Tips and Techniques for Windchill Workflow

The following tips and techniques are intended as a collection of useful observations about workflow. It is particularly focused on a business object maturing through its lifecycle automated by a workflow.

General Guidelines:

1. Workflow Applicability

Workflows are typically applied in concert with lifecycles on a business object instance basis (i.e. document, part, change request, etc.). Keep in mind when designing workflows that processes (instances of workflow) apply to revisions. This means that:

• A process instance will execute once for all iterations of a given object revision
• Different revisions will go through different process instances

Workflows can also be instantiated independent of a business object, but this is an atypical case. Typically the motivation for having a workflow is that there is a conceptual data packet that is routed through a process and matures through some sense of state.

Within Windchill PDMLink, workflows are typically initiated upon object creation. Within Windchill ProjectLink, workflows are initiated for an object ad hoc through the use of routing functionality, and also in a coordinated fashion to automate a project plan.

In addition, in Windchill PDMLink, promote functionality launches Windchill ProjectLink route-like functionality, for multi-object state transition management.

Workflows vs. Lifecycles

Workflows automate a company's procedures by defining a set of user-based and system automated tasks, task linkages and routing rules, and an automated delivery user experience. The simplest visibility to process state is actually through the use of lifecycle state. Lifecycle state is directly visible on a business object and can be governed by the workflow process. Lifecycle give the following capabilities:

• Measure and display information maturity level
• Criteria for search
• Method of control access to data by state
• Can be associated to workflows
• A process monitor is available for graphical and detailed process metric information. This is very powerful and rich information. Typically, however, lifecycle state conveys a simplified adequate sense of where an object is in a process.

Lifecycle Managed Objects

Lifecycle managed objects are associated with a lifecycle template and one or more workflow templates on a per state basis. Lifecycle state is important for reporting maturity of data, controlling access to that data by maturity, and providing a state based filter for that data. Where state based behavior is not required, it is recommended to develop objects that inherit higher in the object model. Folder resident objects are an example of a simpler object class. Particularly where a large number of object instances are involved, inheriting from the simplest object class possible offers maximum performance.

Container Strategy

Workflow templates can be stored at three primary levels within Windchill - Site, Organization, and Container. It is recommended that workflow templates be stored at the highest level where reuse is possible. For example, if all organizations can take advantage of a workflow template it should be stored at the site level. Workflow template reuse is a very powerful notion and should be considered. If the workflow template is truly only relevant within a single library, then it should likely be managed within that library. Keeping the workflow template in the lowest level, like a library, if the template is truly not reusable, is a good practice in that it reduces potential clutter across the entire system.

2. Workflow Definition:

Activities

• When defining a new activity, identify the responsible role. The responsible role determines who will be notified if activities are overdue and/or process errors occur.
• Enter an activity description. The description will be displayed for reviewing of the process templates or can be shown in the work list. A URL can be included to reference more detailed documentation, or to provide ready access to stored form templates for attachment to the business object .
• An instruction tells you what to do with the task. Here you should give an explanation as to why a task has been given, when to click the "Task Complete" button, and what will happen after. The more information (html format is supported) you give the better.
• Make sure end-users have the right access - read, modify, and so forth - to the information. Applying these access policies at a domain level provides for easy downstream adjustment of those policies across all existing business objects, however the impacts of changing ACLs on executing workflow processes must be considered.
• In some cases it is necessary to provide ad hoc "activity based" access rights because the access controls are unique to that instance and its activity. In general, these should only be used in the most sophisticated workflow situations.

Workflow Variables

• It is recommended not to use "space" characters in workflow variable names. Windchill allows using "spaces" in workflow variables. However, these will not be usable in workflow robot java code.
• Java primitives are the mostly highly used and recommended workflow variables. They have the advantage of an out of the box user interface in workflow task forms.
• Windchill workflow's capability to track a variable is very powerful. Because workflow variables are serialized when persisted into the database, it is very important to pay attention to how that class is serialized. If the serialization signature changes between releases of your process, you risk having process instances that contain serialized variables that are not decodable. So, if you change the object's serialization signature, it is important to provide serialization support for all previous signatures.

