Remove letters tabs and spaces from a string with numbers and letters while creating a vector with the remaining numbers

Hello,

I have a file called "SSL.dat". The second line has some numbers I need but it's full of spaces and letters.

fidssl = fopen('SSL.txt','r');
line1ssl=fgetl(fidssl);              
line2ssl=fgetl(fidssl);              %line where the data I need is

This line has the following contents:

mass(Msun)  1.000       9.547922048e-4  2.858857575e-4  4.366245355e-5  5.151391279e-5  8.3459e-9       3.040433607e-6  1.660477111e-7  2.448326284e-6  3.227778035e-7  6.607313077e-9

There are spaces, not tabs, separating each column. I need to remove the letters and split each mass value into a vector so that the vector index matches the column number of said value.

For example, the value "1.000" would match mass(1) and so on.

Can anyone help?

[Stephen23]

Use textscan. It should do the job perfectly. If you have any questions about making it work please upload your file by clicking the paperclip button above the textbox.

Also note that MATLAB uses one-based indexing, so mass(0) would be an error.

[Frederico Arez]

will try thank you

[Frederico Arez]

I attached the file. textscan works great on the first line, but I can't get it to work on the second line (the one in the question). What format do I use and how do I then insert it into a vector?

When I used name = textscan(line1SSL,'%s') on the first line of the file it gave me a "name{}{}".

[dpb]

Excepting if really is space-delimited file you've got a problem because the first line (after the header) you posted has one more column of data in it than the rest. You'll have to treat the two record types separately as there's no way to encode an empty blank field in C (and hence Matlab) input formatting rules, a very big weakness, unfortunately.

[Stephen23]

That is a very badly formatted file: it uses multiple spaces (and tabs) as the delimiter. Therefore in general a missing field cannot be detected using a simple delimiter (e.g. comma), nor by a fixed width because the combination of tabs and spaces actually means that each column has a different number of characters, depending on the row. Ouch.

Untested, but this should get you started:

opt = {'MultipleDelimsAsOne',true};
hdr = repmat('%s',1,12);
fmt = ['%12c%12c',repmat('%f',1,10)];
fid = fopen('SSL.txt','rt');
H = textscan(fid,hdr,1,opt{:});
C = textscan(fid,fmt,opt{:});
fclose(fid);

Answer 1

[Frederico Arez]

I got a solution. When I use textscan() it keeps everything store as strings. All I have to to is convert them to doubles like so,

fidssl = fopen('SSL.dat','r');
line1_ssl=fgetl(fidssl);
planet_name_list = textscan(line1_ssl,'%s');
line2_ssl=fgetl(fidssl);
masslist = textscan(line2_ssl,'%s');

for i=1:12
   mass(i)=str2double(masslist{1}{i+2})
end

seems to work

[dpb]

See alternative solution that reads the file directly; returns eleven (11) elements/row with NaN missing-value indicator for records past the first. I'm a little surprised; not sure precisely why this is...

Answer 2

[dpb]

Actually, to my surprise textscan did a better job than I expected...

>> v=mat2cell(textscan(fid,['%*s' repmat('%f',1,11)],'headerlines',1,'collectoutput',1))
v =
  1.0000    0.0010    0.0003    0.0000    0.0001    0.0000    0.0000    0.0000    0.0000    0.0000    0.0000
  5.2043    9.5820   19.2294   30.1036   68.1435    1.0000    0.3871    0.7233    1.5237   39.2644       NaN
  0.0488    0.0557    0.0444    0.0112    0.4325    0.0167    0.2056    0.0068    0.0933    0.2447       NaN
  1.3046    2.4853    0.7726    1.7680   43.7407    0.0001    7.0050    3.3946    1.8499   17.1514       NaN
188.1847  320.3479  142.9558  267.7650  194.2978  357.5452  174.7959   50.0115   19.3565   15.0233       NaN
275.0659  336.0136   96.5417  265.6496  150.8411  322.6181   29.1243   55.1860  286.5374  113.7629       NaN
100.4917  113.6430   73.9893  131.7939   36.1284  140.2998   48.3305   76.6678   49.5620  110.2869       NaN
>> v(1,end)
ans =
 6.6073e-09
>>