MATLAB: Does ‘varfun’ exclude rows with an empty cell

MATLAB

When I use 'varfun' with a grouping variable that contains empty char row vectors (''), the resulting table excludes these rows. See the following example:
 
tblInput = table({'';'';''},[10;20;30],[40;50;60],[7, 1;8, 2;9,3]);
tblInput.Properties.VariableNames = {'X','Y','Z','Quantity'};
strGroupingVariables = {'X','Y','Z'}; 
tblEmpty = varfun(@nansum,tblInput,'GroupingVariables',strGroupingVariables, 'InputVariables', {'Quantity'})
result:
 tblEmpty =
   0×5 empty table
Is this expected behavior? How do I get the resulting table to include the columns of 'X', just as an empty string?

Best Answer

The behavior you are experiencing is expected behavior. When creating a column consisting of a cell array with empty strings, these are actually treated as empty char row vectors rather than empty strings. Since these are treated as empty char row vectors, they are treated as a missing value. Because of this, there is no "group" to assign them to when passing the variable 'X' to the 'GroupingVariables' parameter, so the entire row is excluded. 
A workaround would be to convert the variable 'X' from a cell array to a string array, as follows:
 
tblInput = table({'';'';''},[10;20;30],[40;50;60],[7, 1;8, 2;9,3]);
tblInput.Properties.VariableNames = {'X','Y','Z','Quantity'};
tblInput.X = string(tblInput.X);
strGroupingVariables = {'X','Y','Z'}; 
tblEmpty = varfun(@nansum,tblInput,'GroupingVariables',strGroupingVariables, 'InputVariables', {'Quantity'})
result:
 tblEmpty =
   3×5 table
     X     Y     Z     GroupCount    nansum_Quantity
     __    __    __    __________    _______________
     ""    10    40    1              8             
     ""    20    50    1             10             
     ""    30    60    1             12
Related Question