"Westhead, Martin
\(Martin\)" <westhead@avaya.com>
13/12/2005 17:57
To
Geoff Judd/UK/IBM@IBMGB
cc
<dfdl-wg@ggf.org>
Subject
RE: [dfdl-wg] Telcon Notes
Hi Geoff,
I’ll certainly put this on
the agenda. If anyone has anything else please let me know.
Also some comments:
1. I’d
like to understand how variables relate to parameters, specifically:
a. Could
we actually define variables to be user-defined parameters? (if not, why
not?)
b. If
(a) doesn’t work can we have variables use the same scoping rules as parameters?
GJ: This is a good point which
I have omitted from the semantics document. However I had considered this
when looking at late binding of schemas and runtime schemas. I think variables
in DFDL should be similar to XSLT parameters. That is the value of a variable
can be set externally although the mechanism for this will not be specified.
If a value of a variable is set externally it is the equivalent of binding
a value to the variable so it cannot have another value bound to it. Also
I think that for a valoue to be set externally the declaration of the variable
should be made at the schema level.
2. In
the extensibility group we have been talking about reference direction.
It was proposed that XPaths in a DFDL document can only refer to elements
higher up the document (i.e. backwards only) they cannot reference things
below them (which will not yet have been parsed). With the use of “hidden
layers” this can be done without loss of generality.
I would suggest that variable
references should work this way. I.e you are only allowed to refer to data
that has already been parsed.
GJ: I agree that only variables
that have been declared and have either a default value or a value bound
to them can be referenced. If this is not the case it will result in a
runtime error. Does this answer your question or have I mis-understood
what you mean ?
Thanks,
Martin
From: Geoff Judd [mailto:JUDDG@uk.ibm.com]
Sent: Tuesday, December 13, 2005 9:34 AM
To: Westhead, Martin (Martin)
Cc: dfdl-wg@ggf.org
Subject: Re: [dfdl-wg] Telcon Notes
Martin,
Please can you add a discussion
of the semantics of DFDL Variables to the agenda of tomorrow's telecon.
I have attached a document with some thoughts that I have had on the subject.
Regards,
Geoff Judd
Websphere MQ Integrator Development
IBM UK Ltd
Hursley
"Westhead, Martin
(Martin)" <westhead@avaya.com>
Sent by: owner-dfdl-wg@ggf.org
30/11/2005 18:12
To
<dfdl-wg@ggf.org>
cc
Subject
[dfdl-wg] Telcon Notes
Present
-------
Mike Bekerle
Tara Talbott
Martin Westhead
Geoff Judd
Steve Hanson
Bob McGrath
Reports from Active Trackers
----------------------------
Scoping tracker is being worked on - no news.
Extensibility group is moving forward - nothing to report yet
Discussion of Geoff's documents
-------------------------------
Variables doc
Mike observed that variables could be used in place of his extensibility
proposal on annotations.
Martin asked about whether we wanted to include this in the standard. Is
the complexity worth the addition - the feeling of the meeting was
probably it was.
Martin proposed that instead of binding a variable value to an element,
the variable value should be set with an XPath expression.
Discussed the question as to whether the variable value can be reset,
whether it can/should be scoped in the same way as annotation
attributes.
Next step will be for Geoff to revise the document to look at cases
where value may need to be reset and start to explore some of the
semantic issues involved.
Choices doc
Discriminator type - what if discriminator is not a string? For
efficiency can we handle discriminators that are integers?
Suggest that instead of discriminators being computed to strings they
are computed to an XPath value with its implied type so that numeric
discriminators do need to be converted to strings.
Wildcard docs
Group seemed fairly happy with this proposal as it stands.