Re: [xml-dev] Why Multipath (LCA) Hierarchical Query Processing Works Au

[Peter Hunsberger <peter.hunsberger@gmail.com>]

On Sun, Aug 2, 2009 at 10:25 PM, <mike@adatinc.com> wrote:
> Basic ANSI SQL inherent hierarchical processing using the Left Outer Join to
> model and processes hierarchical structures is basically quite obvious and
> empirical and I have covered this in my previous SQL/XML articles. The proof
> for multipath hierarchical query processing which requires Lowest Common
> Ancestor (LCA) processing occurring naturally in ANSI SQL is not that
> obvious because it is truly quite amazing since it was never designed into
> ANSI SQL and is quite complex to perform. Empirically it can be proven from
> its results, but it would be very nice to know how and why it is working so
> we can absolutely trust the results. I have written an article describing
> the how and why of natural LCA processing in ANSI SQL and have appropriately
> entitled it “The Ghost in the Machine”. It can be located below.
>
> The Ghost in the Machine
> http://www.tdan.com/view-articles/11069
>
> This LCA processing in XQuery is not automatically performed today and is
> too complex to do with procedural navigation. This problem has been
> researched academically and attempted solutions use LCA functions that have
> to be inserted correctly by the query user which takes away for its ease of
> use and schema-free purpose. My work with LCA processing has shown that LCA
> processing can involve nesting LCA’s that I do not necessarily see occurring
> in this LCA XQuery research limiting their future solutions to more simple
> queries. ANSI SQL performing LCA processing automatically has no multi-path
> LCA query limitations. This has been referred to a LCA query processing, at
> least three decades ago.

Mike,

Perhaps I'm missing something here, surely you're not really
suggesting that we should code outer joins to manage hierarchical
structures in a relational database?  Why not just use Celkos set /
subset tree management (or other related algorithms)?  We've got some
tree structures several 1000 nodes deep and wide managed in a
relational database, can you me exactly how you would go about finding
(for example) all the leaf nodes for such a tree?

-- 
Peter Hunsberger