Translate Graphical XML Query Language to SQLX

1
Translate Graphical XML Query Language to SQLX

• Wei Ni Tok Wang Ling
• Department of Computer Science
• University of Singapore
• niwei, lingtw_at_comp.nus.edu.sg

2
Roadmap

• ORA-SS and overview of our project
• GLASS query
• Translation from GLASS to SQLX
• What is SQLX
• Preprocessing
• Translation
• Conclusion and future works

3
1. ORA-SS

• ORA-SS 6 (Object-Relational-Attribute model for
  Semi-Structured data ) is a rich semantic model
  for semi-structured data
• Represents both the tree-like data structure and
  the relationship types contained in the data set
• Distinguishes relationship attributes from object
  attributes
• Uses Object ID indicating different object
  instances rather than element instances.
• It is different from ER diagram in that ORA-SS
  represent the hierarchical structure of the
  semi-structured data.

1. G. Dobbie, X. Y. Wu, T. W. Ling, M. L. Lee.
   ORA-SS An Object-Relationship-Attribute Model
   for Semistructured Data. TR21/00, Technical
   Report, Department of Computer Science, National
   University of Singapore, December 2000.

4
An XML example
lt!ELEMENT project (Jname?, member)gt
lt!ATTLIST project project_id ID REQUIREDgt
lt!ELEMENT Jname PCDATAgt lt!ELEMENT member
(Mname, age, job_title, publication,
qualification)gt lt!ATTLIST member
member_id CDATA IMPLIED gt lt!ELEMENT
Mname PCDATAgt lt!ELEMENT age PCDATAgt
lt!ELEMENT job_title PCDATAgt
lt!ELEMENT qualification PCDATAgt
lt!ATTLIST qualification degree CDATA IMPLIED

university CDATA IMPLIED
year
CDATA IMPLIEDgt lt!ELEMENT publication
(review)gt lt!ATTLIST publication
pub_id CDATA IMPLIED
title CDATA IMPLIEDgt
lt!ELEMENT review PCDATAgt
The DTD of the example dataset
Object Relations project (J, Jname)
member (M, Mname, age
qualification(degree,
university,
year)) publication (P, title,
(review)) Relationship Relations jm (J,
M, job_title) jmp (J, M, P)
An example ORA-SS schema
The ORDB storage schema of the example XML data
set
5
Overview of our project

• In our project, the XML data is stored in ORDB
  (e.g. Oracle 9i), where
• Relationship types and object classes are
  separately stored
• Relationship type attributes and object
  attributes are differentiated
• Multi-valued attributes and composite attributes
  are stored as nested tables.
• We choose SQLX in our translation rather than
  XQuery because SQLX supports SQL-in-XML-out in
  ORDB (e.g. Oracle 9i). Meanwhile, we also have
  translation to XQuery.

6
2. GLASS query

• GLASS 12 (Graphical query LAnguage for
  Semi-Structured data) is a high expressive
  graphical query language for semi-structured
  data, which is designed on the base of ORA-SS.
• Separates the complex query logic from the query
  graph (the query graph is the graphical part of a
  GLASS query).
• Considers relationship types in querying XML
  data.

1. W. Ni, T. W. Ling. GLASS A Graphical Query
   Language for Semi-Structured Data. DASFAA 2003.

7
2. GLASS query (Cont.)

• A typical GLASS query consists of four parts
• (1) Left Hand Side Graph (LHS graph) denotes
  the basic conditions of a query (which can be
  different from the structure of source schema,
  where we shall do view validation first 3)
• (2) Right Hand Side Graph (RHS graph) defines
  the output structure of the query result
• (3) Link Set specifies the bindings between the
  RHS graph and LHS graph. When two graph entities
  are linked, they are visually connected by a
  line, which means the data type and value of the
  entity in the RHS graph are from the
  corresponding linked entity in the LHS graph.
• (4) Condition Logic Window (CLW) It is an
  optional part where users write conditions and
  constructions that are difficult to draw, which
  includes Logic expressions, Mathematic
  expressions, Comparison expressions and IF-THEN
  statements.

8
2. GLASS query example
Original Schema
Example 1. Find the member whose age is less
than 35, and he either has taken part in less
than 5 projects or written more than 6
publications in some of the projects he attended
display the member id and name.
The hierarchical position of member and project
is swapped in this query.
The GLASS query of Example 1.
9
2. GLASS query example
Original Schema
Example 1. Find the member whose age is less
than 35, and he either has taken part in less
than 5 projects or written more than 6
publications in some of the projects he attended
display the member id and name.
The GLASS query of Example 1.
10
3. Translation from GLASS to SQLX

• What is SQLX 7
• SQLX (aka. SQL/XML) is an XML-related
  specification expanded on SQL.
• SQLX combines the features in both XML document
  processing and the traditional SQL
• Three fundamental SQLX functions
• XMLELEMENT
• XMLATTRIBUTES
• XMLAGG
• Why SQLX
• An extended standard on SQL, which is supported
  by Oracle, IBM, Microsoft, etc.
• It can be processed by ORDB systems.

