Search for a part of a string in cell of cell array

Hello everyone,

i've got some trouble with searching for a string in a cell array. I've tried looking it up, but I didn't find a solution which is working for me.

In fact, I got so confused with cell arrays or cell-in-cell-arrays that I don't know which one I am searching in... When I use the class-function on my 1x68 cell "VarNames" it says:

class(VarNames)
ans = cell
class(VarNames(1))
ans = cell
class(VarNames{1})
ans = char

So do I have a cell-in-cell-array or just a cell array with strings? And how do I search in it? It looks like this:

VarNames = 

Columns 1 through 5

    'Trend 1 - abc def'     'Trend 2 - ghi mmm jkl'     'Trend 3 - mno ttt kkk pqr'     'Trend 4 - stu vwx aaa'     []

Columns 6 through 10

    'Trend 6 - yzz yxw'     'Trend 7 - vut rrr srq'     'Trend 8 - pon mlk'     'Trend 9 - jih gfe'

So there are empty cells in it as well as a non-constant amount of chars. If I want to look for the cell-index in which the string "vwx" is included, how do I do it? I tried

find(strcmp('vwx',VarNames),1))
ans =

     Empty matrix: 1-by-0

I also tried lots of other functions like strncmp, regexp or using cellfun to access the cell array, but I can't get it to work. Any advice?

Thanks in advance!

J.

[Stephen23]

Cell array indexing is simple:

• curly braces {} refer to the cell contents.
• parentheses () refer to the cells themselves.

Note that this usage of parentheses () is consistent in all of MATLAB: if you use them to access part of a numeric matrix, you get a numeric matrix. Use them to access part of a table, you get a table. Use them to access part of a call array, you get a cell array. Ditto character, etc.

This is totally different to curly braces {}, which refer to the data inside a cell array or table.

[dpb]

"Cell array indexing is simple:"

Excepting when it isn't, of course... :) The biggest complaint I have is one can't use the two-level indexing expressions when the cell reference is anything but a single cell--

>> C={rand(3)};
>> C{2}=sqrt(C{1})
C = 
  [3x3 double]    [3x3 double]
>> C{2}(:,1)  % get the first column of the second cell... ok
ans =
  0.8098
  0.1890
  0.9215
>> C{:}(:,1)  % what if want the same column from all cells???
Bad cell reference operation. 
>>

[Guillaume]

You probably know that but that's because C{:} is a list and mathworks never went further in implementing lists.

There's a lot of indexing that breaks down in matlab if you try to chain it.

[dpb]

Yes, just pointing out one of the many "issues" that can easily arise wherein cell addressing isn't necessarily so easy as it appears when described in the simplest of terms...

[Stephen23]

So people can read it: Cell array indexing is simple:

• curly braces {} refer to the cell contents.
• parentheses () refer to the cells themselves.

Note that this usage of parentheses () is consistent in all of MATLAB: if you use them to access part of a numeric matrix, you get a numeric matrix. Use them to access part of a table, you get a table. Use them to access part of a call array, you get a cell array. Ditto character, etc.

This is totally different to curly braces {}, which refer to the data inside a cell array or table.

[dpb]

Yes, that's true as far as it goes. The problems arise as for the OP when one actually tries to make use of the content inside the cell array under general programming constructs; then the interactions and gotcha's! arise.

As in the example we finally got to on down; it turns out that OP placed an empty numeric cell inside the rest of a cell array of strings. That, when addressed by the string functions, hiccuped (not surprisingly, agreed) but it isn't at all apparent at the higher level where things went south to the newer user.

So it's not that the individual components are difficult granted, it's that in application one gets into situations that aren't that easy to solve. Some of these are errors as in OP's case which he solved by correcting the underlying issue (altho he didn't really recognize the issue precisely as being type, not content); others are that when addressing portions of interest for a given computation the result isn't always just a cell or a whatever, it's a list of whatevers which has it's own set of rules. On top of which there's the problem you can't always even generate the desired list as illustrated above.

Answer 1

[dpb]

It's a cellstr array, and yes, it can be confusing how to address them. The content of the cell is a character string referenced by "the curlies" where as the cell or a set of cells are addressed with regular parens. There's just no difference in what V or V(1) are; the latter is just a single cell of the former array. The V{} notation, however, returns the CONTENT of the cell which is something entirely different (albeit that may also be a cell or a double or a handle or, as in this case, a string (which is, in turn, an array of char()).

