Rockset

Rockset is a real-time analytics database for fast queries on fresh data from databases, lakes and streams.

Authentication Methods

Rockset connections support API Key authentication.

Supported Data Types

Rockset is a document store and does not impose static typing on columns.

Because Salesforce fields expect consistent data types to be used, it is your responsibility to ensure that this occurs.

If you aren't careful with this, you may see runtime errors like:

The "amount__c" field is of type number, but the value from the external system is "hello"

The Omnata Setup process will make every effort to choose an appropriate data type, but in some situations it relies on a sample of values that don't cover the full range of possibilities.

If you prefer to guarantee readability and sacrifice some filtering capability, you can manually change the Salesforce field type to Text.

The following table describes the current mapping between Rockset data types and Salesforce.

Rockset	Corresponding Salesforce type
float	NUMBER_TYPE if the precision is < 18, otherwise STRING_SHORT_TYPE
int	NUMBER_TYPE
bool	BOOLEAN_TYPE (rendered as Checkbox) Note that Salesforce does not support null values here, so you must ensure that none exist.
date	Dates are not supported by Salesforce Connect, but the adapter converts to DATETIME_TYPE (at midnight) for convenience. This makes them usable but potentially misleading.
datetime	DATETIME_TYPE
timestamp	DATETIME_TYPE
time	STRING_SHORT_TYPE
string	STRING_SHORT_TYPE if the max length is <255, otherwise STRING_LONG_TYPE

Improving View Performance

Rockset does not provide column metadata for views via the DESCRIBE statement, which we need to fetch whenever the platform cache is cleared.

For performance reasons, we recommend creating a collection named <view name>_sample containing a sample of the records in your view. If you don't do this, we'll sample the view directly to determine the column data types which may be slow on larger datasets.

To do this, first query your view with a limit of 20 records and download the result as a JSON file.

Note that these records will be used to infer the column schema of your view, so if you have columns that contain a lot of null values, you may wish to take a larger sample or filter the view. There is no performance impact for using a larger sample, as we'll only be using DESCRIBE on this sample collection.

Then create a Collection, upload the JSON file, and click "Next" through to step 3.

Ensure your new collection is in the same workspace as your view, and give it the same name with the _sample suffix:

Once this is done, refresh the External Objects in the Omnata UI and the warning should disappear.