Parallel Activities

• Use an "and" or an "or" connector when several parallel activities join into a single branch. (Figures 1 and 2)

Loop Links

• The first link in the workflow path which may be used more than once should be identified as "loop link." (Figure 3)

Lifecycle State Mapping

It is recommended in almost all workflow implementations that you associate a single parent workflow with a lifecycle to control all of an object's possible states. Here are some of the reasons behind this recommendation:

• It's easier to manage one workflow association with the lifecycle instead of a separate one for each phase.
• Variables sharing across all activities.
• Variables sharing across the different lifecycle's states.
• It is possible to define loops to any activity when using a single workflow. It is not possible to loop among activities in separate processes.
• Better performance results from less process overhead.
• It's easier to manage and monitor fewer running process instances.

The parent workflow need not be large or complicated. It can be broken down into blocks and sub-processes so that you can obtain the same reuse and complexity reduction benefits with a single parent workflow as you can with a workflow associated with each phase and gate.

Role Resolution

Roles for an activity are resolved at run-time so that the workflow can be re-used. Role resolution is just one mechanism for assigning participation to an activity.

Roles are resolved from a team assigned in that activity by name or variable, or from the object's team. The object's team is resolved using the team template, the lifecycle template and, for Windchill PDMLink and Windchill ProjectLink, the context team.

• The object type object initialization rules declare a lifecycle and team template for that object.
• On instantiation of the object, a team instance is created for that object and is initialized.
• Team roles are determined by the combination of the lifecycle template and team template roles. Note that this does not include the container roles.
• Role membership is determined by the combination of the container team and the team template.
• If the role does not exist in the team template, then the role membership is determined by the combination of the container team and the lifecycle template.
• Note that if the container team or the team template is modified after object creation, the object's team must be augmented (refreshed) if it is desired that those changes always remain current within the process.

Workflow Java Code

Note that activity Java code is limited to 2000 characters, so you will need to create classes for more complex processing.

Static Code

• Use static code calling helpers themselves calling server's methods.
• Java code in workflows should be minimized. So clean workflow coding would contain primarily calls to helper classes. This enables the reuse of workflow helper capability and also improves testing efficiency through modularization.

Exception Handling

Properly deal with exception handling, as incorrect exception handling gives:

• Hidden exceptions and no action taken
• Workflow that won't complete
• Unwanted behaviors

A good practice is to:

• Return a "result" attribute in order for the workflow process to be routed specifically .
• Return an "error Message" attribute in order for the next step of the workflow to have some information to give.

Tests

• If the workflow contains pieces of Java code, first test this code separately from the workflow using a small. stand-alone Windchill client application. Also execute "Validate Code" frequently to make sure you code is functional.

Synchronizations

• Avoid class-based synchronizations. Class-based synchronization typically creates excessive event firing. Typically when class-based synchronization is used, object-based synchronization would be even more effective

Workflow Testing

• Iterations: Perform a check-in on the workflow before testing it. If you forget to check-in the workflow template, you won't be able to modify it anymore until the already launched workflow instances are terminated.
• Flow Logic: Reviewer should check for reasonable flow logic making sure there are no un-terminated paths.
• All looping paths should contain a loop link indicated in red in Figure 4.
• Multiple paths should converge with an and/or connector. For instance Figures 5 should look like Figure 6.
• Multiple paths may converge on tasks if only one path can fire per loop. In Figure 7 three paths converge but only one can fire at a time.
• Latest workflow instance: Lifecycle should always point to the latest workflow iteration.
• Validate workflow: In the workflow editor execute the Validate All command.
• Syntax check: In the workflow editor execute Check Syntax for all expression robots, task tallying, Java code, etc.
• Data: If the workflow modifies the data it applies to, or even modifies the state, using csv load files to create the test data can be helpful. This allows for quick deleting and re-loading of the data for other tests.
• Principal Setting: If expression code is used to set a lifecycle or similar function make sure this is done as the object creator not as the default administrator. Figure 8 contains the code:

wt.org.WTPrincipal creator=(wt.org.WTPrincipal)wtcr.getCreator().getObject(); wt.session.SessionHelper.manager.setPrincipal( creator.getName() );

wt.lifecycle.LifeCycleHelper.service.reassign(wtcr,wt.lifecycle.LifeCycleHelper.service.getLifeCycleTemplateReference("My Lifecycle Template"));

• Set State checking: In the case of a version/iteration controlled object such as WTDocument there should be code for checking whether the object is shared and/or checked in prior to each set state robot. (Figures 9 and 10)
• Proxys: If workflow proxys are used and you are using loadfiles, make sure the proxies are imported first.