• The row from which to extract a value is determined by the order in which the rows are organized at the time that the function is executed. 

• If you are working on a randomly generated sample of your dataset, the values that you see for this function might not correspond to the values that are generated on the full dataset during job execution.
  

• In addition to the column to which to apply the function, the function accepts two other parameters: 
  • The first integer parameter defines the number of rows before the row being tested to scan for a non-empty value.
  • The second integer parameter defines the number of rows after the row being tested to scan for a non-empty value. 
  • If no non-empty value is found within these rows, the empty value remains empty.
  • The default values are -1 and 0 respectively, which performs an unlimited search before the row for a non-empty value to use to fill.
• You can use the group and order parameters to define the groups of records and the order of those records to which this function is applied.
• This function works with the following transforms:
  • Window Transform
  • Set Transform
  • Derive Transform

Output: Returns all values from the myNumber column with any null cells filled by the most recent non-blank value.

Output: Returns all values from the myNumber column with any null cells filled by the most recent non-empty value within the window 5 rows before the current row and 4 rows after it.

For more information on the order and group parameters, see Window Transform.

col_ref

Name of the column whose values are filled when null. 

• Multiple columns and wildcards are not supported.

int_rows_before

Number of rows before the current row to scan for the most recent non-empty value.

• Default value is -1, which scans all preceding rows.
• 0 does not scan before the current row.

int_rows_after

Number of rows after the current row to scan for the most recent non-empty value.

• Default value is 0, which does not scan rows after the current row. 
• -1 scans all following rows.

Example - Fill with quarterly forecast values

Your monthly sales data includes amount sold for each month. However, the source system only provides the quarterly forecast as part of the data for the first month of each quarter. You can use the FILL function to add the prior forecast to each month's data. 

Source:

Transformation:

You can use the following transformation to fill the prior forecast value for each blank month in the Forecast_Qtr column. Note that the order parameter must be set to Date to establish the proper sorting:

You can now delete the Forecast_Qtr column and rename the generated window column to the deleted name.

To see how you are progressing each month, you might use the following transformation, which computes the average forecast for each month:

You can then compare this value to the actual Amount value for each month:

Results: