Skip to main content
Version: Next

Python Script

Runs Python against the message. It is the general-purpose node — where you compute, reshape, decide and filter when no purpose-built node fits.

Type: python · Category: Transform · Ports: one input · one output

The Python Script node as it appears on the pipeline canvas.The Python Script node as it appears on the pipeline canvas.

When to use it​

  • Computing something from several inputs — a rate, an average, an OEE figure
  • Reshaping a payload before a database write or an API call
  • Filtering: return None and the branch ends
Reach for something else when

For a per-item transformation over an array, Batch Transform runs the items in parallel. For a database write, use the SQL node rather than system.db — only the node gets store-and-forward.

Settings​

SettingKeyDefaultPurpose
Codecode—The transform function.
Timeouttimeout60 sFail the node after this long.

Example — computing OEE from a tag read​

# Supported modules: system, math, json, datetime
# Globals: payload (data), message (metadata)
import datetime


def transform(payload, message):
    """
    :param payload: the primary data object from the previous node
    :param message: the full message, including metadata and context
    :return: the value passed to the next node
    """
    speed = system.tag.read("Line1/Filler/Motor1/Speed")
    if speed["quality"] != "Good":
        return None                      # returning None ends this branch

    target = payload[0]["targetRate"]    # from the SQL node upstream
    run_minutes = payload[0]["runMinutes"]

    availability = run_minutes / 60.0
    performance = speed["value"] / target if target else 0.0

    return {
        "line": "LINE1",
        "availability": round(availability, 4),
        "performance": round(performance, 4),
        "oee": round(availability * performance, 4),
        "timestamp": datetime.datetime.now(datetime.timezone.utc).isoformat(),
    }

Gotchas​

  • Returning None ends the branch. This is the cleanest filter there is — but a downstream Join waiting on that branch will wait forever, so filter before a split, not between split and join.
  • Set the timeout shorter than whatever is waiting. A node still running when the caller has given up is pure cost.

See also​