Convert a fixed width char array into a column vector

Hello! I have troubles converting numbers from a char array to double format. The char array always has the dimensions 1x48. Every 8 characters represent an integer. Sometimes there are spaces between numbers and sscanf fails to give the results I need. For example:

aa = ' 1703434 42      1012275140184521401845314018473';
sscanf(aa,'%8d')
ans = 
    1703434
         42
   10122751
   40184521
   40184531
    4018473

should return in fact

ans = 
    1703434
         42
   10122751
   14018452
   14018453
   14018473

The reason why I don't use str2double for each 8 characters in 'aa' is that it gets very slow for many conversions. The original file I read also contains other words, and this array 'aa' is only a part of each line I read. So far sscanf was the fastest method and it worked well until this line was encountered.

Could anyone please explain me why sscanf behaves like this? Is there any quick alternative?

Answer 1

[David Hill]

str2num(reshape(aa,8,[])');

[Adri]

I still used sscanf using the reshape idea of yours. It's a little bit faster:

aa = ' 1703434         1012275140184521401845314018473';
tic
for ii = 1:500000
colvec = sscanf([reshape(aa,8,[])', repmat(' ',length(aa)/8,1)]','%8f ');
end
toc
tic
for ii = 1:500000
colvec2 = str2num(reshape(aa,8,[])');
end
toc
isequal(colvec,colvec2)
Elapsed time is 5.666218 seconds.
Elapsed time is 9.619199 seconds.

ans =

  logical

   1

Thank you for your suggestion!

Answer 2

[Hamed Moasses]

Hi !

This Answer may help you.

bests

Answer 3

[Stephen23]

This will be quite efficient:

>> aa = ' 1703434 42      1012275140184521401845314018473';
>> vec = sscanf(sprintf('%c%c%c%c%c%c%c%c,',aa),'%f%*[ ,]')
vec =
    1703434
         42
    1012275
   14018452
   14018453
   14018473

[Stephen23]

Timing for 1e4 iterations:

Elapsed time is 2.80135 seconds. % reshape, repmat, sscanf
Elapsed time is 7.16556 seconds. % reshape, str2num
Elapsed time is 0.92013 seconds. % this answer

[Adri]

I tried this alternative both in my main code and in this example and it seems that on my machine it's slower (weird or not?):

aa = ' 1703434         1012275140184521401845314018473';
tic
for ii = 1:500000
colvec = sscanf([reshape(aa,8,[])', repmat(' ',length(aa)/8,1)]','%8f ');
end
toc
tic
for ii = 1:500000
colvec2 = str2num(reshape(aa,8,[])');
end
toc
tic
for ii = 1:500000
colvec3 = sscanf(sprintf('%c%c%c%c%c%c%c%c,',aa),'%f%*[ ,]');
end
toc
Elapsed time is 5.695740 seconds.
Elapsed time is 9.565251 seconds.
Elapsed time is 15.707137 seconds.

Anyway, thank you very much for your suggestion!