MATLAB: Conversion problem of data types of an array

conversiondata type

Hi all, I read with textscan from a text file to get an array. I received this error report at first

"All contents of the input cell array must be of the same data type."

for my array "data_A".

data_A:
522x1 double       522x1 double       522x1 cell       522x1 double

I tried to do this:

data_A(:,3) = str2double(data_A(:,3));

and I received this report:

"Conversion to cell from double is not possible."

Hence, I tried to use cell2mat:

data_A(:,3) = cell2mat(data_A(:,3));

for which I got the same error again:

"All contents of the input cell array must be of the same data type."

It would be so kind, if anyone could look into the matter. Any help and tips are very much appreciated. =)

Best regards,

Boon

Best Answer

I'm with dpb, fix the code so that it does not make a mess in the first place rather than trying to fix the mess.

It sounds like you're importing some data with something like textscan which needs you to explicitly specify the format of the text file. Since R2013b, matlab has included readtable which is usually capable of figuring out the format (and headers) on its own. Perhaps, moving to more modern constructs would solve the problem.

Most time all that is needed is:

t = readtable(yourfile);

Nothing more, and if readtable gets it wrong, you can guide it with detectImportOptions and specify the actual format for these columns where it gets it wrong.

Related Solutions

MATLAB: Finding some specific numbers in a numeric cell array

Looks like you've nested the cell array; either create an array of cells with the double arrays each per cell or dereference inside your anonymous function--example:

>> dat(1)={randi([7 13],10,1)};   % make sample cell array 
>> dat(2,1)={randi([7 13],8,1)};  % two cells w/ a double array in each
>> whos dat
Name      Size            Bytes  Class    Attributes
dat       2x1               264  cell               
>> dat
dat = 
  [10x1 double]
  [ 8x1 double]
>> cellfun(@(x) find(x==7), dat, 'UniformOutput', false)
ans = 
  [3x1 double]
  [         5]
>> ans{:}
ans =
   7
   9
  10
ans =
   5
>> data={dat}  % now nest that in a single cell...
data = 
  {2x1 cell}
>> cellfun(@(x) find(x==7), data, 'UniformOutput', false)
Undefined function 'eq' for input arguments of type 'cell'.
Error in @(x)find(x==7)
>>

Look familiar??? :)

>> cellfun(@(x) find(x==7), data{:}, 'UniformOutput', false)
ans = 
  [3x1 double]
  [         5]
>> whos data
Name      Size            Bytes  Class    Attributes
data      1x1               324  cell               
>>

NB: The curlies "{:}" to dereference the content of the cell in the latter...

MATLAB: Populating a cell using a loop

textscan returns a 1x55 cell array, where the format string specifies the 55. So your code:

A = textscan(fopen(d),'%f %f %f %f %f %f %f %f %f %f %f %f %f %f %s %s %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %s %s %f %f %f %f %f %f %f %f', 'Delimiter',',','Headerlines',1);

will result in a 1x55 cell array. Then a few lines later you write this:

b = A(n,:);

because A only has one row, as soon as n>1 this will be an error, because C only has one row, and so requesting anything from its (non-existent) second (or higher) row will be an error.

"I want b to be a cell with 55 columns but with n rows. one row for each data file that I run"

I really really really recommend that you don't do that: that would require putting scalar numeric data into the cells of a cell array, which just makes it much harder to work with numeric data. You should really keep the numeric data in numeric arrays, or use a table. Because you have a few columns which are character, this complicates the importing a little bit, but there are reasonable solutions which you should look at:

If you do not need those character columns then get textscan to ignore them with %*s in the format string. Then could trivially get textscan to collect all of the numeric data into one numeric array, using the CollectOutput option. Very simple, but you would lose some data.
Use the CollectOutput option to collect the data into arrays of matching types: this would give you one 1x5 cell array C, containing an Nx14 numeric array, an Nx2 char cell array, an Nx29 numeric array, an Nx2 char cell array, and an Nx8 numeric array (or whatever sizes that format string gives you). I recommend this option.
Use a table. These are a very convenient way for handling mixed data (e.g, numeric, char, categorical) and analyzing it. It has many powerful methods and operators for processing data by groups, and for statistical analyses.

Or your proposal:

If you really want to get all of your data into one cell array (which will make any numeric processing slow, inefficient and complex), then you will need to post-process the data after it has been imported, something like this: detect numeric columns, convert numeric columns to cell array containing numeric scalars (e.g. num2cell), then concatenate all into one cell array. I strongly advise you to avoid doing this.

Best Answer

Related Solutions

MATLAB: Finding some specific numbers in a numeric cell array

MATLAB: Populating a cell using a loop

Related Question