Foofah [1][2] is a programming-by-example data transformation program synthesis system. It is able to generate a data transformation program defined in Professor Joe Hellerstein's Potter's Wheel paper [3] using an input-output example from the end user.

• Linux
• Python 2.7
• g++
• Boost.Python (Mac, Linux)
• setuptools

  $ python -m pip install -U pip setuptools

In fact, other Python modules numpy, tabulate, cherrypy, editdistance, python-Levenshtein , matplotlib are also required. But they could be installed using setuptools in next section.

Build Foofah container

$ docker build -t foofah .

Run Foofah contrainer

$ docker run -p 8080:8080 foofah

Foofah web service will be available at localhost:8080.

$ cd foofah
$ python setup.py install

To test Foofah against individual test case from the console:

$ cd foofah
$ python foofah.py --input <test_file>

Note that each test case must be a json file that contains one json object with two members, InputTable and OutputTable, both of which are 2d array of strings, representing the user-provided input-output example.

• Link to an example test case.
• Link to all benchmark test cases used in our full paper.

To learn other command-line argument options:

$ python foofah.py --help

To interact with Foofah through a web interface (as shown in video):

$ python foofah_server.py

By default, the service will be available at localhost:8080.

Foofah is being developed in the University of Michigan. This work in part supported by National Science Foundation grants IIS-1250880, IIS-1054913, NSF IGERT grant 0903629, a Sloan Research Fellowship and a CSE Department Fellowship

[1] "Foofah: Transforming Data By Example", SIGMOD 17', Zhongjun Jin, Michael R. Anderson, Michael Cafarella, H. V. Jagadish

[2] "Foofah: A Programming-By-Example System for Synthesizing Data Transformation Programs", SIGMOD 17', Demo, Zhongjun Jin, Michael R. Anderson, Michael Cafarella, H. V. Jagadish

[3] "Potter's wheel: An interactive data cleaning system." VLDB. Vol. 1. 2001. Raman, Vijayshankar, and Joseph M. Hellerstein.