[wsbpel] Issue - 11 XPath function summary

["Chris Keller" <chris.keller@active-endpoints.com>]

As promised a summary of a proposal for new XPath functions for solving
issue 11.

Proposal for Issue 11:

Currently an Assign Copy operation has the restriction that a <to>
clause which uses a variable/part and query must return exactly one
node.  If it doesn't then a fault is thrown.  This means that <copy> in
its current form can only be used to update detailed information
contained in a message part.  To address this issue as well as the tight
schedule of the wsbpel TC the following proposal for new XPath functions
are proposed.

The current specification specifies three mandatory XPath functions
which engines must implement to allow access to process information.
The following two new functions would be added to address new node
creation within a <copy> operation.  These functions can have some
restrictions in particular they can only be used within a <to> query
attribute.

bpws:appendNode("childXPath" [, "parentXPath"] [, "siblingXPath"])
bpws:insertNode("childXPath" [, "parentXPath"] [, "siblingXPath"])

childXPath: The path of the child to be created, which must use the
abbreviated form of XPath (e.g. a/b[2]/c).  Note that we could place
other limitations on this path as well.

parentXPath: An XPath expression which evaluates, which must evaluate to
exactly one node.  It is optional and if not present then the context of
the <to> variable part will be used for the requested node creation.  

siblingXPath: An XPath expression which evaluates, which must evaluate
to exactly one node.  It is executed within the context of the the
parent query's result.  It is optional and when present it provides a
pivot point for the new node creation.  If it is not present then an
insert opertaion create a new node as the first child and an append will
create a new node as the last child.

These functions when called will create a new node at the requested
location and return the node.  This will ensure that the query returns
exactly one node.

Given the following starting document fragment, which represents a
message part, here are some examples.

<po>
  <header>
    <poNumber/>
    <city/>
  </header>
</po>

Example: bpws:appendNode("detail", "/po")

Results in:

<po>
  <header>
    <poNumber/>
    <city/>
  </header>
  <detail/>
<po>

Example: bpws:insertNode("detail", "/po")

Results in:

<po>
  <detail/>
  <header>
    <poNumber/>
    <city/>
  </header>
<po>


Example: bpws:appendNode("name", "/po/header", "poNumber")

Results in:

<po>
  <header>
    <poNumber/>
    <name/>
    <city/>
  </header>
<po>


Example: bpws:insertNode("name", "/po/header", "poNumber")

Results in:

<po>
  <header>
    <name/>
    <poNumber/>
    <city/>
  </header>
<po>


These examples would be used as part of a Assign/Copy operation.

<assign>
  <copy>
     <from.../>
     <to variable="..." part="..." query="examples"/>
  </copy>
</assign>

Chris