Using sscanf and NaN

Greetings,

So if i have the following:

a = '21414839.260   112535757.19909      2294.184    21414838.320    87690207.10148 1787.672'

Using:

 A=sscanf(a,'%f')

Will return:

A =

     1.0e+08 *

      0.2141
      1.1254
      0.0000
      0.2141
      0.8769
      0.0000

But if If the last 3 numbers are missing from the line :

a = '21414839.260   112535757.19909      2294.184                                        '

I want this to be returned:

 A =

     1.0e+08 *

      0.2141
      1.1254
      0.0000
      NaN
      NaN
      NaN

How can I make this possible?

Answer 1

[Geoff Hayes]

If you know that sscanf of the input string is to return 6 numbers, and A is only populated with three, then you could do the following

 A = [A; repmat(NaN,6-length(A),1)];

[Sebastian Ciuban]

Actually, I want to return NaN where is white space instead of numbers. The size of A doesn,t matter.

e.g. 6 numbers

a = '21414839.260   112535757.19909      2294.184    21414838.320    87690207.10148 1787.672'

Applies the command:

A=sscanf(a,'%f')

Returned:

A =
     1.0e+08 *
      0.2141
      1.1254
      0.0000
      0.2141
      0.8769
      0.0000

eg. with 5 numbers

a = '21414839.260   112535757.19909      2294.184    21414838.320    87690207.10148       '

Applies the command:

A=----------

To be returned:

A =
     1.0e+08 *
      0.2141
      1.1254
      0.0000
      0.2141
      0.8769
      NaN

And so on..If I have 4,5 numbers..where the readinf of line interceps white spaces it must return NaN

[Geoff Hayes]

But if there is just white space and no other indication, how is this function (or any other) supposed to know that there should be five numbers or six or four or none? Look at a - it is just a string with some numbers separated by whitespace which may have a block of trailing spaces when there are missing numbers. If the size of this block is x characters how do you determine how many numbers should be there? If all of your numbers had the same digit count, then you could probably make a guess…but without any other information…

How is a determined?

[Sebastian Ciuban]

The function is this:

function [obs] = readRINEXobs()

%Opening the selecting menu
[file,path] = uigetfile('*.**o');

%Opening the file
fid = fopen(strcat(path,file),'r');

%Reading until END of HEADER 
a = fgets(fid);
while (~strcmp(a(61:63),'END'))
a = fgets(fid);
end
k = 0;
while ~feof(fid)

a = fgets(fid);
sat=a;
index=strfind(a,'G'); 
[~,n]=size(index); 

for i=1:n
    A(1,i)=str2num(sat(index(i)+1:index(i)+2));
    a = fgets(fid);
    A(2:6,i) = sscanf(a,'%f');
    a = fgets(fid);
    A(7,i) = str2num(a);

end
k = k+1;

obs(k).data=A;

end

[Geoff Hayes]

Is the problem at the line

 A(2:6,i) = sscanf(a,'%f');

where you try to populate five values in A but the sscanf returns less than five (leading to the Subscripted assignment dimension mismatch.? If that is the case, then you have made an assumption on the number of floats that you are trying to read in from a and that is 5. So the above is still applicable

% read what you can from the string and format as numbers
temp     = sscanf(a,'%f');

% initialize the ith column of A from rows 2 through 6 with the five
% values from temp, or those values that you could read plus NaNs
A(2:6,i) = [temp; repmat(NaN,5-length(temp),1)];

[Sebastian Ciuban]

Thank you! This is it :)

[Geoff Hayes]

You're welcome!

[Sebastian Ciuban]

Can I ask your assistance again for a small thing?

[Geoff Hayes]

For sure!

[Sebastian Ciuban]

This function reads blocks of data from an ASCII file. Some of them contains white space, problem which was solved with your help and Star Strider's. The thing is I get extra data in the blocks and I can't figure from where.

I think this may be too much to ask, but if I archive the function + data file..Can you hit a "Run" a take a small look?

[Geoff Hayes]

Just attach your files to a comment and I can take a look.

[Sebastian Ciuban]

The function will return a structure 1x2880. The number means the blocks of data that have been read..If you open in matlab the file RINEXOBS13.o, and scroll at the end of file you will see that the last block of data has 8 satellites..In my function, the last one has 12 satellites..and it does that to many blocks, adding extra information.

[Sebastian Ciuban]

I think the problem is in the FOR loop but I can't figure what exactly

[Sebastian Ciuban]

This is it

[Geoff Hayes]

Hi Ciuban - the problem is that A matrix that is being populated with the satellite information is not being reset/cleared for each block:

 while ~feof(fid)
   % do some stuff
   [~,n]=size(index); 
   for i=1:n
     A(1,i)=str2num(….
     % etc.

So once this matrix reaches size 12 (for whatever iteration of the while loop) then it holds on to that excess data for each subsequent block that has less than 12 satellites. When the last block of eight satellites is evaluated, A keeps the last four from some previous block(s). To get around this, just re-size the A matrix prior to entering the for loop since you know how many rows (7) and you know the columns (n):

while ~feof(fid)

    tline = fgets(fid);
    sat=tline;
    index=strfind(tline,'G'); 
    [~,n]=size(index); 

    A = zeros(7,n);  % add this line !!!!

    for i=1:n
       % etc.

Pre-allocation of a matrix is usually a good idea - it prevents a matrix from being resized on subsequent iterations (may improve speed), and clears those pesky warnings in the editor.

Also, don't forget to add in the fclose(fid); statement. Hope the above helps!

[Sebastian Ciuban]

I don't know how to thank you for your patience and help!

[Geoff Hayes]

It was fun!

Answer 2

[Star Strider]

Use sscanf with two output arguments:

% Known input length
L = 6;
a = '21414839.260   112535757.19909      2294.184'
[A,  knt] = sscanf(a, '%f');
A(knt+1:L) = NaN;

produces:

A =

      21.4148e+006
     112.5358e+006
       2.2942e+003
               NaN
               NaN
               NaN

[Sebastian Ciuban]

Thank you for your answer!

[Star Strider]

My pleasure!