TMQL: Last Changes and Issues
After the Oslo meeting in March we have made the following changes to the TMQL specification:
- aligned the data types with CTM
- added at least N and at most N quantifiers (TMCL needs it)
- removed all ontological definition stuff (predicate, function definitions, etc.) from the environment map (this may at some point go into a TM ontology language)
- simplified the grammar somewhat
- added more XML support (as decided in Oslo)
- added most of the formal semantics
- function invocations are now defined on tuple sequences, i.e. a function defined on a tuple will be invoked for each tuple in the sequence. It is elegant, although I expect Java developers to faint.
- Reformulated the predefined functions in CTM (or what I thought is CTM :-)
I also tried to get rid of the term characteristic in the spec. There it is simply used as an superordinate concept for names and occurrences (no associations are involved here) to simplify the description. Without it, the text became incredibly repetitive and some of the meta grammar broke, so I gave up that plan.
Open issues (apart from what may emerge):
- the formal semantics for atomification (and de-atomification) is missing, also that for auto-atomification
- the sorting of binding sets in the context of asc and desc vectors has to be ironed out without breaking the variable elimination mechanism
- the top-level evaluation function might need a bit more consideration
I hold a private copy of the current draft.