Lifecycle of a Plugin

Prerequisites

This overview refers to Plugins, in reference to any new Plugin class developed with the SDK.

Overview

The lifecycle of a Plugin in Alteryx Designer can be broken down into 2 different types of runs:

  1. Workflow Run

  2. Update Only Run

A workflow run is the simpler of the 2, so let’s start with that.

Workflow Run

When you select the Run button in Designer, an instance of your Plugin class is created for each of your tools on the canvas.

The Plugin is constructed via its __init__ method, just like in normal Python. An instance of a Provider class is given to the plugin. Any resources required to read from and write to Designer can be obtained via methods/properties on the Provider.

In Alteryx Designer, input anchors can have one or many connections (controlled via the AllowMultiple flag in the tool XML definition). If you have any input connections, the next stage in the lifecycle consists of each of those input connections being opened, and data being pushed to each one. From the Plugin perspective, the on_input_connection_opened method is called for each connection that gets opened (the parameter it receives is the connection that was opened), and then (typically, before the next connection is opened), data packets are streamed to the plugin. Receiving record packets causes the on_record_packet method to be called. The parameter that on_record_packet receives is the connection that a new record packet is available on.

Once all record packets for all connections are received, the Plugin’s on_complete method is called in order to do any cleanup.

If your Plugin class does not have any input anchors, the lifecycle is the same, however, on_input_connection_opened, and on_record_packet never get called, since there are no connections/received records. This means that any records that you wish to output must be pushed during the on_complete method (you shouldn’t push them during __init__, for reasons described below).

In order to push records to your Plugin’s output anchors, they must first be opened with metadata. This Metadata object describes the record schema for that output anchor (in other words, field names, types, etc.).

See this sequence diagram for a visual representation of this lifecycle:

Workflow Run Sequence Diagram

Workflow Run Sequence Diagram

Update Only Run

Update only is a mode that runs in Designer any time…

  1. A new tool is added to the canvas.

  2. Any tool on the canvas has a change in configuration (typically, via the user interface or Configuration panel).

The purpose of this run mode is to generate the metadata that each tool will output during the next time a workflow runs. This allows new tools on the canvas to know what columns they can operate on. The metadata that comes out of a tool typically depends on…

  1. The incoming metadata.

  2. The configuration of the tool (via the user interface).

In this mode, a subset of the Plugin methods get called:

  1. __init__

  2. on_input_connection_opened

Since the goal of an update-only run is to build output metadata, the __init__ or on_input_connection_opened methods should run the open method on any output anchors the tool has. Additionally, since __init__ and on_input_connection_opened run in update-only mode, the developer should take care to perform the minimal amount of processing in these methods, since a fast update makes users happy.

See the sequence diagram for a visual representation of this lifecycle:

Update Only Sequence Diagram

Update Only Sequence Diagram