TL;DR

git clone https://github.com/aas-integration/clusterer.git
cd clusterer
gradle jar
java -jar build/libs/clusterer.jar -dirs build/classes/main

Then look at the produced json files to see the clustering.

Usage

The jar takes the following args:

 -cp VAL        : Classpath (optional)
 -cs N          : Clustering strategy between 1 and 4. Default is 3. (default:
                  3)
 -dirs STRING[] : List of directories of classdirs to be passed to soot.
 -out VAL       : Output file name. Default: clusters.json (default:
                  clusters.json)

Output

The tool currently produces several json files. Each json file clusters the class names found in the subfolders. Class names are clustered if we consider the classes to be similar. The notion of similarity is different for each produced json file. All strategies start by splitting the last part of the class name into words based on its camel casing. Then it stems them and sorts them alphabetically.

strategy 1

Takes the list of stored and stemmed words and concatenates them to a string using semicolon as a separator. This string used used as a hash key to add the class name to a dictionary: string -> list<string> .

strategy 2

Takes the list of stored and stemmed words for the current class and the super class, removes all words that occur in the super class list from the current list and then concatenates the remainder like strategy 1.

strategy 3

Takes the list of stored and stemmed words for the current class and the super class, and uses the intersection of both. If the intersection is empty, it uses the original list. Then concatenates the remainder like strategy 1.

strategy 4

Same as 3, but for each work in the stored and stemmed words, we look up in a thesaurus for the list of synonyms and replace the word by its lexicographically smallest, stemmed, synonym. Finally, we sort the list again.