Introduction

Flow is a framework for writing tools, daemons and transfer plugins for Viz One, Vizrt’s Media Asset Management software. To communicate with Viz One, the Python SDK python-one is used. Please contact Vizrt for your copy and more information.

The idea behind Flow is to make it simple to integrate with Viz One on an enterprise level. The treats promoted by the framework are:

• Scalability (multi-threading and multi-server)
• Client statelessness
• Simplicity (one application does one thing)

To achieve this, all applications are limited to a certain structure:

Source -> Worker

The Source

Let’s start with the Source. The task of the Source is to gather information in one way or another, do as little with it as possible, and hand it over to the Worker. The Source runs in a single thread, which is the reason for keeping the processing within it as simple as possible. Input to Sources can be anything, but usually Stomp, Files or Standard Input.

Examples of built-in sources are:

• Asset Entry Listener flow.source.asset.AssetEntryListener Reacts on creations, updates and deletions of Asset Entries.
• Unmanaged Files Listener flow.source.location.UnmanagedFilesListener Reacts on creations, updates and deletions of Unmanaged Files.
• Generic Stomp Listener flow.source.stomp.GenericStompListener Subscribes to a manually given Stomp URL.
• Interval flow.source.interval.Interval Spawns a worker every nth second.
• STDIN flow.source.local.STDIN Reads from Standard Input and spawns a worker.

You can also quite easily create your own Source class.

The Worker

The Flow process will at start-up create a thread-pool that is used to run worker threads. Workers are essentially classes with a start method and a SOURCE class specified. Optionally you can also have a configure method that will be run once on start-up.

The start method will take an object as argument. This is the interface between the Source and the Worker. The Source will call the start method with one object as argument as a new thread for each event it produces. If there are no free workers, the event will be held until there are, and then called. There is also a time out for execution time, meaning the waiting time cannot be infinitely long.