How to combine series of consecutive numbers in Matlab?

I have following data:

s{1} = 1:10;
s{2} = 19:27;
s{3} = 29:41;
s{4} = 59:67;
s{5} = 71:89;
s{6} = 91:110;
s{7} = 119:210;
s{8} = 214:290;

and I want to combine sets with difference less than 4

   diff = s{2}(1)-s{1}(end)
  if diff <4
  combine s{1}and s{2}

My out should be:

s{1} = 1:10;
s{2} = 19:41;
s{3} = 59:67;
s{4} = 71:110;
s{5} = 119:210;
s{6} = 214:290;

[Stephen23]

27 and 29 are not consecutive, so where does the 28 come from in the output? Or 90? Do you want any missing values to be added to the data in order to have all consecutive integers within that range?

[Nikita Agrawal]

Hi, thanks for responding. I would like all the acceptable gaps to be filled with consecutive numbers

Answer 1

[Stephen23]

Bs = cellfun(@(v)v(1),s);
Es = cellfun(@(v)v(end),s);
D = Bs(2:end)-Es(1:end-1);
X = cumsum([1,D>=4]);
Bz = accumarray(X(:),Bs(:),[],@min);
Ez = accumarray(X(:),Es(:),[],@max);
Z = arrayfun(@(b,e)b:e,Bz,Ez,'uni',0);

Giving:

>> Z{:}
ans =
    1    2    3    4    5    6    7    8    9   10
ans =
   19   20   21   22   23   24   25   26   27   28   29   30   31   32   33   34   35   36   37   38   39   40   41
ans =
   59   60   61   62   63   64   65   66   67
ans =
    71    72    73    74   ...  106   107   108   109   110
ans =
   119   120   121   122   ...  205   206   207   208   209   210
ans =
   214   215   216   217   ...   286   287   288   289   290

[Nikita Agrawal]

Thanks a lot!

Answer 2

[John D'Errico]

Loops are not terrible things, to be feared in MATLAB. Yes, it is true that often you can gain in speed by replacing nested loops, especially deeply nested loops. But a simple loop?

Never bother to optimize code that is not a time or memory sink. You just waste programmer time, which is far more valuable than a tiny amount of computer time. So unless your code is seen to be a bottleneck, then you are just wasting your time (and ours.) As well, often highly vectorized code is hard to read and understand.

So just write as a simple loop. BUT DON'T USE A VARIABLE NAMED DIFF. Never use variables with the same name as a function name. Especially one you will use often.

for n = 2:numel(s)
  if s{n}(1)-s{n-1}(end) < 4
    s{n} = s{n-1}(1):s{n}(end);
    s{n-1} = [];
  end
end

When that is done, some of the cells will now be empty. Just delete all empty cells, if that is a problem. That you can do simply enough in a vectorized way.

s(cellfun(@isempty,s)) = [];

[Nikita Agrawal]

I agree to your view and thanks for bringing this into context. I am currently working on a more complicated loop and I see really less difference wrt to loop vs vectorization. Also, good that you pointed out diff not to be used. Thanks!

Answer 3

[Kaninika Pant]

You can do the following in place of combine s{1} and s{2}:

1. Append the data of s{2} in s{1}: s{2}=[s{2}, s{1}];

2. Delete s{2}: s(2)=[];

Now you will have to do all this in a loop. The following would work:

 del=1; %to store indices of the cells which will have to be removed
for i=2:8 
   diff=s{i}(1)-s{i-1}(end);
   if diff<4
       s{i-1}=[s{i-1}, s{i}];   %appending data
       del=[del, i];
   end
end
for i=2:length(del) %starting from 2 since del(1)=1 was only for initialization
    disp(del(i)-i+2)
    s(del(i)-i+2)=[]; %each time you delete the index of the next cell decreases by 1
end