Hasso-Plattner-Institut
Prof. Dr. Felix Naumann
 

Introduction

XML has become the defacto standard for electronic data exchange. Although XML is that popular, not all XML data is accompanied by an appropriate schema. The presence of a schema offers many advantages:

  • XML data can be queried faster, because advance query pruning and query rewriting can be employed.
  • XML data can be validated with respect to conforming to predefined formats. This is especially important when exchanging XML data between parties.
  • Automatic data integration often depends on the knowledge of the schema of the data sources.

In order to gain these advantages, it is useful to extract the schema of an XML document a posteriori automatically. While there are already some algorithms known that fulfill this requirement, we adapted and extended them in such a way that we are able to extract a common schema for a collection of documents instead of just one on the one hand and to be able to do this very efficiently, i.e., XStruct scales very well to input documents' sizes.

XStruct was implemented by Jan Hegewald in the context of a student research project.

XStruct

XStruct is our approach to automatic schema extraction and it is implemented it as a Java program.

Features

XStruct's most outstanding features are:

  • Extraction of general, complete, correct, minimal, and understandable XML Schemas.
  • Extraction of schemas from multiple documents.
  • Extraction of schemas from large documents (> 1 GB).
  • Detection of attributes of elements in the XML data.
  • Detection of datatypes of the XML elements and attributes.

Architecture

The architecture of XStruct can be seen in the following figure:

XStruct's architecture

While reading the XML documents, XStruct processes the data in three modules: the model extraction module infers the content model for each element, the attribute extraction module learns the attributes of each element, and the datatype recognition module determines the datatypes of the elements as well as of the attributes. The model extraction module afterwards passes its output to the factoring module, which consolidates all content models for one element into a common one. Finally the schema printer module takes the output of the factoring, the datatype recognition, and the attribute extraction module and outputs the XML Schema file.

Syntax

XStruct is deployed as a single .jar-file and can be used as described here.

XStruct offers two ways to configure the program. Either you can do it with a configuration property file or with command-line options. Depending on the alternative you prefer, there are two variants of call syntax.

  • Configuration with property file:
       XStruct -conf configurationfile [-output schemafile.xsd] inputfile1.xml inputfile2.xml ...
  • Configuration with command-line options:
       XStruct [options] [-output schemafile.xsd] inputfile1.xml inputfile2.xml ...
    where possible options are the following ones:
    • Threshold value:
      numeric value for threshold used in the inferring of element content models. Default: 2
    • MaxSavedValues value:
      numeric positive or zero value that determines how many different values for an object to store. Larger numbers increase precision in data type detection. Default: 15
    • MaxValuesEnumeration value:
      numeric positive value that determines how many different values may be used for an enumertation before setting a primitive datatype. Default: 10
    • LimitUnboundness value:
      numeric positive value that determines how often an object has to appear to set its maxOccurs to unbound. Default: 10
    • SchemaNamespace value:
      string that specifies which namespace prefix to use when constructing the XML Schema. Default: xsd

If you omit specifying an output file, the resulting schema will be printed onto the standard output.

Scalability

See the following figure to get an impression of XStruct's scalability:

Results of scalability experiments.

Downloads

You can download the paper XStruct: Efficient Schema Extraction from Multiple and Large XML Documents which more deeply describes our ideas, here.

The jar file: XStruct.jar

The manual: XStructManual.pdf

Terms of use

The software is free for academic purposes. We would very much appreciate a short note or feedback on he usage. For commercial use please contact Felix Naumann.

Requirements

In order to use XStruct, you need to have Java 5 installed on your system. Nothing else is required.

Valid XHTML 1.1Valid CSS!