Extract numbers from mixed string

I have a file containing header lines like the following,

Test setup: MaxDistance = 60 m, Rate = 1.000, Permitted Error = 50
Operator Note:  Air Temperature=20 C, Wind Speed 16.375m/s, Altitude 5km (Cloudy)

For a given parameter such as MaxDistance or Wind Speed, I would like to extract its numerical value. This is tricky because sometimes there is an equal sign, space, or units, and sometimes there is not, because different operators enter their notes differently (lesson: next time enforce consistency).

How would I extract the following: All numerical characters (ignoring spaces and equal signs but keeping decimal points) that appear after the string representing the parameter name. Stop when a letter or punctuation mark is reached. In the case of 'MaxDistance', I would obtain 60. In the case of Wind Speed, I would obtain 16.375.

[Albert Yam]

What have you tried?

[Jianming She]

This seems a more general way:

function numArray = extractNumFromStr(str)
str1 = regexprep(str,'[,;=]', ' ');
str2 = regexprep(regexprep(str1,'[^- 0-9.eE(,)/]',''), ' \D* ',' ');
str3 = regexprep(str2, {'\.\s','\E\s','\e\s','\s\E','\s\e'},' ');
numArray = str2num(str3);

Example:

a = 'alpha=-3.5,beta=1e-2. but gamma = -34.1'
numArray = extractNumFromStr(a)

numArray =
   -3.5000    0.0100  -34.1000

Answer 1

[Jan]

Import the file into a string at first, e.g. by fileread. Then you get something like this (if not, please explain all necessary details):

Str = ['Test setup: MaxDistance = 60 m, Rate = 1.000, ', ...
       'Permitted Error = 50 Operator Note:  Air Temperature=20 C, ', ...
       'Wind Speed 16.375m/s, Altitude 5km (Cloudy)'];

Now omit all equal characters:

Str(strfind(Str, '=')) = [];

Finally you can get the values:

Key   = 'MaxDistance';
Index = strfind(Str, Key);
Value = sscanf(Str(Index(1) + length(Key):end), '%g', 1);

"Index(1)" cares for multiple occurences of the key.

[K E]

Lovely, and no regexp required. Lucas, Jan extracts the characters after the Key then gets the first number. From the scanf documentation, the %g format scans in a floating point number.

[Jan]

The removing of the = is clear, I think. Then STRFIND looks for the wanted string. Afterwards the first number behind this string is extracted by SSCANF. Here "behind" means the position, where the string is found plus the number of characters the string have.

[Lorenzo]

This works great! Just a quick question Jan: what if you want to find all the uccurrence of a numeric value between two strings? For instance, let's say you want the numeric values that can be found between MaxDistance and Altitude in the original example (i.e. 60, 1000, 50 ecc ecc...). How can you achieve that?

I tried this:

Key1 = 'MaxDistance'; Key2 = 'Altitude'; Index1 = strfind(file, Key1); Index2 = strfind(file, Key2); Value = sscanf(file(Index1:Index2), '%g',1);

but still I can get nothing but the first value.... Also, I dont know a-priori the number of numbers that can be encontured between the two strings...

Thanks!

Lorenzo

Answer 2

[Stephan Koehler]

Here is a one-line answer str2num( regexprep( Str, {'\D*([\d\.]+\d)[^\d]*', '[^\d\.]*'}, {'$1 ', ' '} ) )

[Alexandre THIBEAULT]

Best answer

[Marco A. Acevedo Z.]

hi, good answer but how to include the - sign (if present). Thanks.

Answer 3

[Freddy]

Maybe a little bit too late, but i like to present you also my ("regexp training"-) solution. :)

A = regexp(Str,'(?<Keyword>(?:\w+\s*\w+))\s*=?\s*(?<Value>\d+\.?\d*)','names');
s = struct();
for i = A, 
  s.(genvarname(i.('Keyword'))) = str2double(i.('Value'));
end

[Albert Yam]

That took a long time for me to understand what you are doing. That's cool though.

