ViST: a dynamic index method for querying XML data by tree structures

1
ViST a dynamic index method for querying XML
data by tree structures

• Authors Haixun Wang, Sanghyun Park, Wei Fan,
  Philip Yu
• Presenter Elena Zheleva, November 2003

2
Overview

• Modeling XML Queries
• Structure-encoded sequences
• Indexing
• ViST
• Experimental Results

3
Modeling XML Queries
4

• DTD of purchase records
• (!ELEMENT purchases (purchase))
• (!ELEMENT purchase (seller, buyer))
• (!ATTRIST seller ID ID location CDATA name CDATA)
• (!ELEMENT seller (item))
• (!ATTRIST buyer ID ID location CDATA name CDATA)
• (!ELEMENT item (item))
• (!ATTRIST item name CDATA manufacturer CDATA)

5
Modeling XML Queries

• Focus in XML query language design ability to
  express complex structural or graphical queries

6
Modeling XML Queries

• Querying XML data finding sub structures of the
  data graph that match the sequence
• Structure-encoded sequences a sequential
  representation of both XML data and XML queries

7
Structure-Encoded Sequences
8
Structure-Encoded Sequences

• Maps the data and the queries
• Matches the subsequence
• Purpose to avoid as many join operations as
  possible
• Def. Sequence of (symbol, prefix) pairs

9
Mapping Data

• Represent XML document/tree in preorder
• Represent in structure-encoded seq

10
Mapping Queries

• Benefit of sequence matching query gets
  processed as whole
• Path Expression

11
Structure-Encoded Sequences

• Query
• Data

12
Querying XML

• through Structure-Encoded Sequence Matching

13
Indexing
14
Role of Indexing

• To provide an algorithm to perform this sequence
  matching
• Desired features for algorithm
• Efficient support for subsequence matching
• Use well-supported DB indexing techniques such as
  B trees
• Allow dynamic index insertion

15
What is indexing useful for

• Auxiliary access structures
• Used to speed up the retrieval of records
• In response to certain search conditions
• Provide efficient support for arbitrary
  structured queries
• Using wild-cards // and

16
Indexing

• State-of the-art approaches
• Indexes on paths
• Indexes on nodes
• Indexes on both (structures) ViST

17
ViST
18
Algorithms

• Naïve Algorithm based on Suffix Trees
• RIST Relationships Indexed Suffix Tree
• ViST Virtual Suffix Tree

19
Algorithm Using Suffix Trees

• Suffix Tree a compact index to all distinct,
  contiguous substrings of a string
• D-Ancestorship in XML doc tree
• Through structure-encoded sequence
• S-Ancestorship in suffix tree

20
Example Using Suffix Trees
21
Algorithm Using Suffix Trees

• Searches
• first by S-Ancestorship searching under suffix
  tree
• then by D-Ancestorship matching nodes and
  prefixes
• Disadvantages
• Costly traverse large portion of subtree
• Most commercial DBMSs do not support

22
RIST Indexing by Ancestor-Descendant
Relationships

• Jumps directly to the nodes Y to which X is both
  a D-Ancestor and S-Ancestor
• Index Construction uses B trees

23
RIST Indexing by Ancestor-Descendant
Relationships

• Subsequence Matching
• Determine D-Ancestorship by prefixes
• Determine S-Ancestorship by label ltnx,sizexgt
• x suffix tree node (root of S-tree)
• nx prefix traversal order
• sizex number of descendants

24
ViST the Virtual Suffix Tree

• Same sequence algorithm as RIST
• BUT supports dynamic insertions
• Uses dynamic method to assign labels
• Once assigned, the labels are fixed and are not
  affected by subsequent data insertion or deletion
• Labeling the suffix tree w/o building it
• Relies on statistical information about the XML
  data

25
ViST the Virtual Suffix Tree

• Index structure contains the sequence
• Sequence to be inserted
• Dynamic scope of x ltnx, sizex,kxgt

26
ViST the Virtual Suffix Tree
27
Experimental Results

• Datasets used
• DBLP CS bibliography DB
• 289,627 records/publications
• Each publication tree of max depth 6
• Avg length of structure-encoded seq 31
• XMARK
• 1 record
• Complicated tree structure
• Synthetic

28
Experimental Results

• Comparison Methods
• Index Fabric Algorithm XML paths
• XISS uses nodes as basic query unit
• ViST appx. 1/10 of time to perform queries due
  to (multiple) join operations

29
Experimental Results - remove

• Index Structure and Size (1/3 less from suffix
  tree)
• DocId B Tree N elements
• Combined D-ancestor and S-ancestor B tree - N x
  L elements
• Index Construction

30
Conclusion

• XML Queries Subsequence Matching
• Advantages of ViST algorithm for subsequence
  matching
• Avoids expensive join operations
• Index on both content and structure of XML
  documents
• B trees supported by disk-based data
• Dynamic data insertion and deletion