CELL2FLOAT

converts cell array into scalar float array (v4.0, jan 2010)

Description

CELL2FLOAT - converts cell array into scalar float array

M = CELL2FLOAT(C) returns a float array M with the same size as the cell array C. C can be any N-dimensional cell array. Cells of C that contain a scalar float (single or double) are put in corresponding locations of M. Cells that contain other datatypes (e.g., strings, arrays, empty, integers, etc.) yield NaNs in the corresponding locations of M.
 
M = CELL2FLOAT(C,F) uses the value F instead of NaN to fill these latter locations. F should be a scalar float (single or double).

M = CELL2FLOAT(C,F,'error') will cause the program to error if a cell does not contain a scalar float. If F is empty ([]), NaN will be used as the filler value.
 
[M,Q] = CELL2FLOAT(C, ..) returns a logical array Q with logical ones (true) where the values of C are floats, and logical zeros (false) elsewhere.

Examples:
       C = {single(1) 2 'x' [] ; 1:3 complex(1,2) uint8(1) Inf}
       M = cell2float(C)
       % -> [ 1.00 2.00 NaN NaN ;
       % NaN 1.00 + 2.00i NaN Inf ]
 
       cell2float({1 2 [] 3 [] 5},999)
       % -> [ 1 2 999 3 999 5 ]
 
       cell2float({1,2,'x'},[],'error')
       % ??? Error using ==> cell2float at 80
       % Not all cells contain a scalar float.
 
     See also cell2mat, cellfun, num2cell

version 4.0 jan 2010

Acknowledgements

The author wishes to acknowledge the following in the creation of this submission:
cell2num
This submission has inspired the following:
Pivot/unPivot

• MATLAB release: MATLAB 6.5 (R13)
• Other requirements: should work in most ML releases
• Other Files: cell2float.m, ; license.txt

[A K]

Works the magic in a short time. I needed to convert a cell vector which had numerous empty cells and a few sparse 1's into an equivalent vector with 1's and 0's. I had not managed to find a fast, vectorized approach until now. Thanks!

[Oleg Komarov]

I would prefer the default behaviour to throw an error if a mixed datatype cell array is supplied.
Its auto-NaN behavior can pass unobserved and can lead to late detection of the error if the input was meant to be a cell array of scalars n the first place.
This means that if i believe my Input is a cell array of scalars but it is not i will notice only by supplying the index output of cell2float and calling an any(IDX).

So i'd like a more transparent syntax as:
cell2float(In,NaN)
and
cell2float(In) --> error if In mixed datatypes.

[Oleg Komarov]

I added on line 49 (shifting the rest down) these lines of code:
if nnz(~cellfun('isclass',C,class(C{1}))) > 0
        error('cell2float:mxdClassInput',...
'Filler not supplied. All contents of C must be of the same data type.')
    end

I'll add your cell2float (modified with those lines) + license to my next submission (instead of using cell2mat) if you agree.

[Oleg Komarov]

 

[Oleg Komarov]

>> cell2float({23,35},'error')
ans =
#rorerror

line 64 isn't consistent with char preallocation.

I substituted line 57 with:
if nnz(~cellfun('isclass',C,class(C{1}))) > 0
        error('cell2float:NonScalerCells','Not all cells contain a scalar float.');
    else Filler = NaN;
    end

Then I deleted line 70-72.
Oleg