anonymous function w/ accumarray

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]', 'VariableNames',{'year','id'})

Hi

I got error message below. Could you help shed light why I can't use accumarray with concatinating text output?

year id

____ __

1 1

1 5

1 9

2 10

2 1

cat = @(x) {sprintf([repmat('%s,',1,length(x)-1) '%s'], string(x))};

However got error below when trying to call

[ uniqYr,~,JGrp] = unique( tdata(:,'year'));
tt= accumarray( JGrp, tdata.id, [], @cat);

Error using cat

Dimension argument must be a real, positive, integer scalar.

Answer 1

[Star Strider]

Naming your function ‘cat’ overshadows the existing MATLAB function by that name, that you inadvertently called when you used the ‘@’ operator in the accumarray call. First, rename the anonymous function, and do not use the ‘@’ in the accumarray call when calling it, since it is already a function handle.

This slight variation on the original code works:

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]', 'VariableNames',{'year','id'})

tdata = 5×2 table
    year    id
    ____    __

     1       1
     1       5
     1       9
     2      10
     2       1

catfcn = @(x) {sprintf([repmat('%s,',1,length(x)-1) '%s'], string(x))};
[ uniqYr,~,JGrp] = unique( tdata(:,'year'));
tt= accumarray( JGrp, tdata.id, [], catfcn)

tt = 2×1 cell array
    {'1,5,9'}
    {'10,1' }

.

[Pete sherer]

Thanks very much

[Star Strider]

As always, my pleasure!

[Pete sherer]

How can i do this operation with the string type variable? say another variable is "name" with values

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]',["a" "b" "c" "d" "e"]', 'VariableNames',{'year','id','name'})

then the accumarray should return

accumarray( JGrp, tdata.name, [], catfcn)

{'a,b,c}

{d,e}

[Star Strider]

The accumarray function only works with numeric values and character vectors, so your example code requires one small change —

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]',['a' 'b' 'c' 'd' 'e']', 'VariableNames',{'year','id','name'})

tdata = 5×3 table
    year    id    name
    ____    __    ____

     1       1     a  
     1       5     b  
     1       9     c  
     2      10     d  
     2       1     e  

catfcn = @(x) {sprintf([repmat('%s,',1,length(x)-1) '%s'], string(x))};

[ uniqYr,~,JGrp] = unique( tdata(:,'year'));

tt = accumarray( JGrp, tdata.name, [], catfcn)

tt = 2×1 cell array
    {'a,b,c'}
    {'d,e'  }

Then, it works!

.

[Pete sherer]

I tried using the cellstr() to apply to the string-type variable before using the step above but it didn't work.

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]',["a" "b" "c" "d" "e"]', 'VariableNames',{'year','id','name'});

tdata.name= cellstr( tdata.name);

catfcn = @(x) {sprintf([repmat('%s,',1,length(x)-1) '%s'], string(x))};

[ uniqYr,~,JGrp] = unique( tdata(:,'year'));

tt = accumarray( JGrp, tdata.name, [], catfcn)

Error using accumarray

Second input VAL must be a full numeric, logical, or char vector or scalar.

[Star Strider]

It doesn’t work because it remains a cell array. Use the char function instead —

tdata = table([ 1 1 1 2 2]', [1 5 9 10 1]',["a" "b" "c" "d" "e"]', 'VariableNames',{'year','id','name'});
% tdata.name= cellstr( tdata.name)
tdata.name = char(tdata.name)

tdata = 5×3 table
    year    id    name
    ____    __    ____

     1       1     a  
     1       5     b  
     1       9     c  
     2      10     d  
     2       1     e  

catfcn = @(x) {sprintf([repmat('%s,',1,length(x)-1) '%s'], string(x))};
[ uniqYr,~,JGrp] = unique( tdata(:,'year'));
tt = accumarray( JGrp, tdata.name, [], catfcn)

tt = 2×1 cell array
    {'a,b,c'}
    {'d,e'  }

.