Schema Model Generator

Skill: Schema & Model Generator

Given a Go SDK contract (DTOs), generate the corresponding Terraform schema attributes and model structs with high fidelity.

Prerequisites

• SDK DTO struct fields have been identified (from #skill:sdk-contract-navigator)
• You know the item archetype (basic, definition, properties, etc.)
• If the issue contains a 🌳 DTO Nesting Depth Map, use it to determine:
• One model struct per tree node that introduces []Type or *Type
• Each such struct needs its own set() method
• Which fields use supertypes.ListNestedObjectValueOf (nested objects) vs supertypes.ListValueOf (scalars) vs supertypes.SingleNestedObjectValueOf (optional nested)

Step 1 — Classify Each SDK Field

For each field in the SDK DTO struct, determine its category using the "Attribute Behaviors" table in .github/instructions/schema-model-patterns.instructions.md.

Classification criteria:

| Category | Criterion |

| ------------------------- | -------------------------------------------------- |

| Read-only | Only appears in Get response Properties struct |

| Create-time-only | Appears in CreationPayload and cannot be updated |

| Updatable required | User must provide, can be changed after create |

| Updatable optional | User may provide, can be changed after create |

| Optional with default | Server sets a default if not provided |

Step 2 — Generate Model Struct Fields

Map each SDK field type to the corresponding Terraform model type using the "SDK Type → Model Type Mapping" table in .github/instructions/schema-model-patterns.instructions.md.

Every field must have a tfsdk:"<snake_case_name>" tag.

Naming Conventions

• Model struct: <item><purpose>Model — e.g. lakehousePropertiesModel, lakehouseConfigurationModel
• Nested model: <item><nested>Model — e.g. lakehouseSQLEndpointPropertiesModel
• Field names: PascalCase in Go, with tfsdk:"snake_case" tag
• SDK PascalCase → Terraform snake_case: OneLakeFilesPath → onelake_files_path

Example Properties Model

type lakehousePropertiesModel struct {
    OneLakeFilesPath      types.String                                                              `tfsdk:"onelake_files_path"`
    OneLakeTablesPath     types.String                                                              `tfsdk:"onelake_tables_path"`
    SQLEndpointProperties supertypes.SingleNestedObjectValueOf[lakehouseSQLEndpointPropertiesModel] `tfsdk:"sql_endpoint_properties"`
    DefaultSchema         types.String                                                              `tfsdk:"default_schema"`
}

Example Configuration Model (CreationPayload)

type lakehouseConfigurationModel struct {
    EnableSchemas types.Bool `tfsdk:"enable_schemas"`
}

Example Nested Model

type lakehouseSQLEndpointPropertiesModel struct {
    ID                 customtypes.UUID `tfsdk:"id"`
    ConnectionString   types.String     `tfsdk:"connection_string"`
    ProvisioningStatus types.String     `tfsdk:"provisioning_status"`
}

Step 3 — Generate Model Methods

Generate both directions of mapping: response set() (SDK → TF) and request builders (TF → SDK).

3a. Response set() — SDK → TF (both Fabric Items and non-items)

Every model struct needs a set() method that maps SDK response DTO → TF model.

Top-level set() (with nested objects):

Signature includes context.Context and returns diag.Diagnostics:

func (to *<item>PropertiesModel) set(ctx context.Context, from fab<package>.<DTO>) diag.Diagnostics {
    to.SimpleField = types.StringPointerValue(from.SimpleField)
    // ... other simple fields

    // Handle nested struct
    nestedValue := supertypes.NewSingleNestedObjectValueOfNull[<nestedModel>](ctx)
    if from.NestedField != nil {
        nestedModel := &<nestedModel>{}
        nestedModel.set(*from.NestedField)  // or with ctx if nested has its own nested
        if diags := nestedValue.Set(ctx, nestedModel); diags.HasError() {
            return diags
        }
    }
    to.NestedField = nestedValue

    return nil
}

Leaf set() (no nested objects):

Simpler signature without context.Context or diag.Diagnostics:

func (to *<nestedModel>) set(from fab<package>.<NestedDTO>) {
    to.ID = customtypes.NewUUIDPointerValue(from.ID)
    to.StringField = types.StringPointerValue(from.StringField)
    to.EnumField = types.StringPointerValue((*string)(from.EnumField))
}

Setter patterns by type: Use the "Setter Pattern" column in the "SDK Type → Model Type Mapping" table in schema-model-patterns.instructions.md.

