As discussed on the call, here is some more in-line below in green and tagged SMH+. A couple of other things when writing this up. 1) I think separatorPolicy="required" is misleading and I'm sure contributed to Tim's questions about behaviour. Here we are using "required" to mean that all delimiters are needed, even when the data itself is not required. I think we should use "always". 2) I'd forgotten that there is also separatorPolicy="suppress". In this case, any missing element does not get a separator. The spec states this " implies the children of the sequence must have dfdl:initiator specified. " but it does not say whether the omission of an initiator is a schema definition error. Should it be? Regards Steve Hanson Programming Model Architect, WebSphere Message Brokers, OGF DFDL WG Co-Chair, Hursley, UK, Internet: smh@uk.ibm.com, Phone (+44)/(0) 1962-815848 ----- Forwarded by Steve Hanson/UK/IBM on 25/11/2009 16:11 ----- From: Steve Hanson/UK/IBM To: dfdl-wg@ogf.org Date: 25/11/2009 12:37 Subject: Re: [DFDL-WG] Omitted array occurrences Tim, Alan - my thoughts on this in blue (SMH). Regards Steve Hanson Programming Model Architect, WebSphere Message Brokers, OGF DFDL WG Co-Chair, Hursley, UK, Internet: smh@uk.ibm.com, Phone (+44)/(0) 1962-815848 From: Alan Powell/UK/IBM@IBMGB To: Tim Kimber/UK/IBM@IBMGB Cc: dfdl-wg@ogf.org, dfdl-wg-bounces@ogf.org Date: 19/11/2009 17:30 Subject: Re: [DFDL-WG] Omitted array occurrences Sent by: dfdl-wg-bounces@ogf.org Tim Comments below Need more discussion on this Alan Powell MP 211, IBM UK Labs, Hursley, Winchester, SO21 2JN, England Notes Id: Alan Powell/UK/IBM email: alan_powell@uk.ibm.com Tel: +44 (0)1962 815073 Fax: +44 (0)1962 816898 From: Tim Kimber/UK/IBM@IBMGB To: dfdl-wg@ogf.org Date: 19/11/2009 12:07 Subject: [DFDL-WG] Omitted array occurrences What should the DFDL unparser do when some or all of the elements of an array are missing? I have found the following statements in v0.36 which seem relevant: Section 5.2.1 The minOccurs value is used: · to determine if an element declaration or reference is scalar or array · to determine the required minimum number of occurrences of an array both when parsing and unparsing Section 16.13 ( Note : this definition of 'required' is a repeat of the defintion in section 3 ) Definition: 'required' We define the term 'required' as follows: · A scalar element is required. · An element of a fixed-occurrence array is required. · An element of a variable-occurrence array is required if its index is less than or equal to the value of minOccurs. All other elements are not required. ... On unparsing, if an element is required, and is not part of the logical data and the element has a default value specified then it is used, otherwise it is a processing error. Section 17.3.1 : Sequence groups and separators re: the combination of separatorPolicy="suppressAtEnd" and sequenceKind="ordered": All separators must be found in the data except that when the sequence has trailing optional items, the separators are suppressed for any final missing items. Note suppressAtEnd can only be used when there is no clash with delimiters from the containing structure. My interpretation of the specification is: a) if separatorPolicy="require" then the unparser should output a separator for all missing required elements ( whether array members or not ) Is this an additional definition of a 'required' element? In which case the default value should be output. (interestingly because default is a schema property rather than a dfdl property you cannot set a default default.) SMH: The definition of 'required' relates to the data. Here we are talking about whether to output syntax. Strike 'required' from Tim's interpretation and you have the correct interpretation. b) if separatorPolicy="suppressAtEnd" then the unparser should output a separator for all non-trailing missing required elements Should set the default for any required element so it won't be missing. "On unparsing, if an element is required, and is not part of the logical data and the element has a default value specified then it is used, otherwise it is a processing error. " SMH: Tim's interpretation is not complete. The correct interpretation is "...then the unparser should output a separator for all missing elements in the sequence up to and including the last required element.". It is only optional elements beyond the last required element that benefit from this property. c) separators for missing elements must be output regardless of whether the element is required/optional, simple/complex, does/does not have a default value etc. I assume this because the term 'missing' is used rather than the very clearly-defined term 'required'. Missing just means not in the infoset and is orthogonal to optional/required. If you accept this is an additional definition of required then no. But it then forces you to set defaults for minOccurs=0 elements which will only be used in this circumstance. I'm not sure what the default for complex elements would be: all the children must have a default? . SMH: If c) is trying to say that once you have decided, via a) and b), that a separator is needed, then whether it is simple/complex, does/does not have a default, is irrelevant, then I agree. Reading between the lines, I also infer the following rules: d) if an array has maxOccurs="unbounded" and it is missing from the infoset then the unparser will not output any separators for the array e) if an array has maxOccurs!="unbounded" and it is missing from the infoset then the unparser will output a separator for each missing occurrence ( so it will output maxOccurs separators ). If minOccurs > 0 then use default. If minOccurs= 0 then output nothing. I don't think maxOccurs has any effect. SMH+: The behaviour when dealing with a repeating element (minOccurs, maxOccurs) is analogous to dealing with a sequence. You treat up to and including minOccurs as 'required', and anything beyond as 'optional'. Then you apply separatorPolicy property. So "suppressAtEnd" means you only output delimiters up to an including minOccurs, and "required" means you output delimiters up to and including maxOccurs. There's clearly a problem with the combination of maxOccurs="unbounded" and separatorPolicy="required" - this should be a schema definition error. SMH+: It is possible that some models are pretty ambiguous, and that we could be outputting something that is very difficult to parse. If it is possible to use the full DFDL armoury of parsing techniques (speculation, backtracking, data patterns, remodelling as choice and discriminators, etc) then that is a 1.0 limitation. f) if an element contains a child group, and none of the group members are present in the infoset, then the group is 'missing' and the unparser will output a separator for it. Not sure SMH: This is establishing 'missing' for a local group. Sounds right to me. The separator will be output according to a) and b). But because a local group is (1:1) in DFDL, in practice you will always get a separator. SMH+: If a local group needs to be optional it must be wrapped in a complex element. Suggested changes to the specification: - As a minimum, I think it would be useful for the specification to include a definition of 'missing'. 'Not in the infoset' SMH: That's fine for unparsing only. - DFDL does not allow min/maxOccurs on groups, so they implicitly have cardinality 1:1. Specification should specify the behaviour of the unparser when none of a group's members are present in the infoset. Agree. - The wording in 17.3.1 could be more accurate. I don't think the word 'optional' should be there ( if validation is off then the unparser will tolerate missing required elements -No. 'required' is not part of vaildation). I think the words 'trailing' and 'final' are intended to mean the same - we should standardize on 'trailing'. SMH: I agree the words could be improved. See my b) words above for example. regards, Tim Kimber, Common Transformation Team, Hursley, UK Internet: kimbert@uk.ibm.com Tel. 01962-816742 Internal tel. 246742 Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU -- dfdl-wg mailing list dfdl-wg@ogf.org http://www.ogf.org/mailman/listinfo/dfdl-wg Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU -- dfdl-wg mailing list dfdl-wg@ogf.org http://www.ogf.org/mailman/listinfo/dfdl-wg Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU