Thread Subject: How to 'unnest' a cell array of cells?

Here's an example using output of REGEXP's token-search
facility:

string = 'blah bleh blah'
pattern = 'bl(.)h'
t = regexp(string,pattern,'tokens')

t =

    {1x1 cell} {1x1 cell} {1x1 cell}

>> t{:}

ans =

    'a'


ans =

    'e'


ans =

    'a'


class(t{1})

cell


I would like to have the tokens packed in a 'non-nested'
cell array. Is the loop below the best approach?

n = length(t);
c = cell(n,1);
for i = 1:n
    c{i} = cell2mat(t{i});
end
c

c =

    'a'
    'e'
    'a'

class(c{1})

char

Thank you.

Hi dimitri,

try [t{:}] on your nested cell.

- diederick

Ooops! :) Thank you, Diederick.

But what would you suggest when there are multiple strings
on input:

My example:
strings = {'b=1 c=2'; 'b=2 c=1'; 'b=3 c=2'};
pattern = 'b=(\d) c=(\d)';
t = regexp(strings,pattern,'tokens');

>> [t{:}]

ans =

    {1x2 cell} {1x2 cell} {1x2 cell}

I am looking for the kind of output that is generated by:
[b,c] = cellfun(@getbc, strings)

b =

     1
     2
     3


c =

     2
     1
     2

with:

function [b,c] = getbc(str)
t = regexp(str, 'b=(\d) c=(\d)', 'tokens');
tt = [t{:}];
b = str2num(tt{1});
c = str2num(tt{2});

Anyway to do it without using cellfun?
Rik

Try using regexp's named tokens feature:

strings = {'b=1 c=2'; 'b=2 c=1'; 'b=3 c=2'};
t = regexp(strings, 'b=(?<b>\d) c=(?<c>\d)', 'names')
t = [t{:}]
b = [t.b]'
c = [t.c]'

Documentation of named tokens is here:
http://www.mathworks.com/access/helpdesk/help/techdoc/matlab_prog/f0-42649.html#bqm94nz-1

-=>J

Thanks, but note that this doesn't work for b, c > 9:

clear all
strings = {'b=1 c=2'; 'b=2 c=10'; 'b=11 c=2'};
t = regexp(strings, 'b=(?<b>\d{1,}) c=(?<c>\d
{1,})', 'names');
t = [t{:}];
% the following does not produce the desired result:
b1 =[t.b]'
c1 =[t.c]'

% it requires something like
[bb{1:length(t)}] = deal(t.b);
[cc{1:length(t)}] = deal(t.c);
b2 = char(bb)
c2 = char(cc)

or ?


"Jason Breslau" <tendiamonds@mathworks.com> wrote in
message <fj96o9$hju$1@fred.mathworks.com>...
> Try using regexp's named tokens feature:
>
> strings = {'b=1 c=2'; 'b=2 c=1'; 'b=3 c=2'};
> t = regexp(strings, 'b=(?<b>\d) c=(?<c>\d)', 'names')
> t = [t{:}]
> b = [t.b]'
> c = [t.c]'
>
> Documentation of named tokens is here:
>
http://www.mathworks.com/access/helpdesk/help/techdoc/matla
b_prog/f0-42649.html#bqm94nz-1
>
> -=>J

cellfun(@(c) c{1}, t, 'UniformOutput', false)

Bruno

cellfun(@(c) c{1}, t, 'UniformOutput', false)

Bruno

cellfun(@(c) c{1}, t, 'UniformOutput', false)

Bruno

This will work:

strings = {'b=1 c=27'; 'b=2 c=1'; 'b=34 c=2'};
t = regexp(strings, 'b=(?<b>\d+) c=(?<c>\d+)', 'names');
t = [t{:}];
char(t.b)
char(t.c)

-=>J

"Rik " <rdewijn@xs4all.nl> wrote in message
<fjau2t$665$1@fred.mathworks.com>...
> Thanks, but note that this doesn't work for b, c > 9:
>
> clear all
> strings = {'b=1 c=2'; 'b=2 c=10'; 'b=11 c=2'};
> t = regexp(strings, 'b=(?<b>\d{1,}) c=(?<c>\d
> {1,})', 'names');
> t = [t{:}];
> % the following does not produce the desired result:
> b1 =[t.b]'
> c1 =[t.c]'
>
> % it requires something like
> [bb{1:length(t)}] = deal(t.b);
> [cc{1:length(t)}] = deal(t.c);
> b2 = char(bb)
> c2 = char(cc)
>
> or ?
>
>
> "Jason Breslau" <tendiamonds@mathworks.com> wrote in
> message <fj96o9$hju$1@fred.mathworks.com>...
> > Try using regexp's named tokens feature:
> >
> > strings = {'b=1 c=2'; 'b=2 c=1'; 'b=3 c=2'};
> > t = regexp(strings, 'b=(?<b>\d) c=(?<c>\d)', 'names')
> > t = [t{:}]
> > b = [t.b]'
> > c = [t.c]'
> >
> > Documentation of named tokens is here:
> >
> http://www.mathworks.com/access/helpdesk/help/techdoc/matla
> b_prog/f0-42649.html#bqm94nz-1
> >
> > -=>J
>

Thanks, it does ...

Interestingly, according to the "Quantifiers" table in the
Matlab regular expression
expr+ is equivalent to expr{1,}, but I guess here it isn't.

Rik

"Jason Breslau" <tendiamonds@mathworks.com> wrote in
message <fjbrdv$9h$1@fred.mathworks.com>...
> This will work:
>
> strings = {'b=1 c=27'; 'b=2 c=1'; 'b=34 c=2'};
> t = regexp(strings, 'b=(?<b>\d+) c=(?<c>\d+)', 'names');
> t = [t{:}];
> char(t.b)
> char(t.c)
>
> -=>J
>
> "Rik " <rdewijn@xs4all.nl> wrote in message
> <fjau2t$665$1@fred.mathworks.com>...
> > Thanks, but note that this doesn't work for b, c > 9:
> >
> > clear all
> > strings = {'b=1 c=2'; 'b=2 c=10'; 'b=11 c=2'};
> > t = regexp(strings, 'b=(?<b>\d{1,}) c=(?<c>\d
> > {1,})', 'names');
> > t = [t{:}];
> > % the following does not produce the desired result:
> > b1 =[t.b]'
> > c1 =[t.c]'
> >
> > % it requires something like
> > [bb{1:length(t)}] = deal(t.b);
> > [cc{1:length(t)}] = deal(t.c);
> > b2 = char(bb)
> > c2 = char(cc)
> >
> > or ?
> >
> >
> > "Jason Breslau" <tendiamonds@mathworks.com> wrote in
> > message <fj96o9$hju$1@fred.mathworks.com>...
> > > Try using regexp's named tokens feature:
> > >
> > > strings = {'b=1 c=2'; 'b=2 c=1'; 'b=3 c=2'};
> > > t = regexp(strings, 'b=(?<b>\d) c=(?<c>\d)', 'names')
> > > t = [t{:}]
> > > b = [t.b]'
> > > c = [t.c]'
> > >
> > > Documentation of named tokens is here:
> > >
> >
http://www.mathworks.com/access/helpdesk/help/techdoc/matla
> > b_prog/f0-42649.html#bqm94nz-1
> > >
> > > -=>J
> >
>