'orientation', 'lower left', ...
           'format', 'raw', ...
           'distance', 0.7, ...
           'field_of_view', [0.3, 0.4, 0.7], ...
           'dimension', [200,100, 3], ...
           'data_unit', 'meters');
nwb.stimulus_presentation.set('StimulusPresentation', stimulus);

Error Message

Matrix dimensions must agree.

Error in types.util.checkDims>getIsSizeMatch (line 51)
tf = length(expectedSize) <= length(actualSize) && all(openSizeMask | expectedSize == actualSize);

Error in types.util.checkDims (line 17)
        isSizeMatch = getIsSizeMatch(expectedSize, valueSize);

Error in types.core.OpticalSeries/validate_data (line 61)
        types.util.checkDims(valsz, validshapes);

Error in types.core.TimeSeries/set.data (line 90)
        obj.data = obj.validate_data(val);

Error in types.core.TimeSeries (line 59)
        obj.data = p.Results.data;

Error in types.core.ImageSeries (line 20)
        obj = obj@types.core.TimeSeries(varargin{:});

Error in types.core.OpticalSeries (line 15)
        obj = obj@types.core.ImageSeries(varargin{:});

Error in opticalSeriesTest (line 8)
stimulus = types.core.OpticalSeries(...

Operating System

Windows

Matlab Version

R2019a

Code of Conduct

[X] I agree to follow this project's Code of Conduct
[X] Have you ensured this bug was not already reported?

lawrence-mbf commented 1 year ago

Hi @KyzarNexus ,

This just has to do with how MATLAB orders its data when writing using the HDF5 library. As a result, the dimension validation is intentionally flipped for better compatibility with HDF5, C, and Python and so we don't have to do rehapes on large data.

KyzarNexus commented 1 year ago

Hi @lawrence-mbf ,

Thanks for clarifying!

NeurodataWithoutBorders / matnwb