Database-Supported XML Processors

(1)

Database-Supported

XML Processors

Prof. Dr. Torsten Grust

[email protected]

(2)

Part I

(3)

Outline of this part

(4)

A Word About Myself

Torsten Grust

Originally from Hannover

1989–1994 Student of Computer Science @ TU Clausthal 1994–2004 Database Research @ U Konstanz

1999 Promotion

2000 Visiting Scientist @ IBM, DB2 Everyplace 2004 Habilitation

2004-2005 Professor @ TU Clausthal 2005-2008 Professor @ TU M¨unchen since 9/2008 Professor @ U T¨ubingen

(5)

Welcome to this Course . . .

We will userelational database technology to develop a highly efficient, scalable processor for XML languages like XPath, XQuery, and XML Schema.

This means that

1 _{you will get to know these XML technologies quite well, and}

2 _{you can apply and deepen your (rusty?) knowledge of RDBMSs in a}

(6)

Relational XML Processing

XML Processors≡Tree Processors

⇒This is a course on Relational Tree Processors. Relational Tree Encoding E

• •kkkk • • • S S S S •ww •ww GG• • G G • E → · · ·

Map tree queries into relational queries over tree encodings: Tree tree query //

E

Tree

Rel

relational query //Rel E−1

(7)

Compiling XQuery to Relational Algebra (1)

Input: XQuery Expression

Query against an Internet auction database (think eBay): How many auction items are listed in each of the site’s [geographical] regions?

for $r in doc(”auction.xml”)/site/regions/* return count($r//item)

Tree query: Note how this query usestree navigation operators/

(read: child) and// (descendant) to explore the input XML documentauction.xml.

(8)

(9)

Compiling XQuery to Relational Algebra (3)

Output: Relational Algebra (MonetDB’s Dialect)

1 . . .

2 a0000 := a0004.reverse ().sort ().reverse ();

3 a0000 := a0000.CTrefine (a0003);

4 a0000 := a0000.CTrefine (a0002);

5 a0000 := a0000.mark (0@0).reverse ();

6 a0001 := a0000.leftjoin (a0002);

7 a0005 := a0000.leftjoin (a0004);

8 a0006 := a0000.leftjoin (a0003);

9 . . .

10 a0003 := count(a0004.reverse ());

11 a0007 := a0003.reverse ().mark (0@0).reverse ();

12 a0008 := a0003.mark (0@0).reverse ();

13 . . .

(10)

Pathfinder

For about 61_/₂_{years now, work is underway to design and build the}

purely relationalXQuery processor Pathfinder. Joint work with a couple of brilliant guys from

Pathfinder generates an internal algebraic representation of XQuery expressions and then emits

1 _{MIL code}_{for consumption by MonetDB/XQuery, or}

2 _{SQL:1999 code}_{to be executed by off-the-shelf RDBMS,} _e.g._,

(11)

Pathfinder & IBM DB2 vs. 110+ MB of XML

(12)

Hands On!

In a sense, this course is an in-depth tour of the techniques and concepts behindPathfinder.

Because Pathfinderhas been under development since 2002, the system is already usable and provides an ideal playgroundfor us.

Available under the Mozilla OSS License

www.pathfinder-xquery.org www.monetdb-xquery.org

Source code and installers for Unix (Linux, Mac OS X), Windows.

(13)

Organisatorisches

Termine

Zeit Ort

Vorlesung Do,13:15–14:45 Sand 6/7, kleiner H¨orsaal ¨

Ubung Di, 13:15–14:45 Sand 6/7, kleiner H¨orsaal (Jan Rittinger)

Homepage + Material zur Vorlesung

www-db.informatik.uni-tuebingen.de/teaching/ws0809/dbxml Folien [PDF] zum Download verf¨ugbar (ca. einen Tag vor Termin).

(16)

Wie profitiert man von dieser Vorlesung?

¨

Ubungsaufgaben und Klausuraufgaben werden sichsehr ¨ahneln.

Aktiv dabei sein!

¨

Ubungen starten n¨achsten Dienstag (28. Oktober) Beispiele nachvollziehen und eigene Experimente starten:

Michael Kay’sSaxon (www.saxonica.com)

Pathfinder

Klausur/mdl. Kolloq zum Ende des Semesters bestehen. “Sprechstunde” nutzen

Fast immer, wenn die T¨uren zu unseren B¨uros (Sand 13, B312 und B318) offen stehen. Effektiv sind das 90 % unserer Anwesenheitszeiten.

(17)

Questions?

Questions . . . ? Comments . . . ? Suggestions . . . ?

Database-Supported XML Processors