The row from which to extract a value is determined by the order in which the rows are organized at the time that the transform 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.
- If the previous value is missing or null, this function generates a missing value.
- You can use the
orderparameters to define the groups of records and the order of those records to which this transform is applied.
- This function works with the following transforms:
window value:PREV(myNumber, 1) order:Date
Output: Generates the new column, which contains the value in the row in the
myNumber column immediately preceding the current row, when ordered by
window value:PREV(col_ref, k_integer) order: order_col [group: group_col]
|col_ref||Y||string||Name of column whose values are applied to the function|
|k_integer||Y||integer (positive)||Number of rows before the current one from which to extract the value|
For more information on the
group parameters, see Window Transform.
For more information on syntax standards, see Language Documentation Syntax Notes.
Name of the column whose values are used to extract the value that is
k-integer values before the current one.
- Multiple columns and wildcards are not supported.
|Required?||Data Type||Example Value|
|Yes||String (column reference)|
Integer representing the number of rows before the current one from which to extract the value.
- Value must be a positive integer. For negative values, see NEXT Function.
k=1represents the immediately preceding row value.
- If k is greater than or equal to the number of values in the column, all values in the generated column are missing. If a
groupparameter is applied, then this parameter should be no more than the maximum number of rows in the groups.
- If the range provided to the function exceeds the limits of the dataset, then the function generates a null value.
- If the range of the function is valid but includes missing values, the function generates a missing, non-null value.
|Required?||Data Type||Example Value|
Example - Examine prior order history
The following dataset contains orders for multiple customers over a period of a few days, listed in no particular order. You want to assess how order size has changed for each customer over time and to provide offers to your customers based on changes in order volume.
When the data is loaded into the Transformer page, you can use the
PREV function to gather the order values for the previous two orders into a new column. The trick is to order the
window transform by the date and group it by customer:
window value: PREV(OrderValue, 1) order: Date group: CustId
window value: PREV(OrderValue, 2) order: Date group: CustId
rename col: window to: 'OrderValue_1'
rename col: window1 to: 'OrderValue_2'
You should now have the following columns in your dataset:
The two new columns represent the previous order and the order before that, respectively. Now, each row contains the current order (
OrderValue) as well as the previous orders. Now, you want to take the following customer actions:
- If the current order is more than 20% greater than the sum of the two previous orders, send a rebate.
- If the current order is less than 90% of the sum of the two previous orders, send a coupon.
- Otherwise, send a holiday card.
To address the first one, you might add the following, which uses the
IF function to test the value of the current order compared to the previous ones:
derive type:single value: IF(OrderValue >= (1.2 * (OrderValue_1 + OrderValue_2)), 'send rebate', 'no action') as: 'CustomerAction'
You can now see which customers are due a rebate. Now, edit the above and replace it with the following, which addresses the second condition. If neither condition is valid, then the result is
send holiday card.
derive type:single value: IF(OrderValue >= (1.2 * (OrderValue_1 + OrderValue_2)), 'send rebate', IF(OrderValue <= (1.2 * (OrderValue_1 + OrderValue_2)), 'send coupon', 'send holiday card')) as: 'CustomerAction'
After you drop the
OrderValue_2 columns, your dataset should look like the following. Note that since the transforms with
PREV functions grouped by
CustId, the order of records has changed.