Code Completion Recommender - ASE19 Project

This repository is a group project at the UZH for the Advanced Software Engineering course. The goal is to create a code recommender system.

Our teams recommender is based on the following paper: https://doi.org/10.1109/SANER.2016.80

Setup / Installation

The project consists of two maven projects in the folders evaluation and src respectively. The first project is the evaluation project which runs an evaluation on the RACK recommender using the second project as a dependency. The second project is the RACK recommender implementation usable as a library.

If you want to run the evaluation you have to run mvn install on the library project, to install in your local repository, so that the evaluation project can access it as a dependency.

Maven Repository

If you just want to consume the library, then you can use our package at bintray. The bintray repository has a big blue set me up button at the top right which explains how to use the package as a client.

Using the library

Properties

Before using the library, you need a Java.util.Properties object. You can read this from a file or build it in code, whatever is better for you. The following properties can be set:

Property	Value / Explanation	Required
generate-model	true or false (false if not set) if this is set to true, then a new file in database-file is generated	✗
context-path	path to folder with the context data from KaVE	if generate-model = true
apis	comma separated list of assembly names to generate the model from	if generate-model = true
delta	int value (check paper for meaning)	✔
lamda	double value (check paper for meaning)	✔
database-file	path to sqlite db file (generated model)	✔
use-KKC	true or false (false if not set) check paper for KKC definition	✗

Creating the Recommender

After you created the properties object, you can instantiate the recommender like this:

//props is your property object
KaveContextModule module = new KaveContextModule(props);

//create the recommender
DefaultRecommender rec = new DefaultRecommender(module);

Depending on the properties set, this will create your model or just use the one provided with the database file.

Using the Recommender

After instantiating the recommender, you can use it like this:

//the recommendation needs a KaVE context to work with
Set<Pair<IMemberName, Double>> reccommenderResult = rec.query(context);