KNIFE - Java code analysis toolkit for dependency resolution

Simple source code analysis for dependency resolution. This is a bit of a foray into the world of Java compiler tools for me. It's also the first time I've really used Guice, so sorry if it's a bit messy.

Using KNIFE

KNIFE currently features a few simple analyses, which are called by:

    java -jar KNIFE.jar <analysis> [args] [file1 [file2 [...]]]

The analyses are either over a corpus of java files or over a maven artifact with its submodules.

Java analyses

reftree: Follow the reference tree

Picks out the -n most references classes and follows their dependencies recursively until all dependencies are enumerated. Can be used to find natural slicing points in large code trees.

Parameters

• -n How many of the most referenced classes should be used as a starting point. Default is 3.

refcount: Reference congestion

Displays a list of all classes used by the input files, ordered by how many times they are referenced, in ascending order.

Parameters

• -n Show which classes reference it for all classes that have less than or equal to n references. Default is 3.

importedby: Show which classes references a class

Finds out which classes references a certain class. Similar to Ctrl+Shift+G in Eclipse.

Parameters

• -i The referenced class to be analysed. Needs to be a qualified name.

imports: Global list of all imports

Gives a list of all the imports of all the files listed in the input.

Maven analyses

Maven analyses take a pom.xml as an input, and then analyses this and the maven artifacts given as modules of this pom (recursively).

longestpath: Calculate longest dependency chain

Walks through the artifact graph constructed from the POM and finds the longest chain of dependencies in the graph.

Parameters valid for all analyses

Parameters here can be given to any of the analyses.

• -x A package root to exclude from the list of input files. Can be used if you give a large input tree from find as input but want to slice away some part of the package structure, eg com.mypackage.impl.*

Building KNIFE

KNIFE is built with Maven. I use version 2.2.0 to build it.

Testing KNIFE

KNIFE uses data driven testing with JUnit4 and Yaml-specced files to easily create system tests. See src/test/resources/simple-test/spec.yaml for an example of a collection of spec tests.

To create a new test, you can either put it in one of the existing spec.yaml-files, which all operate on the files inside their directory, or create a new directory with its own spec.yaml- file. If you create a new directory, it must (for now) be listed in the file src/test/resources/spec-tests.

The format of the spec file is as follows:

description: Describe the test case
analysis: analysis to run
options: the command line after the analysis
expectFail: true/false if you expect the test to fail
expected: |
    a string of output you expect the
    test to yield, split across multiple lines.