Re: TMQL: Direct (super|sub)types?
"Andrew S. Townley" <[email protected]>
| Newsgroups | gmane.text.xml.xtm.general |
|---|---|
| Message-ID | <[email protected]> |
On 2 Mar 2011, at 10:04 AM, Lars Heuer wrote: > Hi Rani, > >> A simple use case will be just to provide a graphical visualization of >> the hierarchies in the topic map. > [...] >> Or maybe, you prefer to avoid from having "direct-supertypes" and >> "direct-subtypes" in TMQL because we can always use association pattern >> as an alternative: > > I see. Well, I find the term "direct-" awkward. TMQL introduces a > terminology which does not exist outside of TMQL (neither in TMDM nor > TMRM nor (to my knowledge) in any other knowledge representation > system) and I wonder if the mentioned use cases are common enough to > satisfy a shortcut in TMQL. I guess my answer would be "no" and I > wouldn't make these axes/paths first class citizens in TMQL. I'm with Lars here. I would expect an easy way to get all of the classes explicitly referenced by the author, e.g. all of the isa values, but I'm not sure that I would expect direct support for walking the inheritance tree via TMQL or any other query language. Having an API to provide a tree (to support the potential multi-inheritance) or a list (for single inheritance) is perfectly acceptable. Otherwise, if such a beast was required for TMQL, I would expect it to be called inheritance or something like that. Also, I'd say that since there were originally looping constructs available in TMQL, why wouldn't a recipe/idiom for obtaining the inheritance for a particular topic be sufficient? TMQL isn't supposed to augment, not replace a host language, right? If so, then having API support for all of the less-frequent operations or even for the entire ontology interaction (which is one of the options in my system) seems to meet the needs of the use case Rani mentioned. The direct vs. indirect is very awkward, and I don't see why it should be a first-class citizen of TMQL either. You should be able to retrieve the information required via more primitive path navigations if you wanted to stay in TMQL, or you always have API support of the host environment. Cheers, ast -- Andrew S. Townley <[email protected]> http://atownley.org