1. Information technology -- Database languages --
   SQL -- Part 14 XML-Related Specifications.
   ISO/IEC 9075-142003

11
3. Translation from GLASS to SQLX Preprocessing

• Expansion of the simple projection
• Expansion of the abbreviated RHS graph
• Construction of the condition tree from LHS graph
  and CLW.
• In comparison with the condition tree in TQL
  13, our condition tree
• Contains the information of relationship types
• Can represent aggregation and restructuring (such
  as swap)
• Expresses logic in CLW.

1. Y. Papakonstantinou, M. Petropoulos and
   V.Vassalos. QURSED Querying and Reporting
   Semistructured Data. ACM SIGMOD 2002, Jun 4-6,
   Madison, Wisconsin, USA.

12
3. Translation from GLASS to SQLX
(Preprocessing) The Generation of
condition tree
Original Schema
Step 1. Initializing the condition tree.
The condition tree is initially a copy of the LHS
graph from the original GLASS query.
The CLW remains unchanged in this step.
The original GLASS query
The condition tree after Step 1.
13
3. Translation from GLASS to SQLX
(Preprocessing) The Generation of
condition tree
Step 2. Decomposing the LHS graph (making
duplicate nodes as necessary).
Link to the member in the RHS graph
The condition tree after Step 1.
The condition tree after Step 2.
14
3. Translation from GLASS to SQLX
(Preprocessing) The Generation of
condition tree
Step 3. Adding the logic expressions (in CLW)
into the condition tree.
If there are any existential and universal
quantifiers in CLW, they will be added also in
this step.
The condition tree after Step 3.
The AND node here is necessary to differentiate
the following two different query logic - A AND
B OR C - A AND (B OR C)
The condition tree after Step 2.
15
3. Translation from GLASS to SQLX
(Preprocessing) The Generation of
condition tree
Step 4. Eliminating universal quantifiers in the
condition tree. (not applied in this
example)

• Comments
• The condition tree is a combination of LHS
  graph and CLW, where we generate the WHERE
  clauses in the SQLX expressions directly by
  traversing the condition tree instead of the
  original LHS graph and/or CLW.

The final condition tree.
16
3. Translation from GLASS to SQLX Translate
Condition Tree and RHS graph into SQLX

• Traverse the expanded RHS graph in depth-first
  order and generate the select statements of the
  SQLX.
• Generate where clauses
• Render join for parent-child/ancestor-descendant
  relations (such relations on a XPath are
  performed by joining a series of tables in ORDB)
• Specify conditions for linked nodes by traversing
  the condition tree.

17
3. Translation from GLASS to SQLX Translate
Condition Tree and RHS graph into SQLX
Traverse the (expanded) RHS graph and obtain the
query block below
Preprocessed GLASS query for translation
The SQLX query expressions
SELECT XMLELEMENT(NAME member,
XMLATTRIBUTES (M1.M AS member_id)
XMLELEMENT (NAME Mname, M1.Mname) ) FROM
member M1 WHERE
For Each Node N in DF-Traversing the expanded
RHS graph CASE (N is the root of RHS) THEN
generate a block of SELECT XMLELEMENT
SELECT XMLELEMENT (NAME N ) FROM WHERE CASE
(N is an attribute in XML) THEN generate a block
of SELECT XMLATTRIBUTES SELECT
XMLATTRIBUTES ( AS N ) FROM WHERE CASE
(other kind of N) THEN generate a block of
SELECT XMLAGG(XMLELEMENT) //i.e. N is an element
but not the root SELECT XMLAGG
(XMLELEMENT (NAME N ) FROM WHERE)
18
3. Translation from GLASS to SQLX Translate
Condition Tree and RHS graph into SQLX
Traverse the (expanded) RHS graph and obtain the
query block below
The SQLX query expressions
SELECT XMLELEMENT(NAME member,
XMLATTRIBUTES (M1.M AS member_id)
XMLELEMENT (NAME Mname, M1.Mname) ) FROM
member M1 WHERE
M1.age lt35 AND ( M1.M IN (SELECT M FROM member
WHERE
(SELECT COUNT(J) FROM jm
WHERE M
jm.M)lt5) OR M1.M IN (SELECT M
FROM member
WHERE (SELECT J FROM project
WHERE(SELECT
COUNT(P) FROM jmp
WHERE jmp.J
project.J
AND jmp.M
member.M)gt6)) )
19
4. Conclusion future works

