I need to export from a cell array the char 'CL'; the problem is that I also have the 'CL/imp' chars in the array. I tried using the command regexp(str​,'CL','mat​ch'), but it takes also the from the strings containing 'CL/imp'. What should I do?

str='CL,EXAMPLE,CL/imp,example2';
% I just need to export 'CL', not 'CL/imp'
ans=regexp(str,'CL','match')
ans= 'CL'  'CL'

Answer 1

[Stephen23]

>> C = regexp(str,',','split');
>> C{strcmp('CL',C)}
ans = CL

[Alcapalb]

Thank you Stephen; do you know if there's a stronger way to do it? I mean: I need to export also other different chars in the same string at different positions. For example 'abc' and 'abcother' and sometimes I couldn't use the commas as "separator" to export just the first char ('abc') from the second ('abcother')

[Stephen23]

"do you know if there's a stronger way to do it?"

Sure, once you accurately specify the defining features of the substrings that you want to obtain, e.g. characters they must contain, cannot contain, etc. Writing regular expressions depends on first having a clear definition of what needs to be achieved, and so far your explanation of "For example 'abc' and 'abcother' " gives no useful information whatsoever: how am I supposed to generalize those? Do you only want to identify sequential alphabetic characters? If so, why do you not want to match EXAMPLE ?

I have plenty of experience with regular expressions, and I am happy to help you write one. But you need to put in some effort and clearly define what distinguished those substrings from everything else in the string.

[Alcapalb]

Ok, I will try to make my issue more clear. What distinguish some substrings from others could be inside the string is the final part of the substring. I mean, they have just the beginning of the subtring in common (e.g. "abc" and "abcother" have in common 'abc' or "CL" and "CL/imp" have in common 'CL' or "TLC/A" and "TLC/AC" have in common 'TLC/A'). My problem comes if I have to isolate/export the first of the chars ("abc" or "CL" or "TLC/A"), because writing the code shown in my first question, I see in the ans string also all the second chars. Immagine I have this: cell='CL,EXAMPLE,CL/imp,example2' 'try,TLC/A,TLC/AC' 'bonus' 'abc,example,other,ole'

I want to process this cell several times: occasionally, for example, if I need to export 'CL'; or 'TLC/A' I face problems to distinguish from substrings with the same first part like 'CL/imp' or 'TLC/AC'

[Stephen23]

@Alcapalb: A few things need to be clarified:

1. What defines the boundaries of these substrings? Earlier you wrote that '...sometimes I couldn't use the commas as "separator"', which implies the you do not want to use the comma to define the boundaries of these substrings. In this case, what is stopping us from matching the 'le' in 'example' to the 'le' in 'ole' ?:

'abc,example,other,ole'

If the comma does not always indicate the boundaries, then what characters are permitted in the substrings themselves? Or does the comma always indicate the boundaries of a substring?

2. Does the shorter substring always precede the longer substring?

3. Do you want to match case, or shall case be ignored when matching?

[Alcapalb]

@Stephen, I'm sorry about being not clear.

1. I use the command readtable and then table2cell to take and process data from an excel file. Afterwards, I focus on one of the cell array coming out from table2cell. So, the length of the array is known; each string of this cell array is composed in various way by different chars (sometimes empty, too): that's why I wrote that commas cannot be used always as boundary, because I just take out data from an excel column and I don't know actually which info are there. I just want to tell matlab to give me back another array with the same length of the processed one, containing just a specific substring like e.g.'CL' or 'TLC/A'. In other words I have to filter the excel column asking each time for a different substring ('CL', 'TLC/A' and so on).

My problem is that when I ask matlab with the command ans=regexp(str,'CL','match') to take 'CL' out of the cell array, it includes also 'CL/imp' which I don't want to be included. And the same happens for all the chars have a part in common, like 'TLC/A' and 'TLC/AC' and so on.

