The dynamic indexing method shown in the following example applies a trailing XPath to the result of bwps:getVariableData(), instead of using an XPath as the last argument of bpws:getVariableData(). The trailing XPath makes reference to an integer-based index variable within the position predicate (that is, [...]).

<variable name="idx" type="xsd:integer"/>
...
<assign>
  <copy>
    <from expression="bpws:getVariableData('input','payload'
       )/p:line-item[bpws:getVariableData('idx')]/p:line-total" />
    <to variable="lineTotalVar" />
  </copy>
</assign> 

Assume at runtime that the idx integer variable holds 2 as its value. The expression in the preceding example within the from is equivalent to that shown in the following example.

<from expression="bpws:getVariableData('input','payload'
       )/p:line-item[2]/p:line-total" />

There are some subtle XPath usage differences, when an XPath used trailing behind the bwps:getVariableData() function is compared with the one used inside the function.Using the same example (where payload is the message part of element "p:invoice"), if the XPath is used within the getVariableData() function, the root element name ("/p:invoice") must be specified at the beginning of the XPath.

The following example provides details.

bpws:getVariableData('input', 'payload','/p:invoice/p:line-item[2]/p:line-total')

If the XPath is used trailing behind the bwps:getVariableData()function, the root element name does not need to be specified in the XPath.

For example:

bpws:getVariableData('input', 'payload')/p:line-item[2]/p:line-total

This is because the node returned by the getVariableData() function is the root element. Specifying the root element name again in the XPath is redundant and is incorrect according to standard XPath semantics.