Data related to clone set cluster process (section 2 in paper)

The following are data related to the steps outlined in our paper entitled An Information Retrieval Process to Aid in the Analysis of Code Clones for the clustering of clone sets. The Windows Research Kernel 1.0 source code is only available for non-commercial use in academia. For more information about how to get the Windows Research Kernel 1.0 source code, please see the Windows Research Kernel web site at http://www.microsoft.com/resources/sharedsource/Licensing/researchkernel.mspx. Because of this limitation, every effort is made to remove any source code related information from the data below. If you have received approval to access the source code, we can provide you with more detailed data. For this, please contact Robert Tairas at tairasr@cis.uab.edu.

2.1 Clone Detection (Step 1)

The results of the clone detection tool CCFinder is given below. The textual format is the format used in the subsequent steps:

• Textual format - File used to obtain clones (file names removed from the results)
• Grapical format - For information purposes only, a different format for the output

2.2 Clone Set Filtering (Step 2)

The following contains the clone set totals for the directories in the NT Kernel.

• Clone set totals

2.3 Source Code XML Representation (Step 3)

N/A

2.4 Generating the Term-Document Matrix (Step 4)

The following are the Term-Document matrices for each directory that will be clustered. The format of files representing the matrices is each line contains a row index, column index, and total occurences of the term.

• Memory management
• Registry configuration
• Process/thread support
• Security functions
• I/O management
• All five directories

2.5 Computing the SVD of the Matrices (Step 5)

The following are the lower dimensionality matrices for each directory that will be clustered. These matrices, ie. Xhihat, contains the format of original the Xhihat matrix transposed, as Cluto works on the rows as vectors instead of the columns. The first row contains the matrix dimensions.

• Memory management (Rank: 15)
• Registry configuration (Rank: 15)
• Process/thread support (Rank: 10)
• Security functions (Rank: 10)
• I/O management (Rank: 10)
• All five directories (too big: 59MB file)

2.6 Clustering the Clone Sets (Step 6)

The following are the clustering assignments from Cluto for each of the vectors (i.e., clone sets) in the matrices from the previous step. This mapping is a simple text file containing a cluster number in each line where each line represents a clone set. Also included are the Cluto informational output report for each clustering.

• Memory management (Number of clusters: 30) Output Report
• Registry configuration (Number of clusters: 30) Output Report
• Process/thread support (Number of clusters: 20) Output Report
• Security functions (Number of clusters: 20) Output Report
• I/O management (Number of clusters: 20) Output Report
• All five directories (Number of clusters: 30) Output Report
• All five directories (Number of clusters: 50) Output Report

The following are the mapping of the clustering assignments above with the clone set ID for each directory.

• Clone set to cluster mapping

2.7 Generating Cluster Reportsd for Analysis (Step 7)

The following is a sample cluster report in HTML. The references to the file names and source code have been removed. Various information about the cluster is available, such as the clone ranges and their membership to the clone sets. These listings of clones are ordered both by clone set and by their location in the source code. In the middle of these listings are the files that contain the clones, and the clone sets listed in "Similar Clone Sets" section, which takes into account the two types of clone sets described at the beginning of Section 4 (i.e., linked clone sets and sliding clones).

• Example cluster report for SE-11 (cluster 11 of the security functions diretcory)

This project is supported by NSF grant CPA-0702764.