Serializable vs Non-Serializable

Serializable vs Non-Serializable Arguments

Microsoft Workflow Foundation (MWF), the technology on which OpenRPA is based, categorizes variables/arguments into two different groups: Serializable and Non-Serializable. For the majority of workflows, understanding the differences is not necessary, but there are cases where knowing the technical limitations of non-serializable objects might explain why your workflow or process is not working as expected.

In short: non-serializable objects, as the name suggests, cannot be serialized. This means these objects cannot be saved or transmitted over a network. Consequently, a workflow state cannot be saved if the workflow contains any non-serializable variable/argument, and invoking workflows with non-serializable arguments is not supported.

Here are some limitations users may face when dealing with workflows containing non-serializable variables/arguments and some suggested workarounds:

Saving workflow states

When OpenRPA is connected to an OpenCore, workflow states are saved automatically in OpenCore whenever specific activities¹ are reached. These states exist to indicate the current situation relevant to a workflow instance, such as the designer layout (activities, sequences) being run, the variables and arguments and their current values, etc. Hence, if the workflow contains non-serializable objects such as a DataTable, the state cannot be saved.

¹ - All the activities that “go idle” (For instance Detector, Delay, Persist and all the Invoke activities

Workaround: Split complex workflows into smaller workflows, leaving the smaller workflows to manage the non-serializable objects. That way if an unexpected interruption occurs, you will not lose all the data.

Remote OpenRPA / invoking with non-serializable arguments

Invoke OpenCore / Node-RED invoke and data return

If you wish to use the one of the Invoke Activity, be aware that the workflow being invoked may not use non-serializable arguments. As the destination computer / OpenRPA is different than the source OpenRPA, the non-serializable arguments cannot be passed around and thus the invoking will fail.

Similarly, when using the Invoke OpenCore activity to invoke a flow in Node-RED, non-serializable objects are not supported as arguments. An exception is the DataTable type; OpenRPA will attempt to convert these to JArray Objects before contacting Node-RED and convert it back to DataTable when the data is returned / node workflow out is used.

Workaround: Use the activities from OpenRPA.OpenFlowDB toolbox to upload files/update data/create entities and collections in OpenCore MongoDB. Then, on the destination computer, access the data by querying OpenCore MongoDB. This way, the entire Data/DataSet/DataTable (any non-serializable object) is stored within MongoDB and the parameter passed during the invoke can be the _id or some other identifier.

Workaround2: Convert the non-serializable object into a serializable one, like a base 64 string, then pass it as a parameter and convert it back if needed.