Mining for Program Structure

We present a new mining algorithm for discovering program structure built on a warehouse of program-analysis data. We make use of statement definition-use data along with variable typing information to learn about the structure of legacy code. Using concept analysis, we develop an interactive knowledge discovery framework into which different types of analysis data can be integrated. This work is the first in a series of data warehousing experiments that is designed to tap the wealth of information hidden in existing program-analysis data.