Skip to main content

Eval Node

The Eval node is a powerful transformation component in ZephFlow that applies expressions to modify, filter, or enrich incoming records. It uses the Fleak Eval Expression Language (FEEL) to transform data as it flows through your pipeline.

For more details of using FEEL, please refer to FEEL Reference

Eval node requires the transformation result to be a dict/map/object or an array. If the output is an array, each dict/object element in the array is treated as an output message. Primitive elements in the array will cause errors.

Primitive values such as strings, numbers and booleans will cause the evaluate to fail.

Key Features​

  • Transform data structures with concise expressions
  • Create new fields or modify existing ones
  • Apply conditional logic to record processing
  • Extract values from complex nested structures
  • Convert between data types
  • Manipulate strings, arrays, and dictionaries

Attaching Eval Node in a DAG​

jobContext:
  otherProperties: {}
  metricTags: {}

dag:
  - id: "eval_node"
    commandName: "eval"
    config:
      expression: "your FEEL"
    outputs:
      - "next_node"

Basic Usage​

Transform Example​

dict(
  user_id=$.user.id,
  total_amount=$.items[0].price * $.items[0].quantity,
  timestamp=ts_str_to_epoch($.created_at, "yyyy-MM-dd'T'HH:mm:ss")
)

This transforms each record by creating a new structure with the user's ID, calculated total amount, and a properly formatted timestamp.

Field Enrichment Example​

dict_merge(
  $,
  dict(
    full_name=$.first_name + " " + $.last_name,
    is_valid=$.age >= 18 and $.country == "US"
  )
)

This preserves the original record and adds two new fields: full_name and is_valid.

Config Object​

FieldTypeRequiredDefaultDescription
expressionStringYes—FEEL expression that transforms the input record

Examples​

Data Normalization​

dict(
  event_type="user_login",
  user=dict(
    id=$.userId,
    email=lower($.email),
    role=case(
      $.permissions == "admin" => "ADMIN",
      $.permissions == "editor" => "EDITOR",
      _ => "USER"
    )
  ),
  metadata=dict(
    source=$.src,
    timestamp=ts_str_to_epoch($.event_time, "yyyy-MM-dd HH:mm:ss.SSS")
  )
)

Record Filtering with Conditional Output​

case(
  $.status_code >= 400 => dict_merge(
    $,
    dict(
      error_type=case(
        $.status_code >= 500 => "SERVER_ERROR",
        _ => "CLIENT_ERROR"
      ),
      severity=case(
        $.status_code >= 500 => "HIGH",
        _ => "MEDIUM"
      )
    )
  ),
  _ => null
)

This transforms error records by adding error classification fields, and returns null for non-error records (which can be filtered out by a null filter node).

Processing Array Data​

arr_foreach(
  $.logs,
  log,
  dict(
    timestamp=ts_str_to_epoch(log.time, "yyyy-MM-dd'T'HH:mm:ss.SSSZ"),
    level=upper(log.level),
    service=$.service_name,
    message=log.msg,
    has_error=str_contains(log.msg, "error")
  )
)

This transforms an array of log entries into a structured format with additional derived fields.

  • filter: Filter records based on specific conditions using FEEL
  • assertion: Validate that records meet specific conditions using FEEL