Excerpt |
---|
This example illustrates how to extract values from a column. |
Source:
The following dataset contains counts of support emails processed by each member of the support team for individual customers over a six-month period. In this case, you are interested in the total number of emails processed for each customer.
Unfortunately, the data is ragged, as there are no entries for a support team member if he or she has not answered an email for a customer.
custId | startDate | endDate | supportEmailCount |
---|---|---|---|
C001 | 7/15/2015 | 12/31/2015 | ["Max":"2","Ted":"0","Sally":"12","Jack":"6","Sue":"4"] |
C002 | 7/15/2015 | 12/31/2015 | ["Sally":"4","Sue":"3"] |
C003 | 7/15/2015 | 12/31/2015 | ["Ted":"12","Sally":"2"] |
C004 | 7/15/2015 | 12/31/2015 | ["Jack":"7","Sue":"4","Ted":"5"] |
If the data is imported from a CSV file, you might need to make some simple Replace Text or Pattern transformations to clean up the data to look like the above example.
Transformation:
Use the following transformation to extract just the numeric values from the supportEmailCount
array:
D trans RawWrangle true Type step WrangleText extractlist col:supportEmailCount on:`{digit}+` p01Name Column p01Value supportEmailCount p02Name Pattern matching elements in list p02Value `{digit}+` SearchTerm Extract matches into Array
You should now have a column extractlist_supportEmailCount
containing a ragged array. You can use the following transformations to convert this data to a comma-separated list of values:
D trans | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
D trans | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
D trans | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Convert the column to String data type.
You can now split out the column into separate columns containing individual values in the modified source. The limit
parameter specifies the number of splits to create, resulting in 5 new columns, which is the maximum number of entries in the source arrays.
D trans | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
You might have to set the type for each generated column to Integer. If you try to use a New Formula transformation to calculate the sum of all of the generated columns, it only returns values for the first row because the missing rows are null values.
In the columns containing null values, select the missing value bar in the data histogram. Select the Replace suggestion card, and modify the transformation to write a 0
in place of the null value, as follows:
D trans RawWrangle true p03Value ismissing([extractlist_supportEmailCount3]) Type step WrangleText set col: extractlist_supportEmailCount3 value: '0' row: ismissing([extractlist_supportEmailCount3]) p01Name Columns p01Value extractlist_supportEmailCount3 p02Name Formula p02Value '0' p03Name Group rows by SearchTerm Edit column with formula
Repeat this step for any other column containing null values.
You can now use the following to sum the values in the generated columns:
D trans | ||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Results:
After renaming the generated column to totalSupportEmails
and dropping the columns used to create it, your dataset should look like the following:
custId | startDate | endDate | supportEmailCount | totalSupportEmails |
---|---|---|---|---|
C001 | 7/15/2015 | 12/31/2015 | ["Max":"2","Ted":"0","Sally":"12","Jack":"6","Sue":"4"] | 24 |
C002 | 7/15/2015 | 12/31/2015 | ["Sally":"4","Sue":"3"] | 7 |
C003 | 7/15/2015 | 12/31/2015 | ["Ted":"12","Sally":"2"] | 14 |
C004 | 7/15/2015 | 12/31/2015 | ["Jack":"7","Sue":"4","Ted":"5"] | 16 |
D s also | ||
---|---|---|
|