Re: TMQL: Direct (super|sub)types?

On 5 Mar 2011, at 12:15 PM, Robert Cerny wrote:

> Am 05.03.2011 um 00:30 schrieb Benjamin Bock:
> 
>>>       s1 // types for all types,
>>> 
>>>       s1 / types for direct types
>>> 
>> 
>> That didn't fit my initial mental model but maybe it's a good idea.
>> I'll at that to my think-about-it list. Thanks.
> 
> Just for the records: When writing '//types', i feel like something is missing between the two slashes (also in XPATH). I prefer Andrew's first proposal '/*types' for 'all types', because the star has the connotation of 'deep' or 'transitive' for me. Yet, i find the current proposal '/types' for 'all types' and the less frequent (so longer is ok) '/direct-types' for direct types, very good.

To be fair, I think it was Patrick who suggested it first.

Here we're back to aesthetics vs. functionality.  I like the '//' better because I know and have wrestled with XPath a good bit and it kinda means the same type of thing.  Reusing norms and precedents from the wild where possible is a good thing! :)

I could live with '/*' but I'd still separate them with whitespace either way (because that's just the kinda guy I am...).  The only time I've ever encountered '/*' is as the start of a C style comment, and that would be the first thing I'd think of.

I actually wouldn't care as long as I could be able to specify both modes in the same query, e.g. something like:

	select $x where ($x // types) - ($x / types) => bag of all types for $x which aren't explicit types

Of course, I'd also then expect to be able to do t / ROLE and t // ROLE for arbitrary roles as well.  I think that kind of thing's already supported in the proposals though (must check....)

ast
--
Andrew S. Townley <[email protected]>
http://atownley.org