Tukwila Data Integration System

Welcome to the home of the Tukwila Data Integration System, hosted on Sourceforge. Tukwila is a research prototype XML processor and data integration engine developed at the University of Washington and the University of Pennsylvania, primarily by Prof. Zachary Ives, in conjunction with Profs. Alon Halevy and Daniel Weld.

Tukwila is the core query engine for the Sagres, Piazza, and BioMediator projects at the University of Washington, and it will also be at the heart of several emerging projects at the University of Pennsylvania.

We will be releasing Tukwila under the GNU Public License (GPL) in the very near future. Please stay tuned...

What Does Tukwila Do?

Tukwila is an XQuery processor for network-bound data: it accepts queries in a subset of the XQuery language, fetches XML data from URI sources specified in the query, and performs the XQuery over them. Tukwila has many unique features:

• In contrast to almost all other XQuery engines, Tukwila executes queries in a streaming fashion: it begins producing answers incrementally as it is still reading data from the source(s).
• Unlike many XML processors, Tukwila can support joins and grouping operations. It has the most common SQL-style query operations built in.
• Tukwila includes an advanced query optimizer, as well as state-of-the-art adaptive query processing features that allow it to incrementally re-optimize queries in the middle of execution. See Zack Ives' research papers for more information. You can also find a description of an earlier version of the Tukwila project at this site.

Using Tukwila

This version of Tukwila does not include a query reformulation module, though one exists from the Piazza project and might be ported to this framework.

Tukwila is divided into two major components: (1) the query execution engine, which runs as a server in the background, and (2) the front-end, which is a Java-based GUI and a Java JAR file that can be linked with your code.

Linux Execution Engine

Coming soon.

Windows Execution Engine

Decompress Tukwila into a new directory. Run Tukwila.exe from the same directory, which will invoke the Windows execution engine.

GUI

The GUI can be run with:

./TukwilaGUI [{hostname}]

where {hostname} defaults to localhost and is the name of the machine running the execution engine.

To test your connection, open Maya1.xq or GS.xq and run it.

Java API

The API is documented in InvokeXQuery.html.

Web Status Monitoring

Go to http://{engine hostname}:7777 to see the system status.

Source Code Internal Documentation

See internals.html for a discussion of the internal class hierarchy structure and source code for Tukwila.