In the previous release, RAQL treated data from datasets that did not include a schema as untyped. In 3.7, the RAQL Engine now derives a schema based on the first data row in the dataset.

In most cases where a schema is not available, the datatype for columns is considered to be string. The data in the first row also defines the entire set of expected columns and the structure for all subsequent rows.

This derived schema can cause errors when the dataset is queried if the first row of the dataset is not a good representation of the dataset.

To prevent errors, the best practice is to provide schemas when datasets are loaded whenever possible. See Providing Dataset Path and Datatype Information in a Schema for more information on how to add schema information.