Workflow patterns

##Evaluation of HyperFlow with respect to workflow pattern support (http://www.workflowpatterns.com/)##

###Data patterns###

1. Task Data
2. Block Data
3. Scope Data
4. Multiple Instance Data
5. Case Data
6. Folder Data
7. Workflow Data
8. Environment Data
9. Data Interaction - Task to Task
10. Data Interaction - Block Task to Sub-Workflow Decomposition
11. Data Interaction - Sub-Workflow Decomposition to Block Task
12. Data Interaction - to Multiple Instance Task
13. Data Interaction - from Multiple Instance Task
14. Data Interaction - Case to Case
15. Data Interaction - Task to Environment - Push-Oriented
16. Data Interaction - Environment to Task - Pull-Oriented
17. Data Interaction - Environment to Task - Push-Oriented
18. Data Interaction - Task to Environment - Pull-Oriented
19. Data Interaction - Case to Environment - Push-Oriented
20. Data Interaction - Environment to Case - Pull-Oriented
21. Data Interaction - Environment to Case - Push-Oriented
22. Data Interaction - Case to Environment - Pull-Oriented
23. Data Interaction - Workflow to Environment - Push-Oriented
24. Data Interaction - Environment to Workflow - Pull-Oriented
25. Data Interaction - Environment to Workflow - Push-Oriented
26. Data Interaction - Workflow to Environment - Pull-Oriented
27. Data Transfer by Value - Incoming. Yes: by receiving a signal with a data value.
28. Data Transfer by Value - Outgoing. Yes: by sending a signal with a data value.
29. Data Transfer - Copy In/Copy Out
30. Data Transfer by Reference - Unlocked. Yes: data references can be passed in signals instead of actual values.
31. Data Transfer by Reference - With Lock. Yes, indirectly. Support of a mutually accessible data store is considered a separate concern, out of scope of the workflow engine. However, functions can use an external store with a lock mechanism in which workflow processes can share data.
32. Data Transformation - Input. Yes: data can be transformed in the process function, before the actual processing.
33. Data Transformation - Output. Yes: data can be transformed in the process function, after the actual processing and right before invoking the callback.
34. Task Precondition - Data Existence
35. Task Precondition - Data Value
36. Task Postcondition - Data Existence
37. Task Postcondition - Data Value
38. Event-based Task Trigger
39. Data-based Task Trigger
40. Data-based Routing. Yes: this pattern can be implemented with a choice process whose function evaluates a condition based e.g. on input signals and sets the condition flag for output signals accordingly. Example: Streaming Map/Reduce workflow.

###Control patterns###

1. Sequence. Yes: a signal exchanged between two processes automatically establishes a sequence.

2. Parallel Split. Yes: several output ports of a process connected to multiple other processes automatically split the execution into parallel branches.

3. Synchronization. Yes: a dataflow process synchronizes all branches of execution associated with its input ports.

4. Exclusive Choice. Yes: this pattern is enabled by a choice process whose function evaluates the condition and sets the condition flag for one of the output signals.

5. Simple Merge. Yes: both simple and multi-merge can be implemented by having multiple processes produce the same signal to another process (see Multi-Merge pattern below for illustration).

6. Multi-Choice. Yes: the choice process type enables this pattern.

7. Structured Synchronizing Merge

8. Multi-Merge. Yes: both simple and multi-merge can be implemented by having multiple processes produce the same signal to another process, as illustrated below.

9. Structured Discriminator

10. Blocking Discriminator

11. Cancelling Discriminator

12. Structured Partial Join

13. Blocking Partial Join

14. Cancelling Partial Join

15. Generalised AND-Join. Yes: Dataflow process type naturally fulfills this pattern owing to the mechanism of queuing signals built into the HyperFlow model of computation.

16. Local Synchronizing Merge

17. General Synchronizing Merge. Not supported

18. Thread Merge. Yes: this pattern can be implemented using a choice process, as in the AppEngine image gallery workflow where we need to wait for all image processing 'threads' to finish before we can generate the HTML file. To this end, the MergePaths process gets a count of items to be collected (pathCnt) and for every thumbPath received, it adds it to a collection, decreases the value of the pathCnt emits it to itself. When pathCnt reaches 0, the process emits the other signal (thumbPathList) signal, concluding the merge.

19. Thread Split. Yes: the parlevel property of a process enables this pattern, as in the AppEngine image gallery workflow.

20. Multiple Instances without Synchronization

21. Multiple Instances with a Priori Design-Time Knowledge

22. Multiple Instances with a Priori Run-Time Knowledge

23. Multiple Instances without a Priori Run-Time Knowledge

24. Static Partial Join for Multiple Instances

25. Cancelling Partial Join for Multiple Instances

26. Dynamic Partial Join for Multiple Instances

27. Deferred Choice

28. Interleaved Parallel Routing

29. Milestone

30. Critical Section

31. Interleaved Routing

32. Cancel Task

33. Cancel Case

34. Cancel Region

35. Cancel Multiple Instance Activity

36. Complete Multiple Instance Activity

37. Arbitrary Cycles

38. Structured Loop

39. Recursion

40. Implicit Termination

41. Explicit Termination

42. Transient Trigger

43. Persistent Trigger