KDL data model

[nichtich]

KDL specification does not explicitly define a data model but only a syntax that implies some data model of KDL. As long as you parse from KDL syntax into some data structure of a programming language, a KDL data model is not needed. In contrast both KQL and KDL Schema build on a data model. The lack of its explicit definition leaves open handling of some edge cases, such as:

• Is an empty string tag different to no tag ( does ("")value differ from value)?
• Is an empty list of child nodes different from no child nodes (does node {} differ from node)?

The data model should abstract from syntax details such as comments, how nodes are terminated, whether names are given as bare identifiers or given as strings etc.

Here is a draft of a possible data model (without explicit definition of name and value and both questions above answered with no):

• a document is a list of nodes
• a node consist of
  • a tag being a name
  • a name
  • a list of arguments, each being a value
  • a set of properties, each consisting of a name and value, and each name being unique
  • a list of children being a document

[larsgw]

Looks good IMO.

• Is an empty string tag different to no tag ( does ("")value differ from value)?

Yes, I think similarly node names can also be just "" (so the draft does not change). See also #177.

• Is an empty list of child nodes different from no child nodes (does node {} differ from node)?

Not sure, but I do know that kdljs does not distinguish between the two but kdl4j does.

[hkolbeck]

I made the distinction in kdl4j since it seemed odd for someone to specify a present-but-empty child, and so they might mean something specific by it. I am not attached to that choice.

[nichtich]

Whether ("")value differs from value and whether node differs from node {} is a matter of choice of the KDL parser, so it is not specified yet. This is not important for several use cases but at least KDL Query Language should be clear about what to expect.

[tabatkins]

Currently the tests are clear that ("")value and value are different, and node and node {} are different.

[zkat]

and that seems fine to be tbh

[nichtich]

Ok if both child nodes and tags are nullable, the abstract KDL data model should be:

• a document is a list of nodes
• a node consist of
  • a name
  • an optional tag being a name
  • a list of arguments, each being a value
  • a set of properties, each consisting of a name and value, and each name being unique in the set
  • an optional list of children being a document
• a name is a list of Unicode code points
• a value is either
  • a list of Unicode code points
  • a number, consisting of an integer part and a decimal part,
  • a boolean, being one of the special values true and false,
  • or the special value null

KDL defines a syntax of this data model but KQL and KDL purely act on the abstract KDL data structures independently from syntax.