Transformations
Detailed technical documentation on RudderStack Transformations.
Last updated
Was this helpful?
Detailed technical documentation on RudderStack Transformations.
Last updated
Was this helpful?
Transformations is one of RudderStack's key features. It gives you the ability to code custom JavaScript functions to implement specific use-cases on your event data, like:
Filtering/sampling events.
Enriching events by implementing static logic or leveraging an external API.
Cleaning/aggregating data.
Data masking or removing sensitive PII information in the events to ensure data privacy.
Implementing external actions on the events using an API.
Transformations can be used across your Event Stream, Reverse ETL and Cloud Extract pipelines.
You can refer to our Sample Transformations GitHub repository for some useful templates that you can use to create your own transformations.
Transformations only work for the Cloud mode destinations.
To know more about the Cloud mode in RudderStack, check out the RudderStack Connection Modes guide.
Log into your RudderStack dashboard.
Click on the Transformations link from the left panel in the dashboard, as shown:
Click on the Create New option as shown:
Next, assign a name for this new transformation. Enter the transformation function's code in the Transformation window, as shown:
Add your transformation code within the transformEvent
function in the Transformation window. You can also add other functions and call them from within transformEvent
.
Our Transformations GitHub repository contains some useful templates that you can use to create your own transformations.
You can copy-paste the entire code of any of the functions present in this repository into the Transformation window.
Remember to delete the pre-populated transformEvent
function in such cases, before pasting your code.
RudderStack also gives you the ability to test your transformation function with the Run Test option as shown:
In case you want to perform any aggregation/roll-up operation on a micro batch of events, use the transformBatch
function, as shown:
Disclaimer for using transformBatch
:
Make sure you pass on the messageId
from the input event onto the output event. If the messageId
is passed on, the order of delivery of events is preserved. Without it, this order may not be maintained.
It is highly recommended to use transformEvent
in every possible case as it ensures event ordering in all cases.
In case a transformBatch
is required and event ordering is important, make sure to pass on the messageId
.
One of the key features of RudderStack Transformations is the ability to reuse code written for a transformation in other transformations as well. For this, RudderStack lets you create libraries or functions that can be reused in different transformations.
To create a library, follow these steps:
Click on the Transformations link in the left nav bar and go to the Libraries section. Click on the Create New option to add new libraries, as shown:
Add the library's Name, an optional Description, and include the custom functions that you need to reuse across all other transformations.
You can add multiple functions to a single library, as shown:
RudderStack does not support the deletion of libraries as of now. This is to ensure that you don't break any existing transformations that use the library that you might be trying to delete.
To use the libraries in your existing transformation, simply take the name of your library and convert it into camel case without spaces - this becomes your library handle.
For example, if your library name is is rudder email
, then the library handle would be isRudderEmail
.
Let's say you want to import a function called rudderEmail
, which returns true
for the emails from a specific domain (e.g. RudderStack) and false
otherwise, from the is rudder email
library. Also, you want to use this function to filter the events that don't have the email address of the specified domain.
The following code snippet demonstrates how we can implement this use-case:
On running a test, an example event not having the specified email domain is filtered out, as shown:
When importing a single function or multiple functions from a library, it is important to keep the import statement to one line. Breaking the import function into multiple lines will lead to an error.
The following snippets highlight how to properly import functions from a library:
RudderStack injects a function metadata(event)
into your transformations as an argument. This allows you to access the event metadata variables that help you customize your transformations.
metadata()
takes the event as input and returns the metadata of the event.
The following properties, if available, are present in the metadata response:
sourceId
This refers to the ID of the source configured on your RudderStack dashboard. Note that it different from the source Write Key. Refer to the image below for more details.
destinationId
ID of the destination configured on your RudderStack dashboard.
messageId
Corresponds to the unique ID for each event.
sessionId
If sessions are enabled, this corresponds to the value of the session ID.
Since you may not need the metadata in every transformation, you can optionally access it by including it in your function signature as an argument wherever required.
An example of this is as shown below:
You can make any number of external API requests in your transformation functions and use the response to enrich your events data. RudderStack injects an asynchronous fetch(url)
function into your transformations. It makes an API call to the given URL and returns the response in a JSON format.
An example of how to use the fetch
function in transformations is shown below:
To see how this feature is used in more detail, check out the Clearbit enrichment example which uses the fetch
function.
For improved performance, we recommend using batch
API requests instead of a separate API request for each event whenever possible.
FetchV2 is a wrapper around the fetch
call to provide more access to fetch the response properties.
The following properties are present in a fetchV2
response:
status
Status code of fetch response, e.g. 200
.
url
URL of the Fetch API.
headers
Response headers
body
Response body in JSON or TEXT. By default, it is JSON.
Use the try-catch
block to capture the cause for failures during the fetch
call, e.g. when a timeout is applied.
An example of how to use the fetchV2
function in RudderStack transformations is shown below:
You can access event-related logs while running a test by including the log()
function in your transformation code. An example of this is as shown:
On running a test on the above code, you can see the logs in the Logs section of the dashboard, as shown:
Arguments to the log
function can be a String, Number or Object.
You should take into account the memory and time limits when adding a new transformation. Each invocation of the transformation should not exceed the following limits:
Memory Limit
8 MB
Time Limit
4 seconds
The user transformation fails if these limits are exceeded.
For queries on any of the sections covered in this guide, you can contact us or start a conversation in our Slack community.
Source ID