Skip to main content
Version: PromptQL

Recast Types

Overview​

At times, your database schema may not perfectly align with how you want to expose data types in your API. For example, you might have text fields that contain numeric values, JSON fields you want to filter as strings, or custom database types that would work better as standard types in your API.

Your semantic metadata layer allows you to recast these types without requiring any changes to your underlying datasource schema. This powerful feature enables you to unlock type-specific operations like numeric aggregations on text fields or string filtering on JSON data, all through simple metadata configuration changes.

Step 1. Update the representation mapping​

In the <connector-name>-types.hml file for your connector, you'll find definitions and mappings for your data source's types. For the type you wish to recast, you'll need to identify the correct DataConnectorScalarRepresentation and change the value for the representation key.

If you don't know which types are available when recasting, trigger the Hasura VS Code extension by pressing CTRL+SPACEBAR. Check out the example below.

Here we'll change a text type (represented as a string):
kind: DataConnectorScalarRepresentation
version: v1
definition:
  dataConnectorName: pg
  dataConnectorScalarType: text
  representation: string
  graphql:
    comparisonExpressionTypeName: string_comparison_exp
To an int:
kind: DataConnectorScalarRepresentation
version: v1
definition:
  dataConnectorName: pg
  dataConnectorScalarType: text
  representation: int32
  graphql:
    comparisonExpressionTypeName: text_as_int32_comparison_exp
Comparison Expression Type

We're also amending the comparisonExpressionTypeName as these must be unique.

Step 2. Regenerate your metadata​

Update your metadata by running the following:
ddn model add <connector-name> "*"

This will update your metadata to use the new type; you'll find changes in both the <connector-name>-types.hml file for various expressions (such as aggregates and boolean) and any models which use the recast type.

Step 3. Test it out​

Rebuild your metadata:
ddn supergraph build local
Then, restart your services:
ddn run docker-start

If you explore your metadata, you'll see the type recast, allowing you to perform operations that weren't previously possible.