Function

Custom JavaScript or Python transformation function that returns an updated row with optional lineage and audit metadata.

Parameters

• language (Language, string, optional) — JavaScript or Python. Omitted language is JavaScript. Switching does not translate source. Allowed values: javascript, python.
• function (Function Code, string, required) — JavaScript or Python function code. See JSON Configuration for signature and examples. Default: "".

Field selection

• source — Field whose values this operation reads. Use a single field name, a comma-separated list of plain field paths (cash_flow_type, currency) from the multi-select UI, or a regex via Pattern mode. Omit to apply to all fields on the row.
• target — Optional output field. When omitted, the source field is updated in place. Hidden when more than one named source field is selected.

JSON Configuration

{
  "id": "op-function",
  "type": "transformation",
  "action": "function",
  "params": {
    "language": "javascript",
    "function": "({ row }) => ({\n  row: {\n    ...row,\n    total: row.price * row.quantity,\n  },\n  lineage: [\n    {\n      outputField: \"total\",\n      inputFields: [\"price\", \"quantity\"],\n      usage: \"transformation\",\n    },\n  ],\n})"
  },
  "testInputs": [
    "Hello World",
    "MixedCase",
    "sample text",
    "already",
    "123",
    ""
  ]
}

Write Python with params.language set to python (JS/Python control on the Function editor). Use def transform(ctx): and return {"row": …}. Import the Python standard library as needed, from fontana import Currency for money, and import pandas when the function uses pandas. On Operation nodes a Python function returns exactly one row per input row.

Python on Compute nodes

On a Compute node the function still runs once per input row, and ctx["context"] is a dict that persists across every row of that operation. It starts empty for each operation and each run. Return one {"row": …}, a list of {"row": …} for several output rows, or [] for none. To aggregate, collect rows in context, return [] until ctx["rowIndex"] == ctx["numRows"] - 1, then return the summary rows.

def transform(ctx):
    acc = ctx["context"]
    acc.setdefault("rows", []).append(dict(ctx["row"]))
    if ctx["rowIndex"] < ctx["numRows"] - 1:
        return []
    total = sum(r["amount"] or 0 for r in acc["rows"])
    return [{"row": {"rows": len(acc["rows"]), "total": total}}]

Each returned row traces back to the input row whose call returned it, so a summary returned on the last row traces to that row. A row whose function raises an error produces no output and is reported against that input row. One Python interpreter can hold at most 4 GiB, so keep only the fields you need in context; running out of memory stops the node with an error.