[DFDL-WG] Action 157 - nilKind of logicalValue and nilValueDelimiterPolicy

[Mike Beckerle]

Action 157 - nilKind logicalValue and nilValueDelimiterPolicy

Summary: nilKind logicalValue is an "in-band" kind of null indicator.
Having separate delimiter rules for the distinguished null values is
unnecessary complexity.

Discussion: As is, the spec allows these two properties to be combined in
all ways. it is *possible* to implement combinations of
nilValueDelimiterPolicy with nilKind logicalValue, it's just very
complicated. Issuing quality diagnostic messages may be very difficult if
the user specifies ambiguous formats.

Example: initiator="[" terminator="]" separator="," lengthKind="delimited"
but the data contains commas also. as in 123.456,78. If the nilKind is
logicalValue, and one of the nilValues is -1,0 then nilValueDelimiterPolicy
must be "both" or the format won't parse because the initiators and
terminator won't be present to separate the data commas from the separator
commas.

Special case code in every DFDL implementation would have to deal with this
ambiguity, so not only is there additional complexity in just writing a
parser for correct and unambiguous DFDL schemas, if there are ambiguities
like the above, there is yet further complexity in issuing a quality
diagnostic message.  That is all not to mention the test cases needed to
insure conformance.

In the absence of a use case motivating this complexity, it is worthwhile
to rule this specific combination of properties out as a schema-definition
error, or simply make the combination such that when logical nil values are
used, the nilValueDelimiterPolicy is ignored. The conservative choice is
one that lets us put back this capability in the future should we discover
a viable use case for it. Hence, a schema-definition error is a better
choice for the errata.

Suggested Errata Language:

When dfdl:nilKind="logicalValue" then dfdl:nilValueDelimiterPolicy must be
"both". It is a schema-definition error otherwise. All values, including
nil indicator values, will have initiators and terminators if those
delimiters are specified. dfdl:nilValueDelimiterPolicy values other than
"both" are usable only with dfdl:nilKind of "literalValue" and
"literalCharacter".

TBD: should it apply to nilKind "literalCharacter" ?

Alternative:

When dfdl:nilKind="logicalValue" then dfdl:nilValueDelimiterPolicy is
ignored. All values, including nil values, will have initiators and
terminators if those delimiters are specified.

-- 
Mike Beckerle | OGF DFDL WG Co-Chair
Tel:  781-330-0412
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/dfdl-wg/attachments/20111122/921cb415/attachment.html