Adding attachments to a HumanTask is a feature that exists in Oracle HWF (Human Workflow) since 10g.
However, in 11g there have been many improvements on this feature and this entry will try to summarize them.
Oracle BPM 11g 11.1.1.5.1 (aka PS4 Feature Pack or PS4FP) introduced two great features:
Ability to link attachments at a Task scope or at a Process scope:
"Task" attachments are only visible within
the scope (lifetime) of a task. This means that, initially, any member of
the assignment pattern of
the Human Task will be able to handle (add, review or remove) attachments. However, once
the task is completed, subsequent
human tasks will not have access to them. This does not mean those attachments got lost. Once
the human task is completed, attachments can be retrieved in order to, i.e., check them in to a Content Server or to inject them to a new and different
human task.
Aside note: a "re-initiated"
human task will inherit comments and attachments, along with history and -optionally- payload. See here for more info.
"Process" attachments are visible within
the scope of
the process. This means that subsequent
human tasks in
the same process instance will have access to them.
Ability to use Oracle WebCenter Content (previously known as "Oracle UCM") as
the backend for
the attachments instead of using HWF database backend. This feature adds all content server document lifecycle capabilities to HWF attachments (versioning, RBAC, metadata management, etc). As of today, only Oracle WCC is supported. However, Oracle BPM Suite does include a license of Oracle WCC for
the solely usage of document management within BPM scope.
Here are some code samples that leverage
the above features.
Retrieving uploaded attachments -Non UCM-
Non UCM attachments (default ones or those that have existed from 10g, and are stored "as-is" in HWK database backend) can be retrieved after
the completion of
the Human Task. Firstly, we need to know whether any attachment has been effectively uploaded to
the human task. There are two ways to find it out:
Through an XPath function:
Checking
the execData/attachment[] structure. For example:
Once we are sure one ore more attachments were uploaded to
the Human Task, we want to get them. In this example, by "get" I mean to get
the attachment name and
the payload of
the file. Aside note: Oracle HWF lets you to upload two kind of [non-UCM] attachments: a desktop document and a Web URL. This example focuses just on
the desktop document one. In order to "retrieve" an uploaded Web URL, you can get it directly from
the execData/attachment[] structure.
Attachment content (payload) is retrieved through
the getTaskAttachmentContents() XPath function:
This example shows how to retrieve as many attachments as those had been uploaded to
the Human Task and write them to
the server using
the File Adapter service.
The sample process excerpt is as follows:
A dummy UserTask using "HumanTask1"
Human Task followed by a Embedded Subprocess that will retrieve
the attachments (we're assuming at least one attachment is uploaded):
and once retrieved, we will write each of them back to a file in
the server using a File Adapter service:
In detail:
We've defined an XSD structure that will hold
the attachments (both name and payload):
Then, we can create a BusinessObject based on such element (attachmentCollection) and create a variable (named attachmentBPM) of such BusinessObject type.
We will also need to keep a copy of
the HumanTask output's execData structure. Therefore we need to create a variable of type TaskExecutionData...
...and copy
the HumanTask output execData to it:
Now we get into
the embedded subprocess that will retrieve
the attachments' payload. First, and using an XSLT transformation, we feed
the attachmentBPM variable with
the name of each attachment and setting an empty value to
the payload:
Please note that we're using
the XSLT for-each node to create as many target structures as necessary. Also note that we're setting an Empty text to
the payload variable. The reason for this is to make sure
the <payload></payload> tag gets created. This is needed when we map
the payload to
the XML variable later.
Aside note: We are assuming that we're retrieving non-UCM attachments. However in real life you might want to check
the type of attachment you're handling.
The execData/attachment[]/storageType contains
the values "UCM" for UCM type attachments, "TASK" for non-UCM ones or "URL" for Web URL ones. Those values are part of
the "Ext.Com.Oracle.Xmlns.Bpel.Workflow.Task.StorageTypeEnum" enumeration.
Once we have fed
the attachmentsBPM structure and so it now contains
the name of each of
the attachments, it is time to iterate through it and get
the payload. Therefore we will use a new embedded subprocess of type MultiInstance, that will iterate over
the attachmentsBPM/attachment[] element:
In every iteration we will use a Script activity to map
the corresponding payload element with
the result of
the XPath function getTaskAttachmentContents(). Please, note how
the target array element is indexed with
the loopCounter predefined variable, so that we make sure we're feeding
the right element during
the array iteration:
The XPath function used looks as follows:
hwf:getTaskAttachmentContents(bpmn:getDataObject('UserTask1LocalExecData')/ns1:systemAttributes/ns1:taskId, bpmn:getDataObject('attachmentsBPM')/ns:attachment[bpmn:getActivityInstanceAttribute('SUBPROCESS3067107484296', 'loopCounter')]/ns:fileName)
where
the input parameters are:
taskId of
the just completed
Human Task
attachment name we're retrieving
the payload from
array index (loopCounter predefined variable)
Aside note:
The reason whereby we're iterating the execData/attachment[] structure through embedded subprocess and not, i.e., using XSLT and for-each nodes, is mostly because the getTaskAttachmentContents() XPath function is currently not available in XSLT mappings. So all this example might be considered as a workaround until this gets fixed/enhanced in future releases.
Once this embedded subprocess ends, we will have all attachments (name + payload) in
the attachmentsBPM variable, which is
the main goal of this sample. But in order to test everything runs fine, we finish
the sample writing each attachment to a file. To that end we include a final embedded subprocess to concurrently iterate through each attachmentsBPM/attachment[] element:
On each iteration we will use a Service activity that invokes a File Adapter write service. In here we have two important parameters to set. First,
the payload itself.
The file adapter awaits binary data in base64 format (string). We have to map it using XPath (Simple mapping doesn't recognize a String as a base64-binary valid target):
Second, we must set
the target filename using
the Service Properties dialog box:
Again, note how we're making use of
the loopCounter index variable to get
the right element within
the embedded subprocess iteration.
Handling UCM attachments will be part of a different and upcoming blog entry. Once I finish will all posts on this matter, I will upload
the whole sample project to java.net.
