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

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

On Sun, Aug 2, 2009 at 11:41 PM, <mike@adatinc.com> wrote:
> Joe Celko's and the other hierarchical processing solutions you mentioned is
> to use external programmable hierarchical programming functions and they
> also do not handle the necessary multiple node type and multiple data
> occurences necessary for XML.

Not sure what you mean by "external programmable hierarchical
programming functions"  You can use set/subset hierarchies with pure
SQL.  Nothing external is required....?

They can certainly handle multiple node types and multiple data
occurrences: just map the hierarchy using a table that is pure set /
subset and use a foreign key from it to a table joining multiple types
and / or multiple occurences as needed.

> Using the Left Out Join in a hierarhical data
> modeling fashion allows full multipath hierarchical processing to be
> transparently performed automatically and correctly including nonlinear
> multipath queries. See my DevX ANSI SQL Hierarchical Processing article at:
> http://www.devx.com/xml/Article/39183/1954
>

I read your articles, frankly I don't get it. Like I said, I can't see
how to manage large tree structures in the way you suggest.  In one
application we manage about 200 different major tree structures, one
instance of these has 60,000+ main nodes in it.  These nodes represent
in the order of 50 major types and probably over 200 subtypes.  The
tree nodes point to in the order of 300,000 data instances (this
particular tree is relatively sparsely populated). All this is managed
with about 10 tables using set / subset.  Any path traversal that I
need for any use case I've ever seen is realizable in pure ANSI SQL.
We create XML from these tables, on demand to build navigation menus,
OWL ontology dumps, and data management screens (which may include
multiple levels of hierarchy depending on the use case).

What am I missing here?

-- 
Peter Hunsberger