From Mike's e-mail:

We need to improve test coverage on dfdl:outputValueCalc.

Here are a few ideas for "torture tests".

I would work on these, but I still have a final unparser feature to code, and a bunch of review-related changes to do.

• • create runtime-valued property and dfdl:outputValueCalc torture tests

These tests are based around:

<element name="xi" type="xs:string" dfdl:length="1" dfdl:lengthKind="explicit"/>
<element name="yi" type="xs:string dfdl:length="1" dfdl:lengthKind="explicit""/>
<element name="zi" type="xs:string" dfdl:length="1" dfdl:lengthKind="explicit"/>

<element name="x" type="xs:string" dfdl:outputValueCalc="

{ dfdl:valueLength(../y) }

" dfdl:initiator="

{ ../xi }

" dfdl:lengthKind="delimited" dfdl:terminator=";"/>
<element name="y" type="xs:string" dfdl:outputValueCalc="

{ dfdl:valueLength(../z}

)" dfdl:initiator="

{ ../yi }

" dfdl:lengthKind="delimited" dfdl:terminator=";"/>
<element name="z" type="xs:string" dfdl:initiator="

{ ../zi }

" dfdl:lengthKind="delimited" dfdl:terminator=";"/>

The dfdl:outputValueCalc's in the above require the unparser to suspend unparsing of x and y, then unparse them first z, then y, then x, but put their output in x,y, z order.

The test is to make sure this happens correctly even if each of x, y, z gets a runtime-valued property's values are all different. E.g.,
<xi>X</xi>
<yi>Y</yi>
<zi>Z</zi>
<z> 0123456789</z>

unparsing shoudl produce

XYZX2;Y10;Z0123456789;

Generalize this idea to all the other runtime-valued properties beyond just initiator.

byteOrder
encoding
outputNewLine
initiator
terminator
separator
escapeCharacter
escapeEscapeCharacter
textStandardDecimalSeparator
textStandardGroupingSeparator
textStandardExponentRep
calendarLanguage

• • create test that dfdl:outputValueCalc freezes the delimiter stacks

similar concept to the above, except the delimiters come from surrounding nests of sequences.

When each unparses (which should be y first then x), it should get the right delimiters. This insures we're freezing the delimiter stack, and the values of the delimiters on them.

• • create test that dfdl:occursIndex freezes properly for OVC

similar concept to the above, but the OVC expressions depend on freezing of different dfdl:occursIndex values. (Two arrays of the same length, contain complex types. Inside those are OVC elements....)