davidbakereffendi/graplhook4j

A Java driver for the GraPL project to provide an interface for connecting and writing to various graph databases based on the code-property graph schema.

Hi there 👋

I'm a PhD student @ Stellenbosch University and Directory of R&D @ Whirly Labs

• 🔭 I’m currently working on Joern and related static analysis research
• 👯 I’m looking to collaborate on code feature mining, partial program analysis, and accelerating program analysis with CPGs
• 📫 How to reach me: [email protected]
• 😄 Pronouns: he/him

graplhook4j's Issues

Create TigerGraphHook

Implement the IHook interface for TigerGraph

Update CPG Schema

Some of the latest changes on ShiftLeft's repository

• JumpTarget introduced
• A few properties added to vertices such as MetaData

Add basic documentation on the README for GraPLHook4j

As the project gains basic capability, a README on how users can get started to use GraPLHook4j.

Create API for Control Structure Vertices

Currently, control structures are built with generic block vertices which are not correct. This means that the driver needs to create a control structure vertex creation and linking.

Ability to obtain the largest AST order from a graph

Due to the nature of j2GraPL's new controller structure, the order is persisted across multiple files in order to keep the order unique. When a graph is saved and then deserialized later to add to, the current order needs to be set in the controller.

The hook should provide a method to obtain the latest order of the graph.

Re-order AST nodes when subgraph removed

If a file or method changes and the old file or method is removed, the order property will no longer make sense. If [x, y] nodes are removed, z needs to take on the value of x and so on. After this order is resolved, the removed file or method can once again be added.

Create schema for TigerGraph

Before we can implement a hook for TigerGraph, we need to create a schema. This can be implemented once AST jump statements are handled.

Use Remote Gremlin connection instead of JanusGraph Client

The JanusGraph client is currently what is used to connect to a JanusGraph instance which allows for the full JanusGraph API but this has more cons than pros for GraPL. The suggested fix is to migrate the current JanusGraph hook to use the remote connection API which is more graph database neutral but means that testing and implementation needs to be refactored.

This does, unfortunately, mean that we lose the ability to serialize and deserialize the graph database outside of the backend server.

Improve Assign Block Efficiency By Starting at Smaller Seeds

Right now the efficient variant of assignBlock starts at the method label but it stores full path which means that we can start at the smallest seed which is the namespaces and travel down those to the method.

NamespaceBlock resolution issues

NamespaceBlock resolution with FileVertex has come up in tests with j2GraPL. This has also exposed an ID duplication when serializing the graph to GraphML.

Migrate to Gradle

Following the changes in j2GraPL, this repository should also move to Gradle.

Basic control vertex to block/method join functionality

Add the ability to project jump statements and condition vertices to the graph.

Handle JanusGraph Transaction failures better

• Using JanusGraph + Scylla + ES has occasionally thrown a Lock contention exception.
• More details to handle transaction failures are here: https://docs.janusgraph.org/basics/transactions/

Test compatibility with other versions of Java

GraPLHook4j's minimum version is Java 8 and the CI/CD only tests on this. To be thorough it would be useful to have Travis test up until at least JDK11 which, at the time of writing, is the latest LTS version of Java still supported.

Be able to check if there is a block with a given order

When using assignToBlock there is no way to know if there is a block with that order in the graph database so we may have to use try-catch blocks which is a bit inefficient. There should be a method to check if there is a block with an associated order.

Add formal CONTRIBUTING to GraPLHook4j

Until now, an informal way of contributing and prefixing commit messages has been used. In order to keep the commit history clean a more formal CONTRIBUTING file should be added.