AS4 Progress

1
AS4 Progress

• Te Li
• 10/9/2006

2
Agenda

• Motivation
• Architecture
• Approach
• Junctions
• Discussion

3
Motivation

• Vast amount of XML streams SOAP messages, online
  auctions, financial tickers,
• Access XML documents sequentially
• No XML database system at hand
• On tape
• Query language for XML streams
• XPath?
• XQuery! (may need to be extended for windowing)
• (k-way) joins, self-joins, torch-joins, and
  (k-way) cross-comparisons
• Result construction with iterations
• Rich set of built-in functions (in XPath2 too)
• User-defined functions and modules
• Strongly-typed and incorporated XML Schema

4
Architecture

• Its the goal.
• For now, Optimizer and Combiner are not touched.
• Step by step consider one-one case first.
• The brain of the AS4 system is Executor.

5
Approach

• Node-Tag Flow Network
• Data flow XML stream-gtnode-tag flow-gtresult
  stream
• A node-tag flow contains the following items
  start tags, end tags, attribute nodes, and text
  nodes.
• The network components a source node, channels,
  junctions, and an output node.

6
Junctions

• Self-joins
• K-way joins
• K-way torch-joins
• K-way comparisons

7
Self-Join Junctions

• The simplest type of junctions
• If no sorting on the join column (d), hash table
  would be enough. Otherwise, it is a special 1-way
  join.

8
K-Way Joins

• This is a 2-way join query (two 1-way join
  junctions). A 2-way join query can have one 2-way
  join junction.
• How to process it? The data structure junction
  table

pn
_at_id
ca
_at_pn
it
_at_id
_at_it
t(n)
t(n)
Output variables
9
K-Way Torch-Joins

• A k-way torch-join junction is a junction that
  has a self-join followed by a k-way join.
• A special case of k-way join

10
K-Way Comparisons

• The junction condition is not just an equality
  test, can be any one of gt, gt, lt, lt, , and !.
• Contains k-way joins and k-way non-join
  comparisons.
• So the key point to solve this problem.

11
Discussions?