I tested with 2.4.0. If I remove the CRCRLF separator on line 33 that
you have highlighted then I get an error about no forward progress:
$ daffodil --version
Apache Daffodil (incubating) 2.4.0
$ daffodil parse -Dheader=absent -s samp.dfdl.xsd samp.txt
[error] Parse Error: Choice dispatch branch failed: List(Parse Error: No
forward progress.
Schema context: sequence[1] Location line 33 column 12 in
file:/home/slawrence/downloads/samp/samp.dfdl.xsd
Data location was preceding byte 37)
Schema context: choice[1] Location line 31 column 14 in
file:/home/slawrence/downloads/samp/samp.dfdl.xsd
Data location was preceding byte 37
[error] Parse Error: No forward progress.
Schema context: sequence[1] Location line 33 column 12 in
file:/home/slawrence/downloads/samp/samp.dfdl.xsd
Data location was preceding byte 37
Do you have any other changes to the schema?
On 4/19/22 11:05 AM, Attila Horvath wrote:
Keep in mind I'm using Daffodil 2.4.0
I understand and your explanation makes sense; however...
When I remove highlighted DFDL separator attributes on line #33 per
suggestion:...
image.png
I get identical result:...
image.png
I.E. <CR><CR><LF> still consumed prior to 'tailing-rec' element processing.
Unexpectedly, the only way to get the 'expected' result is by adding additional
<CR><CR><LF>:...
image.png
You've confirmed my understanding of 'infix'; however, not getting expected
results.
Thx - Attila
On Tue, Apr 19, 2022 at 10:40 AM Steve Lawrence <[email protected]
<mailto:[email protected]>> wrote:
Thanks for the provided schemas. This helps quite a bit.
In this case, the "record" element does not consume the final CR CR LF
because it is an infix separator, as you expect.
However, you have an CR CR LF separator on the outer sequence on line
33, so you have something like this:
<xs:sequence dfdl:separator="%CR;%CR;%LF;"
dfdl:separatorPosition="infix">
<xs:sequence>
<xs:element name="MU16Type" ... />
</xs:sequence>
<xs:element name="trailing-rec" ... />
</xs:sequence>
So this expects a CR CR LF in between your MU16Type element and a
trailing-rec element. That is what actually consumes the final CR CR LF
and then the trailing-rec element happily consumes no data and you get
the NIL element.
So the suggestion if you want the CRCRLF element to be parsed in the
tailing-rec is to remove the CR CR LF separator from the outer sequence.
Note that things get a bit tricky if you do that. In that case, the
final trailing CR CR LF is not consumed as an infix separator, as
expected. And then we do get a single trailing-rec element that consumes
the CR CR LF and we get the <CRCRLF> element. But then because
trailing-rec is unbounded, and happy to try to parse more trailing-rec
elements. But a trailing-rec element can consume no data because of the
NIL element. This would mean Daffodil could get stuck in an infinite
loop happily consuming no data. Daffodil detects this and errors with a
"No forward progress" message. Simply removing the NIL element from the
TailType-TLE fixes this issue, this way it's not possible for
TailType-TLE to consume no data.
- Steve
On 4/19/22 10:12 AM, Attila Horvath wrote:
> ALCON
>
> Same question/issue - see attached samp schema/data.
>
> Thx - Attila
>
> On Tue, Apr 19, 2022 at 10:03 AM <[email protected]
<mailto:[email protected]> <http://apache.org <http://apache.org>>> wrote:
>
>
> > I can't reproduce this with simpler examples. Can you
provide your
> > actual schema/data?
> >
>