I put a subset into a shorter variable name V here for illustration --

>> whos V   % a cell array of nine elements
Name      Size            Bytes  Class    Attributes
V         1x9               832  cell               
>> v=V(1); whos v  % make an object of the first cell -- same excepting just one element now
  Name      Size            Bytes  Class    Attributes
  v         1x1                94  cell               
>> v=V{1}; whos v  % samething for the _content_ it contains a character string (array of char)
  Name      Size            Bytes  Class    Attributes
    v         1x17               34  char               
>> v=v(1); whos v  % the content within -- one element is just a single letter now...
  Name      Size            Bytes  Class    Attributes
    v         1x1                 2  char               
>> v
v =
T
>> V(1)
ans = 
    'Trend 1 - abc def'
>> V{1}
ans =
Trend 1 - abc def
>>

Note in particular the difference between the last two -- the cell content is displayed inside single quotes which delimit a cellstring; the content is shown as simply the string of characters themselves; no quotes.

Cell strings are a great convenience as you can reference them with a single subscript and many functions are enhanced to handle them. But, it is somewhat roundabout road to get results such as you're looking for from the search functions, etc., granted because operations on cell arrays tend to return cell arrays.

To find your substring, try something like

>> ~cellfun(@isempty,strfind(V,'vwx'))
ans =
   0     0     0     1     0     0     0     0     0

which is a logical array of T for the cell index(/indices) which contain the subject string, F elsewhere. You can get the single numeric array index by wrapping this in a call to find --

>> find(~cellfun(@isempty,strfind(V,'vwx')))
ans =
   4
>>

Note here that cellfun is the way to get the results of the test for the various cell content compacted into an "ordinary" vector from the cell array of individual results. It takes some getting used to, but with familiarity it becomes more or less natural.

[Johannes Elfner]

Thanks for your help! I already tried using cellfun but I wasn't able to get a result due to empty cells in my cell array. Is there any workaround?

[dpb]

I'm sure there is but unless you show us the exact code used and the actual data values, "not so much". What's wrong with the example I posted above?

[dpb]

In rereading your initial posting looking for further klews as to where you might have gone wrong I see something I missed earlier -- in

find(strcmp('vwx',VarNames),1))

you've used strcmp which is for an exact match of the two strings so the empty matrix result is the expected result. NB: the use of strfind to find a substring within another string. Also note that you have the shorter pattern for which you're searching in the position of the searched-within string so that the search is backwards--you're looking for the whole string inside the pattern. This will fail.

[Johannes Elfner]

Yea, strcmp was not the right choice... Especially if used like above. But I tried it in all kinds of combinations and the way I wrote it down here was just what I thought it would be the easiest way to show my problem. And well, I failed at that... ;) I've been using strcmp and mostly strncmp in the right way in my code, but still didn't get any results. I guess it was due to empty cells like in cell 3 of V:

V = 

    'abc'    'abc'    []    'abc'    'abc'

My workaround is to fill all empty cells with strings:

IndexEmptyFields = cellfun('isempty',V);
V(IndexEmptyFields) = {'-'};

Using this workaround searching the cell-array works fine using either your cellfun-code or strncmp. Thanks for your help!

[dpb]

V = 'abc' 'abc' [] 'abc' 'abc'

There's your problem, indeed--

>> class(V{3})
ans =
double
>>

You're trying to treat cell content that isn't a string as if it were; "You can't do that!" :) It's the right idea above but to be safer, use the empty string {''} to fill the locations (or, better yet, use that when creating the cellstr array in the first place); then there's still the "isempty" return to identify the locations but the string functions will function as intended.

Answer 2

[Guillaume]

VarNames = {'Trend 1 - abc def', ...
            'Trend 2 - ghi mmm jkl', ...
            'Trend 3 - mno ttt kkk pqr', ...
            'Trend 4 - stu vwx aaa', ...
            [], ...
            'Trend 6 - yzz yxw', ...
            'Trend 7 - vut rrr srq', ...
            'Trend 8 - pon mlk', ...
            'Trend 9 - jih gfe'};
  VarNames(cellfun(@isempty, VarNames)) = {''}; %replace empty matrices by empty strings
  idx = find(~cellfun(@isempty, strfind(VarNames, 'vwx'))) %find cells that match

As to indexing () vs {}:

• () returns a portion of the cell array. It always output a cell array, even if you only ask for one cell. You use this when you want to manipulate the container and don't care about what's in the cells.
• {} returns the content of the cell array. It is always of the type of whatever is inside the cells. You use this when you want to access what is in the container.