Re: transitively closed relation

From: --CELKO-- <71062.1056_at_compuserve.com>
Date: 20 Feb 2002 15:27:58 -0800
Message-ID: <c0d87ec0.0202201527.37c3bf05_at_posting.google.com>


>> Transitive closure relation has been traditionally viewed as an
incremental evaluation structure over adjacency list. ... Now, lets drop adjacency list altogether, and declare transitive closure as "the final relational representation" for hierarchy! <<

But the number of rows in the TC table is the (number of nodes -1), which can get large very fast. It also dos ot support aggregation up the heirarchy.

>> Now deleting the node no longer has the anomaly that you mentioned
for
adjacency list. Also, if we want to find immediate parent or children of the node, we can query it with a single join. <<

True. Itzak Ben-Gan uses a string with the full path from the root in it for his path enumeration model:

 node path

 ('a', 'a')
 ('b', 'a,b')
 ('c', 'a,b,c') 

then he uses string operators and LIKE predicates for his queries. Received on Thu Feb 21 2002 - 00:27:58 CET

Original text of this message