extracting numbers with decimal places from the body of text file and assigning to a variable

Hi,

I have a text file that I have read in to Matlab as a character array. This file has text written in the body of it, however I am after the specific variables.

I want to extract specific values from the text to assign to their specific variable.

For example my text has something like the following in italics:

Header with text and comments

other text that I am not interested in, etc.

AAA = 18.457

BBB = 34.6

CCC = 4

I would like my results to be a series of variables

AAA = 18.457

BBB = 34.6

CCC = 4

Which I could then use to perform operations on.

I tried using the following:

fid = fopen(“file”,’r’)

text = textscan(fid,'%s','Delimiter','','endofline','');

text = text{1}{1};

fid = fclose(fid);

Expression = ‘AAA = (\d+)';

AAA = regexp(text,expression,'tokens');

However, this only printed out “18” rather than my desired “18.457” (so stopping at the decimal character). Is there a way to extract a number that may or may not have decimal places?

Ideally, I would also make it so it wasn’t sensitive to matching the exact number of spaces after the variable either “e. It just needs “AAA”, rather than “AAA “.

Is there a way to use Matlab to achieve what I want?

Answer 1

[Stephen23]

%str = fileread(..) % <- simpler way to import the file data.
str = sprintf('%s\n','Header with text and comments','other text that I am not interested in, etc.','AAA         = 18.457','BBB         = 34.6','CCC        = 4')

str = 
    'Header with text and comments
     other text that I am not interested in, etc.
     AAA         = 18.457
     BBB         = 34.6
     CCC        = 4
     '

rgx = '^\s*(\w+)\s*=\s*(\d+\.?\d*)';
tkn = regexp(str,rgx,'tokens','lineanchors');
tkn = vertcat(tkn{:}).';
tkn(2,:) = num2cell(str2double(tkn(2,:)));
out = struct(tkn{:})

out = struct with fields:
    AAA: 18.4570
    BBB: 34.6000
    CCC: 4

out.AAA

ans = 18.4570

Personally I would use a different approach: open the file, read the header lines using fgetl, then import the data using textscan. It would probably be easier than messing about with matching number formats (i.e. don't reinvent the wheel).

[Stephen23]

James Browne's "Answer" moved here:

Thanks,

Are you able to explain

rgx = '^\s*(\w+)\s*=\s*(\d+\.?\d*)'

a bit more?

particularly how the "\.?\d*" bit works?

[Stephen23]

That line defines a regular expression:

https://www.mathworks.com/help/matlab/matlab_prog/regular-expressions.html

A breakdown of what it does:

% ^\s*(\w+)\s*=\s*(\d+\.?\d*)
% ^                            match start of line
%  \s*                         match zero or more whitespace
%     (   )                    define 1st token
%      \w+                     match one or more word characters
%          \s*                 match zero or more whitespace
%             =                match equals sign
%              \s*             match zero or more whitespace
%                 (         )  define 2nd token
%                  \d+\.?\d*   match integer or decimal number

The regexp options specify to return the specified tokens and to match the caret ^ to after each newline character (i.e. the beginning of each line). Number matching broken down into its atomic operators:

% \d+\.?\d*
% \d+       match one or more digits
%    \.?    match zero or one decimal point
%       \d* match zero or more digits

If the number is integer then the decimal point and trailing digits are not matched.

[James Browne]

Thanks, I made that work with my code.

I added in "\-?" so the token is now "(\-?\d+\.?\d*)" because I also wanted to include negative numbers as possible outputs.

Instead of pulling out individual variables from the structure array (ie. with out.aaa) is it possible to make each variable in the structure array into a variable along with it's name?

[Stephen23]

"is it possible to make each variable in the structure array into a variable along with it's name?"

Possible yes, but only if you want to force yourself into writing slow, complex, obfuscated, buggy code that is difficult to debug:

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

There are so many reasons why that is a fragile, bad approach to writing your code. For example, consider what your code would do if the header name happens to be the same name as any existing variable: it would simply overwrite that variable without any warning. Such bad code design allows for all sorts of latent bugs that are difficult to track down because they depend on specific data... ugh.

If you know the headers/variables in advance then by all means allocate them explicitly:

https://www.mathworks.com/matlabcentral/fileexchange/26216-structure-fields-to-variables

If you do NOT know the headers in advance then magically creating variables from them would be a fragile, buggy, ugly approach: how would you even know what header had been imported? (trivially easy to do with the structure, quite tricky to do with randomly named variables in a workspace)