• We present the translation from GLASS to SQLX.
• GLASS supports SWAPPING and GROUPING on the base
  of the semantic information in ORA-SS.
• By translating into SQLX, GLASS can be processed
  on ORDB systems such as Oracle 9i.
• The translation method is simple. Compared with
  other XML-to-SQL translation works,
• The ORDB data storage preserves the semantic
  information in ORA-SS differentiates
  relationship type attributes from object
  attributes.
• GLASS query considers the relationship
  information in ORA-SS which are important to
  define the query semantic clearly.
• Our translation is simpler in comparison with the
  literature review in 8 and supports SWAPPING,
  GROUPING (also Quantifiers) in GLASS.

1. R. Krishnamurthy, R. Kaushik, and J. F. Naughton
   XML-to-SQL Query Translation Literature The
   State of the Art and Open Problems. University of
   Wisconsin-Madison

20
4. Conclusion future works

• Future works
• Query optimization on the base of ORA-SS
• Decrease the cost of join
• Cost model and optimized query plan
• Improving our case tool (so far partially
  developed)

21
References

1. S. Ceri, S.Comai, E. Damiani, P.Fraternali, S.
   Paraboschi, and L.Tanca. XML-GL a graphical
   language of querying and restructuring XML
   documents. In Proc. WWW8, Toronto, Canada, May
   1999.
2. S. Ceri, S. Comai, E. Damiani, P. Fraternali, and
   L. Tanca. Complex Queries in XML-GL. SAC(2)
   2000888-893.
3. Y. B. Chen, T. W. Ling, M. L. Lee. Designing
   Valid XML Views. 21st International Conference on
   Conceptual Modeling (ER'2002), pp 463-477,
   October 7-11, 2002, Tampere, Finland.
4. S. Cohen, Y. Kanza, Y. Kogan, W. Nutt, Y. Sagiv
   and A. Serebrenik. Equix Easy Querying in XML
   Databases. In proceedings of Webdb98 The Web
   and Database Workshop, 1998.
5. S. Comai, E. Damiani, P. Fraternali. Computing
   Graphical Queries over XML Data. ACM Transactions
   on Information Systems, Vol. 19, No. 4, October
   2001, Pages 371-430.
6. G. Dobbie, X. Y. Wu, T. W. Ling, M. L. Lee.
   ORA-SS An Object-Relationship-Attribute Model
   for Semistructured Data. TR21/00, Technical
   Report, Department of Computer Science, National
   University of Singapore, December 2000.
7. Information technology -- Database languages --
   SQL -- Part 14 XML-Related Specifications.
   ISO/IEC 9075-142003
8. R. Krishnamurthy, R. Kaushik, and J. F. Naughton
   XML-to-SQL Query Translation Literature The
   State of the Art and Open Problems. University of
   Wisconsin-Madison
9. B. Ludaescher, Y. Papakonstantinou, and P.
   Velikhov. Navigation-driven evaluation of virtual
   mediated views. In Proceedings of the sixth
   International Conference on Extending Database
   Technology (EDBT)(Konstanz, Germany, March),
   Lecture Notes in Computer Science, vol. 1777,
   Springer-Verlage, New York, 2000.
10. L. Mark, etc. XMLApe. College of Computing,
    Georgia Institue of Technology.
    http//www.cc.gatech.edu/projects/XMLApe/
11. K. D. Munroe, B. Ludaescher and Y.
    Papakonstantinou. Blended Browsing and Querying
    of XML in Lazy Mediator System. Konstanz,
    Germany, March 2000.
12. W. Ni, T. W. Ling. GLASS A Graphical Query
    Language for Semi-Structured Data. DASFAA 2003.
13. Y. Papakonstantinou, M. Petropoulos and
    V.Vassalos. QURSED Querying and Reporting
    Semistructured Data. ACM SIGMOD 2002, Jun 4-6,
    Madison, Wisconsin, USA.
14. XQuery 1.0 An XML Query Language. W3C Working
    Draft 22 August 2003 http//www.w3.org/TR/xquery/
15. XML Path Language (XPath) 2.0. W3C Working Draft
    22 August 2003 http//www.w3.org/TR/xpath20/
16. XML Schema. http//www.w3.org/XML/Schema

22
The End

• Thank you very much
• wish you have wonderful time in Beijing.