How to convert a long txt sequence of single digit numbers (e.g. '100101011101010111...') to number array?

Have have sequences of 1's and 0's in a string. Each sequence is on the order of 100 million numbers. I want to convert this char into a numerical array of the same length, so that each individual char character is a single digit number in the array (a 1 or a 0). E.g. for a length 10 char variable:

s = '1000011110'
n = [1 0 0 0 0 1 1 1 1 0] (this is what I want)

Is there any way to do this without extremely long for loops?

Answer 1

[dpb]

s = '1000011110';
n = sscanf(s,'%1d')

n = 10×1

1 0 0 0 0 1 1 1 1 0

Probably fastest. Or you can compare

s = '1000011110';
n =str2num(s.')

n = 10×1

1 0 0 0 0 1 1 1 1 0

If the string data are in a file rather than being generated internally, then use the first technique and read in as numeric to begin with...

fid=fopen('yourfile.txt');
n=fscanf(fid,'%1d');
fid=fclose(fid);

ADDENDUM

Try the following to avoid the i/o library and spoof a typecast...

s = '1000011110';
n=double(s.'-'0')

ADDENDUM SECOND:

Which above can be written as

n=double(s.')-'0';

which then MATLAB syntax will do the same thing without even the explicit cast to double()...

n=s.'-'0';

or the transpose to column that the string solutions need...

n=s-'0';

[Devin Van Elburg]

Perfect, exactly what I was looking for! First option is fastest for me, takes ~25s to convert ~150M char to num array, vs ~38s for second option.

Thank you!

[dpb]

Actually, you might also try something based spoofing a cast() operation to bypass the i/o library...

s = '1000011110';
n=double(s.'-'0')

n = 10×1

1 0 0 0 0 1 1 1 1 0

[Devin Van Elburg]

Wow, this is way faster! Takes ~1s. Thank you!!

[dpb]

Yeah, the i/o library has to be able to parse multiple character strings; taking advantage of the knowledge they're all one character and then the internal storage representation in ASCII code turns it into a simple math problem.

[Devin Van Elburg]

Follow-up question: How can I do the opposite? Is there an efficient way to go from n to s (in my example above)?

[dpb]

Yes. It shouldn't be too hard to figure out the reverse operation of the above...give it a shot, first.

HINT: You do need a function call analogous to using double() above in this one, however, MATLAB won't do the implicit conversion the other way after the needed numeric adjustment; that leaves the value as a default double as does the subtraction above. Think about what the ASCII codes are for internal storage...

[Devin Van Elburg]

Well my approach was originally this:

s = sprintf('%s',num2str(n))

But it takes ~75s to run since the n variable typically around ~10M numbers long

[dpb]

That's not the (direct) complement/inverse of the quick way, though, that's back thru the i/o libraries again...

Again, think about ASCII codes and internal storage --

BIG hint: What would it look like to add back to the number vector what we subtracted from the string to get it?

Or, approach it like an algebra problem -- we wrote above

n=s-'0';

What is the solution of the above equation for s?

[Devin Van Elburg]

Ah, so it's as simple as this:

s = char(n.' + '0');

Very quick and efficient, thank you!

[dpb]

s = char(n.' + '0');

Indeed! :)

If you want the single string instead, drop the .' transpose and leave as row vector instead...