3b. Request Builders — TF → SDK (Create/Update)

Both Fabric Items and non-items need TF→SDK mapping for writable fields. The pattern differs by category:

• Fabric Items: Inline in creationPayloadSetter closure (simple — typically 1-3 fields from configuration model). See fabric-item-patterns.instructions.md § "Closure Examples".
• Non-items: Dedicated request builder structs with set() method that builds the SDK request directly (complex — full request DTOs)

Non-item request builder struct — embeds the SDK request type, set() populates it:

type requestCreate<Type> struct {
    fabcore.Create<Type>Request // embedded SDK request type
}

func (to *requestCreate<Type>) set(ctx context.Context, from <type>ResourceModel) diag.Diagnostics {
    to.DisplayName = from.DisplayName.ValueStringPointer()
    to.Description = from.Description.ValueStringPointer()
    // ... map each writable field into the embedded struct
    return nil
}

Usage: r.client.Create<Type>(ctx, reqCreate.Create<Type>Request, nil)

Inverse mapping rules: Use the inverse of the "SDK Type → Model Type Mapping" table in schema-model-patterns.instructions.md. For each TF type, call its Value*Pointer() method (e.g., types.String → .ValueStringPointer(), types.Bool → .ValueBoolPointer()).

Non-obvious cases:

| TF Model Type | SDK Type | Pattern |

| ----------------------------------------- | ------------- | ------------------------------------------------------------------ |

| types.Int64 | *int32 | ptr.To(int32(from.Field.ValueInt64())) — type narrowing required |

| supertypes.SingleNestedObjectValueOf[M] | *NestedDTO | .Get(ctx) → construct nested DTO from sub-model |

| supertypes.ListNestedObjectValueOf[M] | []NestedDTO | .Get(ctx) → iterate slice, build each DTO |

Reference: internal/services/connection/models_resource_connection.go

Step 4 — Generate Schema Attributes

For Fabric Item resources, schema attributes go in separate functions:

Resource Properties Schema

// schema_resource_<item>.go
func getResource<Item>PropertiesAttributes(ctx context.Context) map[string]schema.Attribute {
    return map[string]schema.Attribute{
        "<field_name>": schema.StringAttribute{
            MarkdownDescription: "<description>.",
            Computed:            true,
        },
        // ... more attributes
    }
}

Resource Configuration Schema (for CreationPayload)

func getResource<Item>ConfigurationAttributes() map[string]schema.Attribute {
    return map[string]schema.Attribute{
        "<field_name>": schema.BoolAttribute{
            MarkdownDescription: "<description>.",
            Required:            true,
            PlanModifiers: []planmodifier.Bool{
                boolplanmodifier.RequiresReplace(),
            },
        },
    }
}

Schema Attribute Type Mapping

Use the "SDK Type → Schema Mapping" table in .github/instructions/schema-model-patterns.instructions.md.

Rules for All Schema Attributes

For attribute behavior flags, plan modifiers, and validators, refer to the "Attribute Behaviors", "Plan Modifiers", and "Validators" sections in .github/instructions/schema-model-patterns.instructions.md.

Additional rules:

• Always use MarkdownDescription (never Description)
• Nested objects: Must include CustomType: supertypes.NewSingleNestedObjectTypeOf<model>
• UUID fields: Must include CustomType: customtypes.UUIDType{}

Example Nested Attribute

"sql_endpoint_properties": schema.SingleNestedAttribute{
    MarkdownDescription: "An object containing the properties of the SQL endpoint.",
    Computed:            true,
    CustomType:          supertypes.NewSingleNestedObjectTypeOf[lakehouseSQLEndpointPropertiesModel](ctx),
    Attributes: map[string]schema.Attribute{
        "provisioning_status": schema.StringAttribute{
            MarkdownDescription: "The SQL endpoint provisioning status.",
            Computed:            true,
        },
        "connection_string": schema.StringAttribute{
            MarkdownDescription: "SQL endpoint connection string.",
            Computed:            true,
        },
        "id": schema.StringAttribute{
            MarkdownDescription: "SQL endpoint ID.",
            Computed:            true,
            CustomType:          customtypes.UUIDType{},
        },
    },
},

Canonical References

• Model struct patterns: internal/services/lakehouse/models.go
• Resource schema patterns: internal/services/lakehouse/schema_resource_lakehouse.go
• Data source schema patterns: internal/services/lakehouse/schema_data_lakehouse.go
• Non-item schema (superschema): internal/services/connection/schema.go
• Request builder patterns: internal/services/connection/models_resource_connection.go