Cell conversion to double

Greetings, Let's say a is a 11x1 cell like this a = '0.000000' '1.000000' '2.000000' '3.000000' '4.000000' '5.000000' '6.000000' '7.000000' '8.000000' '9.000000' '10.000000' and I want to convert it in double. If I try b=cell2mat(a), I got the following error : ??? Error using ==> cat CAT arguments dimensions are not consistent. Error in ==> cell2mat at 85 m{n} = cat(1,c{:,n}); However, I know I can bypass it with a loop with 2 conversion as: for i = 1:length(a) b(i) = str2num(cell2mat(a(i))); end Thus, I wonder if there is a simpler way to do this with an easy one-step function. Regards, Steven

[Barry Swindler]

not sure if you found an answer yet, but a simple thing like this works.

    m=zeros(size(a,1),size(a,2));
    m=str2double(a);

still not just one step, but there's no need for a loop.

hope this helps!

[ayesha abbassi]

B = cell2mat(A). now check its type by writing whos B in command window

[Chrysi K.]

@ayesha abbassi Thank you so much!!! You helped me!!! I had a similar problem!

Answer 1

[Jan]

To convert a cell array of character vectors to numbers, you can use the |str2double| function. This function is the simplest method. C = {'0.000000'; '10.000000'; '100000.000000'}; M = str2double(C) The |cell2mat| function converts a cell array of character vectors to a character array, and only if all the character vectors have the same length. |cell2mat| also preserves the data type of the contents of the cells, so it does not convert characters to numbers. If you need your code to be fast, then use the following code instead. This code is faster than |str2double|: C = {'0.000000'; '1.000000'; '2.000000'; ... '3.000000'; '4.000000'; '5.000000'; '6.000000' '7.000000'; '8.000000'; '9.000000'; '10.000000'}; S = sprintf('%s*', C{:}); N = sscanf(S, '%f*'); Unfortunately |sprintf| seems to forget a proper pre-allocation. This C-Mex is 4 times faster: : S = CStr2String(C, '*'); N = sscanf(S, '%f*'); Timings in 2011b, Core2Duo: n = 100; C = cell(1, n); for iC = 1:n; C{i} = sprintf('%f', i); end tic; for i=1:1000; N = cellfun(@(x)str2double(x), C); end; toc >> 3.61 sec tic; for i=1:1000; N = cellfun(@(x) sscanf(x, '%f'), C); end; toc >> 3.01 sec tic; for i=1:1000; N = str2double(C); end; toc >> 2.79 sec tic; for i=1:1000; N = cellfun(@str2double, C); end; toc >> 2.49 sec tic; for i=1:1000; N = zeros(1,100); for j=1:100; N(j) = sscanf(C{j}, '%f'); end; end; toc >> 1.40 sec tic; for i=1:1000; N = sscanf(sprintf('%s*', C{:}), '%f*'); end; toc >> 0.14 sec tic; for i=1:1000; N = sscanf(CStr2String(C, '*'), '%f*'); end; toc >> 0.071 sec To my surprise a full implementation in C is *slower* than |sscanf(sprintf())|, see . Matlab's sscanf seems to be much better than the MSVC implementation.

[Daniel Shub]

Nice comparison. One of the benefits of my job is access to way more computational power than I need. In general, I rarely worry about efficiency, however a speed up of a factor of 50 is nothing to sneeze at.

[Daniel Shub]

A faster, although not necessarily viable way, is to have the cell array contain numbers and not strings. On my computer

for iC = 1:n; C{iC} = iC; end

tic; for i=1:1000; N = [C{:}]; end; toc

appears fastest, although not by as much as I would have guessed.

[Fangjun Jiang]

+1, Nice work, Jan! str2double() has source code available. It uses sscanf().

[Jan]

@Daniel: Even [C{:}] seems to have problems with the pre-allocation. The C-Mex http://www.mathworks.com/matlabcentral/fileexchange/28916-cell2vec determines the needed output size at first. For the {1 x100} cell containing scalars it is 5 times faster than [C{:}] in 2001b.

[Jan]

Inserted: N(j)=sscanf(C{j}, '%f').

[hello_world]

@Jan great example, but it does not help if the cell contains strings. I have raised a question with that problem at: https://www.mathworks.com/matlabcentral/answers/408675-how-to-convert-a-csv-file-of-cell-array-type-to-double

Can you please help with that?

[Nishaban P.K.]

sir, I use str2double function ,it return some binary values like '00000111' into 111. why?

[Giuseppe Degan Di Dieco]

Dear Jan,

thanks for your tip. Still useful these days.

Best.

[Hiril Patel]

Hello Jan

i have cell of 73x1 of splitted data which i would like to convert to decimal but when I use M = str2double(adc_data). it returns to NaN

here is the string data i splitted

[Christopher Saltonstall]

@Jan I have a similar problem but a slightly different data set. My data set is read from a strangly formated .csv file an yields a table where each cell of the table has an array. When I read one cell of the table, I get something like this but much longer.

C = {'[0 1 2 3 4]'};

How can I convert this to a double array?

This is the only way that I have figured out how to do this but eval does not work with my data size.

A = cell2mat(C)
eval(['data = ' A])

[Walter Roberson]

@Christopher Saltonstall

A = cellfun(@(S)str2num(S,Evaluation="restricted"), TableNameGoesHere.FieldNameGoesHere, 'uniform', 0)

possibly followed by a cell2mat(A) provided that all of the entries are the same size.

Answer 2

[Daniel Shub]

It appears your cell array contains strings and not numbers (doubles).

b = cellfun(@(x)str2double(x), a);

[Fangjun Jiang]

Believe it or not, you can use str2double directly.

a={'1','2'};

str2double(a)

[Daniel Shub]

Wow, another reason to like str2double over str2num.

[Jan]

cellfun(@str2double, C) is faster than str2double(C). Surprising!

[Jan]

"cellfun(@str2double, C)" is faster than "str2double(C)". Surprising! But the indirection "cellfun(@(x)str2double(x), C)" wastes time.

[Fangjun Jiang]

+1, For none-time-critical task, I still vote for using str2double().

[hello_world]

@Daniel Shub: It converts all cell entries (which are string) into NaN values. I have raised a question at: https://www.mathworks.com/matlabcentral/answers/408675-how-to-convert-a-csv-file-of-cell-array-type-to-double

Can you please look into that?