How does it skip over 'Operator Note:' ?

Edit: Never mind I get it. It doesn't have anything for ':'. The '(?:\w' has nothing to do with a ':' in the string, it is grouping the token for 'up to two words'.

Answer 4

[Albert Yam]

This is how I went about it, all steps included even the errors.

teststr = 'Test setup: MaxDistance = 60 m, Rate = 1.000, Permitted Error = 50 Operator Note:  Air Temperature=20 C, Wind Speed 16.375m/s, Altitude 5km (Cloudy)';
regexp(teststr,[\d])
regexp(teststr,['\d'])
regexp(teststr,['\d'],'match')
regexp(teststr,['\d+'],'match')
regexp(teststr,['\d+.?'],'match')
regexp(teststr,['\d+\.?'],'match')
regexp(teststr,['\d+\.?\d?'],'match')
regexp(teststr,['\d+\.?\d+?'],'match')
regexp(teststr,['\d+\.?\d*?'],'match')
regexp(teststr,['\d+\.?\d?'],'match')
regexp(teststr,['\d+\.?\d*'],'match')

[K E]

This is very useful for showing how one would construct a regular expression piece by piece. They are so cryptic if you don't use them much.

[Albert Yam]

Learning is fun.

[G]

Be careful with the last solution :

'\d+\.?\d*'

with the case:

teststr = 'Test setup: MaxDistance = 60 m, Rate = 1.000, Permitted Error = .5 Operator Note:  Air Temperature=-20 C, Wind Speed 16.375m/s, Altitude 5km (Cloudy)';

it doesn't work (negative number and '.xxx' number notation (like Permitted Error & Air Temperature in the sample)).

If someone has already done these cases ...

[G]

solved!

regexp(teststr,'\d+\.?\d*|-\d+\.?\d*|\.?\d*','match')

[G]

Better:

regexp(teststr,'\d+\.?\d*|-\d+\.?\d*|\.?\d+|-\.?\d+','match')

or

regexp(teststr,'-?\d+\.?\d*|-?\d*\.?\d+','match')

remains the -.34e-004 case !

[Angkur Shaikeea]

i need to extract

0.00000 0.00000 0.00000

0.00000 1.00000 0.00000

1.00000 0.00000 0.00000

from a text file containing

.............................................

Nodal positions:

0.00000 0.00000 0.00000

0.00000 1.00000 0.00000

1.00000 0.00000 0.00000

Nodal positions:

0.00000 0.00000 0.00000

0.00000 1.00000 0.00000

1.00000 0.00000 0.00000

Nodal positions:

0.00000 0.00000 0.00000

0.00000 1.00000 0.00000

1.00000 0.00000 0.00000

any help using regexp?

Answer 5

[Dahai Xue]

C.J. Harris, I put your regexp into a function to extract all numbers using regexp. I have hard time to find an array operation that can use the 'a' and 'b' without the loop. Hopefully somebody has ideas. Of course it is not difficult to add more parameters or options to find "certain" numbers with preceding or following landmark strings.

function nums = regExtractNums(str) 
[a,b] = regexp(str, '\d+(\.\d+)?'); 
nums = zeros(length(a),1); 
for k = 1:length(a) 
    nums(k) = str2double(str(a(k):b(k))); 
end
end

Answer 6

[C.J. Harris]

In order to extract a certain value:

Str = ['Test setup: MaxDistance = 60 m, Rate = 1.000, ', ...
       'Permitted Error = 50 Operator Note:  Air Temperature=20 C, ', ...
       'Wind Speed 16.375m/s, Altitude 5km (Cloudy)'];

matchWord = 'Air Temperature';

[a,b]  = regexp(Str,'\d+(\.\d+)?');
strPos = find(a > strfind(Str,matchWord),1,'first');
nValue = str2double(Str(a(strPos):b(strPos)));