How to sort file when reading them with dir ?

I have list of text files , they are ordered like

• doc_1.txt
• doc_2.txt
• ..
• doc_20.txt

when I use

allFiles = dir('*.txt');

it is getting files in this order

• doc_1.txt
• doc_10.txt
• doc_11.txtHow can I get these files in correct order?

Answer 1

[Stephen23]

You could download my FEX submission natsortfiles:

http://www.mathworks.com/matlabcentral/fileexchange/47434-natural-order-filename-sort

and use it to directly sort the output from DIR:

allFiles = natsortfiles(dir('*.txt'));

The function natsortfiles does not perform a naive natural-order sort, but also sorts the filenames and file extensions separately so that the file extension period character does not influence the sort output:

>> A = {'doc_10.txt'; 'doc_20.txt'; 'doc_2.txt'; 'doc_1.txt'; 'doc_11.txt'};
>> sort(A) % gives the wrong order!
ans = 
    'doc_1.txt'
    'doc_10.txt'
    'doc_11.txt'
    'doc_2.txt'
    'doc_20.txt'
>> natsortfiles(A) % the correct number order :)
ans = 
    'doc_1.txt'
    'doc_2.txt'
    'doc_10.txt'
    'doc_11.txt'
    'doc_20.txt'

[Marsge]

Thanks, you save my day!

[Farid]

This saved my life

Thanks

[Chandra Shekhar Lohani]

Thanks @Stephen23 , this function file worked for me with .csv data .

Answer 2

[per isakson]

See

• sort_nat: Natural Order Sort, by Douglas Schwarz
• Natural-Order Filename Sort, by Stephen Cobeldick

and there are more in the File Exchange

[Tural]

I can't use external functions ,is there any other way of doing that?

[per isakson]

"[...] external functions" &nbsp why is that? &nbsp AFAIK: The alternative is to make your own function (or script). Both these FEX-contributions shows how to do that. They are well documented.

Answer 3

[Andrei Bobrov]

n = dir('doc_*.txt');
n1 = {n.name};
z = regexp(n1,'(?<=doc_)\d*(?=\.txt)','match');
z1 = str2double(cat(1,z{:}));
[~,ii] = sort(z1);
n01 = n1(ii);
z2 = z1(ii);
lga = floor(log10(z2)) + 1;
ml = max(lga);
ptr = ['doc_%0',num2str(ml),'d.txt'];
p = z1(ii);
k = lga < ml;
p1 = p(k);
n02 = n01(k);
if any(k)
    for jj = 1:numel(p1)
        movefile(n02{jj},sprintf(ptr,p1(jj)));
    end
end

Answer 4

[Thorsten]

Convert each filename to the corresponding number (strip leading doc_ and trailing .txt, then use double); next, sort the numbers and get the second output of sort, ind; finally use ind to re-order your filenames.

[Stephen23]

As long as there is one (and only one) number per filename, and it is in exactly the same position in each filename...

Answer 5

[Steve Chavez]

Maybe you could use the names of the files

• Choose the order from the name to obtain indexes and then use the indexes to sort the list.

As an example: The names of the variables were chosen assuming that the number in the txts represent days.

Use dir to obtain a list into a struct

list=dir('*.txt')

Convert the struct to cell array

listcell=struct2cell(list)

[a b]=size(listcell)

Create a cell array

    dia_string=cell(b,1)

A simple loop to obtain the order (order of days) from the names. If nombre is 'doc_20.txt' nombre(5:6) is equal to 20

    for k=1:b   
        nombre=listcell{1,k};
        dia_string{k}=nombre(5:6); 
    end

Use sort and save the index

[diaorden ind]=sort(dia_string);

Use the index in the cell array

list_2=list(ind) % list_2 is the list ordered as expected

[Stephen23]

@Steve Chavez: the original question lists some example filenames:

doc_1.txt
doc_10.txt
doc_11.txt

Notice that the number of digit characters is not constant: in some cases one, some two, and perhaps more. Currently your answer takes a fixed number of characters (two) from a fixed location in the string, which means that it will give incorrect sort orders for any numbers with not-two digits, and that if the number is located elsewhere in the string the output will be rubbish. So it is not a very general solution.

However note also that if the files were named with constant-width numbers, as your answer uses, then there is no point in this answer (or any of the answers) because then a simple sort gives the correct output. Compare your answer:

>> C = {'doc_99.m','doc_10.m','doc_50.m','doc_01.m'};
>> [a,b]=size(C);
>> dia_string=cell(b,1);
>> for k=1:b, nombre=C{1,k}; dia_string{k}=nombre(5:6); end
>> [diaorden,ind]=sort(dia_string);
>> C(ind)
ans = 
    'doc_01.m'    'doc_10.m'    'doc_50.m'    'doc_99.m'

which gives exactly the same answer as simply calling sort:

>> sort(C)
ans = 
    'doc_01.m'    'doc_10.m'    'doc_50.m'    'doc_99.m'