Soft Cardinality Constraints on XML Data

How Exceptions Prove the Business Rule

Emir Muñoz Fujitsu Ireland Ltd.

Joint work with F. Ferrarotti, S. Hartmann, S. Link, M. Marin

@ Nanjing, China, 14th October 2013

Contribution

• Introduce the definition of soft cardinality constraints over XML data.

• Efficient low-degree polynomial time decision algorithm for the implication problem.

• Empirical evaluation of soft cardinality constraints on real XML data.

Emir M. - WISE, Nanjing, China, 14th October 2013 2

Outline

1. Introduction

2. Soft Cardinality Constraints

3. The Implication Problem

4. Performance Evaluation

5. Conclusion

Emir M. - WISE, Nanjing, China, 14th October 2013 3

Introduction Concepts

• Cardinality constraints:

– Capture information about the frequency with which certain data items occur in particular context.

• Soft cardinality constraints:

– Constraints which need to be satisfied on average only, and thus permit violations in a controlled manner.

Emir M. - WISE, Nanjing, China, 14th October 2013 4

Introduction Example (1/2)

Emir M. - WISE, Nanjing, China, 14th October 2013 5

Project within a research institute

support research

• Some cardinality constraints:

– Every scientist is a member of 2, 3, or 4 research teams.

– Every technician can work in up to 4 different support teams.

– A project cannot have more than one manager.

– In every team, there should be two employees for each expertise level.

Emir M. - WISE, Nanjing, China, 14th October 2013 6

Introduction Example (2/2)

• Some cardinality constraints:

– Every scientist is a member of 2, 3, or 4 research teams.

– Every technician can work in up to 4 different support teams.

– A project cannot have more than one manager.

– In every team, there should be two employees for each expertise level.

Emir M. - WISE, Nanjing, China, 14th October 2013 7

Introduction Example (2/2)

Probably will be exceptions

Scientist working in 5 research teams or more

Soft constraints

Soft Cardinality Constraints Definition

• Expressiveness from the ability to specify soft upper bounds (soft-max) as well as soft lower bounds (soft-min) on the number of nodes.

• soft-card(Q, (Q´, {Q1,…, Qk})) = (soft-min, soft-max)

• With some sources of intractability

Emir M. - WISE, Nanjing, China, 14th October 2013 8

Context path Target path

Field paths

soft-min = 1

• Every scientist is a member of 2, 3, or 4 research teams. – soft-card(ε, (_.RTeam.Sci, {id})) = (2, 4)

• Every technician can work in up to 4 different support teams. – soft-card(ε, (_.STeam.Tech, {id})) = (1, 4)

• A project cannot have more than one manager. – soft-card(_, (Manager, Ø)) = (1, 1)

• In every team, there should be two employees for each expertise level. – soft-card(_._, (_, {Expertise.S})) = (2, 2)

Emir M. - WISE, Nanjing, China, 14th October 2013 9

Soft Cardinality Constraints Examples

The Implication Problem Definition and Algorithm

• Let be a finite set of (soft) constraints.

• We say that finitely implies , denoted by if every finite XML T that satisfies all also satisfies

Emir M. - WISE, Nanjing, China, 14th October 2013 10

Performance Evaluation Configuration

• We compare the performance against XML Keys

• Machine Intel Core i7 2.8GHz, with 4G RAM

• Documents: – 321gone, yahoo (auction data)

– dblp (bibliographic information on CS)

– nasa (astronomical data)

– SigmodRecord (articles from SIGMOD Record)

– mondial (world geographic db)

Emir M. - WISE, Nanjing, China, 14th October 2013 11

Performance Evaluation Results

Expressivity

Time

Emir M. - WISE, Nanjing, China, 14th October 2013 12

In comparison with previous XML keys

Conclusion

• We introduced an expressive class of soft cardinality constraints, sufficiently flexible to boost XML applications such as data exchange and integration.

• Slight extensions result in the intractability of the associated implication problem.

• We give an axiomatization for this new class.

• Present an empirical performance test that indicate its efficient application in real use cases.

Emir M. - WISE, Nanjing, China, 14th October 2013 13

Discussion

• Questions & Answers

– Soft Cardinality Constraints on XML Data

THANKS! Emir Muñoz emir@emunoz.org

Emir M. - WISE, Nanjing, China, 14th October 2013 14