The window operator produces a result array where each output cell is some aggregate function calculated over the cells contained in a N-dimensional rectangle around the corresponding cell in the source array.

Synopsis

window(array, dim1_preceding, dim1_following, [dim2_low,dim2_high[, ...]], aggregate1(attr_name1)[,aggrgegate2(attr_name2)[, ...]])

Summary

The window operator computes one or more aggregate functions of specified attributes over a fixed-size moving window. The result array has the same size and number of cells as the source array.

Inputs

The window operator takes the following arguments:

Example

To calculate a running average for a 3x3 window on a 4x3 array, do the following:

1. Create an array M with constant values on each diagonal:

   AFL% create array M <val:double>[row=1:4; col=1:3];
   AFL% store(build(M, row+col), M);

   The output is:

   {row,col} val
   {1,1} 2
   {1,2} 3
   {1,3} 4
   {2,1} 3
   {2,2} 4
   {2,3} 5
   {3,1} 4
   {3,2} 5
   {3,3} 6
   {4,1} 5
   {4,2} 6
   {4,3} 7

2. Return the average (mean) values over a  3×3 rectangle moving over M.  The center of the rectangle is the left corner of the input.
   

   AFL% window(M, 0, 2, 0, 2, avg(val)); 

   
   The output is:

   {row,col} val_avg
   {1,1} 4
   {1,2} 4.5
   {1,3} 5
   {2,1} 5
   {2,2} 5.5
   {2,3} 6
   {3,1} 5.5
   {3,2} 6
   {3,3} 6.5
   {4,1} 6
   {4,2} 6.5
   {4,3} 7

   The AFL window operator provides the same functionality as the AQL SELECT ... FROM ... WINDOW statement.

When the Window goes Beyond the End of the Array

To see what happens when the window goes beyond the end of the array, do the following:

1. Consider the following array, A:

   [(),(2),(3),(4),(5)] 

2. Use a window that goes outside the boundary of the array, and join it to A, to construct a new array that has a lag:

   AFL% join(A,window(A,1,0,first_value(x) as lag)); 

   
   The output is:

   i,x,lag
   2,2,2
   3,3,2
   4,4,3
   5,5,4