2. No, it doesn't.

3. Yes, I do.

Hope I was enough clear, now, but please ask me again if you need to know more. Thank you very much

[Stephen23]

@Alcapalb: what I asked was:

1. What I asked about is how you know the boundaries of the substring, e.g.: why do you only consider CL to be a substring. Why not L, or le which is part of ole? Or mp from CL/imp also matches mp from example2. You state that the commas are not the boundaries. What defines the bounds of these substrings?

2. Thank you.

3. That was not a yes/no question.

[Alcapalb]

1. Because CL is the exact acronym I am looking for, as so many others I need to isolate and export. Why not L or le? Because I don't need it. The boundaries of these substrings are defined by their own meaning.

3. You asked "do you want..." and I answered: yes I do, which means yes I want to match case

Thank you, Stephen

[Stephen23]

@Alcapalb:

1. I give up asking about this, and will just assume that the comma defines all substring boundaries. The comma as separator seems to match all of your examples.

4. Does the desired substring always occur at the start of the longer substring?

5. Do you know the substring in advance and just want its location, or are you trying to identify it?

[Stephen23]

Perhaps this does something like what you want:

>> str = 'CL,EXAMPLE,CL/imp,example2';
>> C = regexp(str,',','split');
>> fun = @(s)nnz(strncmp(s,C,numel(s)))>1;
>> C{cellfun(fun,C)}
ans = CL

It identifies any repeated substrings (bounded by commas), where the second occurrence can be part of a longer substring.

[Alcapalb]

1. "I don't know about the boundaries of the substrings, but I know that the substrings only contain the following characters: alphabetic, underscore, or forward-slash".

4. Yes, it does occur at the start

5. I know the substring in advance and want its location

[Stephen23]

Perhaps you want something like this:

>> [mtc,idx] = regexp(str,'CL(?=[^A-Z/_])','match','start','once')
mtc = CL
idx =  1

It gives the start location. It only matches CL when it is not followed by the characters that you have specified.

[Alcapalb]

Thank you very much, Stephen. It works partially. In the input cell array, when 'CL' is not followed by space (if CL is 1x2char), it doesn't give me back 'CL'.

[Stephen23]

Perhaps this:

>> str = 'CL';

>> [mtc,idx] = regexp(str,'CL(?=($|[^A-Z/_]))','match','start','once')
mtc = CL
idx =  1

It would help a lot if you actually accurately specified what you want to match, or perhaps more usefully what you do NOT want to match.

[Alcapalb]

Thank you very much, Stephen. It works, now. Sometimes "beginners" might be not clear when they try to expose their problems.

Answer 2

[Guillaume]

Alcapalb, what Stephen is asking is a clear definition of what need to be matched and what must be rejected as well as what are the boundaries for whatever you want to match. We can make regular expressions match almost anything and reject almost anything but the rules need to be exactly defined.

For example, you can write

matches = regexp(str, 'CL(?!/imp)', 'match')

That would not match 'CL' in 'CL/imp' so would answer your question. However, it still would match 'CL' in 'CL/abc' and in 'abcCLdef' and in 'CLimp', etc. No idea if that's what you want because you haven't clearly defined what is allowed and not allowed.

Perhaps you don't want 'CL' followed by a '/', in which case we change the regular expression to

matches = regexp(str, 'CL(?!/)', 'match')

No more 'CL/abc' but it is still matching 'abcCLdef'.

Perhaps what you want to match is instead

matches = regexp(str, '(?<=^|[^a-zA-Z])CL(?=$|[^a-zA-Z])', 'match')

This would match 'CL' as long as it's preceded by the start of the string or a non-letter and as long as it's followed by the end of the string or a non-letter. A completely different set of rules that results in a completely different regular expression.

All of these could be the answer to your question or it could be something completely different because you haven't specified.

Furthermore, in your question you're asking for the 'match' which is a bit pointless as it's always going to be 'CL'. So there's probably something more about that you've also not stated.

Finally, don't use ans as a variable as it's a special variable that matlab is likely to overwrite.

[Alcapalb]

Thank you, Guillaume. What you wrote will be precious as well.