Loren on the Art of MATLAB

Turn ideas into MATLAB

Note

Loren on the Art of MATLAB has been archived and will not be updated.

Indexing – Mixing It Up

I've written a bunch of posts related to indexing, but I still haven't showed all the flexibility allowed. Today I'm going to add a variant to the list.

Contents

Types of Indexing - a Review

Let's start with indexing using subscripts. Let's start by creating a matrix.

A = [1 2 3 4; 5 6 7 8; 9 10 11 12; 13 14 15 16]
A =
     1     2     3     4
     5     6     7     8
     9    10    11    12
    13    14    15    16

If I want to pick out the element sitting in the second row, third column, I simply write this

el23 = A(2,3)
el23 =
     7

and you see result 7.

If I want another element as well, let's say 4th row, 2nd element, I can certainly follow my last statement with a similar one.

el42 = A(4,2)
el42 =
    14

If I want both those elements extracted, I can try combining their respective row and column indices.

els = A([2 4],[3 2])
els =
     7     6
    15    14

What you see is that MATLAB returns the 4 intersections, not just the first row with the first column index, and same for second. Instead we get each row index combined with each column index. To get only the two elements we originally sought, we can convert the subscripts into another form of index, a linear index. With linear indexing, we think of an array as starting with the (1,1) element, and, going down each row, stringing out the elements into a long column. Then we use the count into that column vector, which has length equal to numRows * numCols. Use sub2ind to make the transformation.

linInds = sub2ind(size(A),[2 4],[3,2])
linInds =
    10     8

and then use these to index back into A,

A(linInds)
ans =
     7    14

getting exactly the values expected.

At least I hope they are the values you expected, instead of 10 and 8. Remember, MATLAB stores array values columnwise, i.e., column followed by column.

Combining Styles of Indexing

Now consider another situation, where you have a matrix and you only want certain columns extracted, depending on the value in the corresponding row. Perhaps, for example, you are using NaN as a placeholder and you only want columns 2 and 4 in rows that do not start with NaN.

One function that will be useful to us here is isnan, and the output of isnan is a logical array (think true and false). If I have a vector D like this,

D = [ 1 NaN 3 -4 NaN pi]'
D =
            1
          NaN
            3
           -4
          NaN
       3.1416

I can identify which entries have NaN values.

nanD = isnan(D);
whos nanD
nanD
  Name      Size            Bytes  Class      Attributes

  nanD      6x1                 6  logical              

nanD =
     0
     1
     0
     0
     1
     0

I can us logical indexing to extract the non-|NaN| entries in D.

nonNaND = D(~nanD)
nonNaND =
            1
            3
           -4
       3.1416

Combine Techniques

I am now ready to combine the ideas of logical indices and regular subscripts. Here's some data.

Data = rand(10,5)
Data =
      0.81472      0.15761      0.65574      0.70605      0.43874
      0.90579      0.97059     0.035712     0.031833      0.38156
      0.12699      0.95717      0.84913      0.27692      0.76552
      0.91338      0.48538      0.93399     0.046171       0.7952
      0.63236      0.80028      0.67874     0.097132      0.18687
      0.09754      0.14189      0.75774      0.82346      0.48976
       0.2785      0.42176      0.74313      0.69483      0.44559
      0.54688      0.91574      0.39223       0.3171      0.64631
      0.95751      0.79221      0.65548      0.95022      0.70936
      0.96489      0.95949      0.17119     0.034446      0.75469

Let me pull out columns 2 and 5, if the first column is at least 0.8. I'll leave the first column as well so you can trace what's happening.

extractD = Data(Data(:,1)>=0.8,[1 2 5])
extractD =
      0.81472      0.15761      0.43874
      0.90579      0.97059      0.38156
      0.91338      0.48538       0.7952
      0.95751      0.79221      0.70936
      0.96489      0.95949      0.75469

Have You Combined Indexing Techniques?

If you have, what were you trying to accomplish when you mixed different kinds of indexing into a single array.




Published with MATLAB® 7.11


  • print

Comments

To leave a comment, please click here to sign in to your MathWorks Account or create a new one.