regexp : get all the match for one expression

Hi, I have this variable :

var= 'bigFamily_middleFamily_littleFamily_childName' 
% each parts of this name can inclus numbers

And I want to extract all the family name :

bigFamily_
bigFamily_middleFamily_
bigFamily_middleFamily_littleFamily_

But with my code :

clear all
clc
var= 'bigFamily_middleFamily_littleFamily_childName'
pattern_family=sprintf('([a-zA-Z0-9]+_)*')
var_family=regexp(var,pattern_family,'match')

I only get the longest family name :

var_family = 

      'bigFamily_middleFamily_littleFamily_'

The number of name can change (can 0,1,2,3 or more). Is it possile to get a vector of match string with different strings ?

Answer 1

[Guillaume]

It's not possible to do what you want with just a regular expression. A character can only be part of one match, so once you've captured 'bigFamily_' you can't capture it again as part of bigFamily_middleFamily_ within the same regular expression.

What I would do is just capture each *family part and rebuild all the possible combinations:

var =  'bigFamily_middleFamily_littleFamily_childName';
subfamilies = regexp(var, '[a-zA-Z0-9]+_', 'match');
combinations = num2cell(logical(tril(ones(numel(subfamilies)))), 2);
families = cellfun(@(c) strcat(subfamilies{c}), combinations, 'UniformOutput', false)

Or if you don't want the trailing '_':

var =  'bigFamily_middleFamily_littleFamily_childName';
subfamilies = regexp(var, '[a-zA-Z0-9]+(?=_)', 'match');
combinations = num2cell(logical(tril(ones(numel(subfamilies)))), 2);
families = cellfun(@(c) strjoin(subfamilies(c), '_'), combinations, 'UniformOutput', false)

[Ortinomax]

Thank you for your answer,

To be complete on my need, I have a list of variable names. And I want to add a new one with his nearest family (the longer one).

So, your code will work perfectly.

EDIT : Only the version with the trailing works here. My Matlab doesn't have the "strjoin" function. but it's good.

Instead of your code I wrote something to work on my issue (but more specific for my case) :

listVar={'bigFamily_middleFamily_littleFamily_childName1',...
    'bigFamily_middleFamily_littleFamilyO_childName2',...
    'bigFamily_middleFamilyO_littleFamily_childName3',...
    'bigFamilyO_middleFamily_littleFamily_childName4'};

var= 'bigFamily_middleFamily_littleFamily_childName';
pattern_family=sprintf('([a-zA-Z0-9]+_)*');

var_family=regexp(var,pattern_family,'match');
fprintf('myVar : %s \n\tpattern_family : %s \n\tvar_family : %s \n\n\n',var,pattern_family,var_family{:});
for i=1:length(listVar)
    anotherVar=listVar{i};
    var_family=regexp(var,pattern_family,'match');

      DONE=100;
      while ~isempty(var_family)&&DONE
          var_family=var_family{end};
          [varfound index_start]=regexp(anotherVar,sprintf('%s([a-zA-Z0-9_])*[0-9]',var_family),'match','start');
          if index_start
              fprintf('VarFound : %s \n of the family : %s\n\n',varfound{end},var_family);
              %         varfound=varfound{end}

              DONE=0;
          else
              DONE=DONE-1;
              var_family=var_family(1:end-1);
              var_family=regexp(var_family,pattern_family,'match');
          end
      end
  end

If a don't find a close family, I recreate a var_family on my var_family. Well, it's not great but it works.

Now I'm adaptating your work for my case.

[Guillaume]

I'm not sure I understand what you're trying to do exactly, but if all you want to do is find the longest partial matches of var in listVar then this would be simpler:

listVar = {'bigFamily_middleFamily_littleFamily_childName1',  ...
           'bigFamily_middleFamily_littleFamilyO_childName2', ...
           'bigFamily_middleFamilyO_littleFamily_childName3', ...
           'bigFamilyO_middleFamily_littleFamily_childName4'};
var = 'bigFamily_middleFamily_littleFamily_childName';

listFamilies = regexp(listVar, '_', 'split');
varFamilies = strsplit(var, '_');
matches = cellfun(@(lf) logical(cumprod(strcmp(lf, varFamilies))), listFamilies, 'UniformOutput', false);
celldisp(matches)
matched_Families = cellfun(@(lf, m) strjoin(lf(m), '_'), listFamilies, matches, 'UniformOutput', false)

[Ortinomax]

It's all I want, but the rest of my code is what it is. I don't have a good list of cells.

But with your first answer I build something I find good.

(And in this code, there is the strsplit function I don't have on this old Matlab.)

Thank for your help.

[Guillaume]

You can replace the strsplit(x, y) with regexp(x, y, 'split')