Transformations & Schema Mapping

Flows embed transform nodes between source nexsets and downstream destinations. Use NexsetCreate.transform (or a reusable transform_id) to describe how records should be reshaped, then attach the resulting nexset to sinks or additional transforms.

from nexla_sdk import NexlaClient
from nexla_sdk.models.nexsets.requests import NexsetCreate

client = NexlaClient(service_key="<SERVICE_KEY>")

transformed = client.nexsets.create(NexsetCreate(
    name="Orders Flattened",
    parent_data_set_id=456789,  # Nexset emitted by a source
    has_custom_transform=True,
    transform={
        "version": 2,
        "operations": [
            {"operation": "flatten", "path": "line_items", "alias": "item"},
            {
                "operation": "rename",
                "fields": {
                    "item.productId": "product_id",
                    "item.quantity": "qty"
                }
            },
            {
                "operation": "compute",
                "expression": "qty * item.unitPrice",
                "as": "line_total"
            },
            {"operation": "filter", "condition": "status != 'CANCELLED'"}
        ]
    }
))

Common operations supported by the public API include:

• flatten: explode nested arrays or objects into distinct rows.
• rename / map: align field names with downstream schemas.
• compute: create derived values using Nexla's expression engine.
• filter: drop records that do not satisfy business rules.
• typecast: coerce data to expected types before validation.

When a transform is reusable across flows, store it as a template and set transform_id instead of duplicating the JSON payload. Copying a flow with FlowCopyOptions(copy_entire_tree=True) replicates the source → nexset → destination path and preserves the transform configuration.