It's likely this is another ICU/lax issue. But I'd like to see the specific error that this gets, and the remaining properties that are in effect, e.g., the lengthKind is of particular importance here.
My point is that Daffodil/DFDL does not use the calendarXYZZY properties to determine the length of the xs:date element. The length is determined only based on the lengthKind and that this is representation 'text' (I assume). The resulting text string, once isolated based on length properties, is only subsequently then interpreted into an xs:date using the dfdl:calendarXXX properties. So if you add whitespace after this date, and those characters are included into the Birthday element text representation based on its lengthKind, then yes, I agree this is about ICU and lax mode not being sufficiently lax. But I'd like to see the error message and rest of the surrounding DFDL schema to confirm. ________________________________ From: Steve Lawrence <slawre...@apache.org> Sent: Wednesday, August 19, 2020 1:34 PM To: users@daffodil.apache.org <users@daffodil.apache.org> Subject: Re: Bug in Daffodil? Like with numbers, we use ICU for parsing calendars. It seems this is another case of ICU and and DFDL describing different behaviors for lenient/strict mode. With numbers, ICU decided that lax behavior is not tightly specified, and so we can't really depend on a particular behavior. I suspect the same goes for calendars. I think the DFDL-WG is working to make that change in the final revision so that DFDL it makes this more clear. On 8/19/20 1:22 PM, Roger L Costello wrote: > The spec says this for calendarCheckPolicy: > > Additional lenient parsing behaviour when in 'lax' mode: > > Leading and trailing whitespace > in the data but not in the pattern > is accepted > > The following element declares a lax policy: > > <xs:element name="Birthday" type="xs:date" > dfdl:calendarCheckPolicy="lax" > dfdl:calendarPatternKind="explicit" > dfdl:calendarPattern="yyyy:MM:dd" > dfdl:calendarFirstDayOfWeek="Sunday" > dfdl:calendarDaysInFirstWeek="5" > dfdl:calendarTimeZone="UTC+6" > dfdl:calendarLanguage="en" > /> > > Daffodil reports an error if I put whitespace following this date: > > 1970:11:06 > > The spec says that whitespace should be allowed with a lax policy. I believe > this is an error in Daffodil. Do you agree? > > /Roger >
