Correction. Final choice example should be this: (Does not have
explicit length. I need to follow my own advice :-)
<choice>
<sequence dfdl:terminator="-"/><!-- no elements at all in this
choice branch -->
<sequence>
<!--
Here you can put your sequence of 8 elements.
My preferred style would still be to use a single element.
-->
<element name="dt" type="xs:dateTime"
dfdl:calendarPatternKind='explicit'
dfdl:calendarPattern="YYYYMMdd'T'HHmmss" dfdl:calendarTimeZone='UTC'
dfdl:terminator="Z"/>
</sequence>
</choice>
On Tue, Nov 8, 2022 at 4:57 PM Mike Beckerle <[email protected]> wrote:
>
> I would certainly call this whole thing variable length because there
> are two representations. Length 1, or Length 16.
>
> But really here's a thought about fixed vs. variable length. Or really
> a thought about fixed length vs. delimited length.
>
> I think of this data as slash separated first, and then the things
> between the slashes must have other characteristics in order to be
> valid, such as the proper length.
>
> In your example, suppose the data is "..../20220321T180521/..."
>
> Looks fine. Oh wait, the "Z" for Zulu is missing.
>
> And that's what the diagnostic wants to tell you. It wants to take the
> slash separators as highest priority, isolate the string
> "20220321T180521" using the following separator, and then tell you
> that value isn't valid for your format because the Z is missing.
>
> But if your DFDL models this data as fixed length, then fixed length
> 16 characters would isolate this string "20220321T180521/" (note the
> slash has become part of the value region of the data). Then the
> diagnostic is going to be confusing because it will be complaining
> about "/" in the value. I would claim that's not a good way to model
> what people describe as a "slash separated format".
>
> In summary, you want to avoid lengthKind 'explicit' in such formats
> and really use lengthKind 'delimited' wherever possible.
>
> As for exactly how I would suggest to model your data, first here's a
> stab at a single element, nillable variation which is pretty clean.
> I'm assuming these live within a sequence having separators.
>
> <element name='dt' type="xs:dateTime" nillable='true'
> dfdl:nilValue='-'
> dfdl:nilValueDelimiterPolicy="none"
> dfdl:calendarPattern="YYYYMMdd'T'HHmmss"
> dfdl:terminator="Z" />
>
> (This assumes your default format for the schema is dfdl:lengthKind
> 'delimited', dfdl:calendarPatternKind='explicit',
> dfdl:nilKind='literalValue' which would be typical for a format like
> this.)
>
> I made the "Z" a terminator since it seems like your dates are always
> going to have Z for Zulu (aka UTC) time zone.
>
> If you want your 8 separate elements, then I would model the
> "no-value" hyphen as below, where no value is represented in the
> infoset by the absence of any date-time element(s).
>
> <choice>
> <sequence dfdl:terminator="-"/><!-- no elements at all in this
> choice branch -->
> <sequence>
> <!--
> Here you can put your sequence of 8 elements.
>
> My preferred style would still be to use a single element.
> -->
> <element name="dt" type="xs:dateTime"
> dfdl:lengthKind="explicit" dfdl:length='15'
> dfdl:lengthUnits='characters'
> dfdl:calendarPatternKind='explicit'
> dfdl:calendarPattern="YYYYMMdd'T'HHmmss" dfdl:calendarTimeZone='UTC'
> dfdl:terminator="Z"/>
> </sequence>
> </choice>
>
>
>
>
>
>
>
> On Tue, Nov 8, 2022 at 2:22 PM Roger L Costello <[email protected]> wrote:
> >
> > Hi Folks,
> >
> > Here is a sample input value:
> >
> > ../20090214T132530Z/..
> >
> > It is a single value, consisting of 8 parts (i.e., a composite field):
> >
> > 2009 is a 4-digit year named DateTimeIso
> > 02 is a 2-digit month named MonthNumeric
> > 14 is a 2-digit day named Day
> > T is the separator between the date and the time, named TimeSeparator
> > 13 is a 2-digit hour named HourTime
> > 25 is a 2-digit minute named MinuteTime
> > 30 is a 2-digit second named SecondTime
> > Z is the time zone, named TimeZoneZulu
> >
> > Here is another value for this field:
> >
> > ../-/..
> >
> > If no data is available, then the field is populated with a hyphen.
> >
> > Is this field fixed or variable length?
> >
> > /Roger
