EXAMPLE - Splitting with Different Delimiter Types

This example shows how you can split data from a single column into multiple columns using the following types of delimiters:

  • single-pattern delimiter: One pattern is applied one or more times to the source column to define the delimiters for the output columns
  • multi-pattern delimiter: Multiple patterns, in the form of explicit strings, character index positions, or fixed-width fields, are used to split the column.

For more information on these methods, see Split Transform.


In this example, your CSV dataset contains status messages from servers in your environment. In this case, the data about the server and the timestamp is contained in a single value within the CSV.

Server|Date Time,Status
admin.examplecom|2016-03-05 07:04:00,down
webapp.examplecom|2016-03-05 07:04:00,ok
admin.examplecom|2016-03-05 07:04:30,rebooting
webapp.examplecom|2016-03-05 07:04:00,ok
admin.examplecom|2016-03-05 07:05:00,ok
webapp.examplecom|2016-03-05 07:05:00,ok


When the data is first loaded into the Transformer page, the CSV data is split using the following two transforms:

splitrows col: column1 on: '\r'

split col: column1 on: ',' quote: '\"'

You might need to add a header as the first step:


At this point, your data should look like the following:

admin.example.com|2016-03-05 07:04:00down
webapp.example.com|2016-03-05 07:04:00ok
admin.example.com|2016-03-05 07:04:30rebooting
webapp.example.com|2016-03-05 07:04:30ok
admin.example.com|2016-03-05 07:05:00ok
webapp.example.com|2016-03-05 07:05:00ok

The first column contains three distinct sets of data: the server name, the date, and the time. Note that the delimiters between these fields are different, so you should use a multi-pattern delimiter to break them apart:

split col:Server_Date_Time delimiters:'|',' '

When the above is added, you should see three separate columns with the individual fields of information. Note that the source column has been automatically dropped.

NOTE: A column name cannot contain the | value, so the source column name cannot be used as the basis for the column names applied to the generated columns. In this case, you must use the rename transform to update the generated columns accordingly.

Now, you decide that it would be useful to break apart the timestamp column into separate columns for year, month, and day. Since the column delimiter of this field is consistently a dash (-), you can use a single-pattern delimiter with the split transform:

split col:date on:`-` limit:2


After you rename the generated columns, your dataset should look like the following. Note that the source timestamp column has been automatically dropped.

Was this page helpful? Let us know how we did:

Send feedback about...

Google Cloud Dataprep Documentation