extract a variable from middle of a line of a .dat file with mixed text and numbers

Hello.

I'm a fairly new user of MatLab and I'm having some difficulty reading data from a .dat file.I want to extract specific values from a line of the file I'm using. That line has both text and the values i want. From the sample I'm posting bellow I want to extract the values of LAT i.e. 50 55.51 and LONG i.e. -40 26.26. I would like to get 50 55.51 -40 26.26 as different variables so I can use them to convert to decimal coordinates 50.93 and -40.44, using something like this:

lat_deg = 50      % extract from file.dat
lat_min = 55.51   % extract from file.dat
lon_deg = -40      % extract from file.dat
lon_min = 26.26   % extract from file.dat

lat = lat_deg+lat_min/60
lon = lon_deg+long_min/60

_________________sample____________________________

STATION 2   DATE 1 27 1999   TIME 14:57 TO 17:08
LAT 50 55.51  LONG -40 26.26   WATER DEPTH 4060 M

   PRESS   TEMP   SALINITY  RHO(T,S,P)
    (M)     (C)   (PSS-78)  (KG M-3)

      2.   18.623   35.658   1025.35
      5.   18.627   35.657   1025.56
     10.   18.623   35.658   1025.79

The computers I use run Matlab R2008a, but sometimes I'll need to use the older Matlab 6.5 version. Any help is appreciated.

edited: to correct LAT/LONG values and indicate Matlab versions I'll be using

Answer 1

[Cedric]

If LAT/LONG can really be anywhere in the file (not on 2nd line as shown in your sample), you could go for a regexp solution:

 >> buf = fileread('myData.dat') ;

 >> pat = 'LAT\s+(-?[\d\.])+\s+(-?[\d\.]+)\s+LONG\s+(-?[\d\.]+)\s+(-?[\d\.]+)' ;
 >> tokens = regexp(buf, pat, 'tokens') ;

 >> lat = str2double(tokens{1}(1:2))
 lat =
   50.0000   55.5100

 >> long = str2double(tokens{1}(3:4))
 long =
  -40.0000   26.2600

There are thousands of ways to implement this with REGEXP; here is another possibility:

 >> buf = fileread('myData.dat') ;

 >> match = regexp(buf, '(?<=LAT\s+)-?[\d\.]+\s+-?[\d\.]+', 'match') ;
 >> lat = str2num(match{1})
 lat =
   50.0000   55.5100

 >> match = regexp(buf, '(?<=LONG\s+)-?[\d\.]+\s+-?[\d\.]+', 'match') ;
 >> long = str2num(match{1})
 long =
  -40.0000   26.2600

Let me know if you want more information about how we are using REGEXP here.

[matlab_student]

I've only tested your solution briefly (the first example), but i think it works for me. I'll experiment more solutions latter. LAT/LONG are always on the 2nd line, as shown in the sample. Which code do you think it's best for my intentions? Yours or the one from Andrei Bobrov? Thanks for the help.

[Cedric]

I think that you can go for simpler than both Andrei and I solutions if LAT/LONG are really always on second line, e.g.

 >> fid = fopen('myData.dat', 'r') ;
 >> fgetl(fid) ;                                      % Skip 1st line.
 >> coord = fscanf(fid, 'LAT %f %f LONG %f %f') ;
 >> fclose(fid) ;

with that:

 >> coord
 coord =
   50.0000
   55.5100
  -40.0000
   26.2600

which you can easily use to compute the final longitude and latitude.

[matlab_student]

Yes, that's simple enough and does what i need. Again, thank you for your helpful input.

[Cedric]

You're welcome!

Answer 2

[Andrei Bobrov]

f = fopen('your_data.txt');
c = textscan(f,'%s');
fclose(f);
c1 = c{:};
d = str2double(c1(bsxfun(@plus,find(ismember(c1,{'LAT' 'LONG'})),1:2)));
d(:,2) = sign(d(:,1)).*d(:,2);
out = d*[1;1/60];

[matlab_student]

Thank you for your input. That code works for most of what i want. But i think it won't run on some machines i use with the older MatLab 6.5 installed. Any way i can use textread or some similar function instead?