README.md

BinEq API

This repository contains a simple Java API to access the BinEq benchmark.

The benchmark has to be downloaded separately into a folder, most API classes expect this folder as input. The core API is defined in a single class io.github.bineq.datasets.DataSet, it has a method to read records from an oracle in the benchmark as stream.

The various benchmark oracles are represented by classes io.github.bineq.datasets.EQ* (equivalence oracles) and io.github.bineq.datasets.NEQ* (non-equivalent oracles), respectively.

The project uses Maven to build, the command is mvn test.

Usage example:

Path benchmark = ...;  // the benchmark has been downloaded into this folder
new EQ() // or some other oracle in package `io.github.bineq.datasets` like NEQ1 etc
    .records(benchmark)
    .forEach (record -> {
        // Bytecodes caches bytes loaded from jars
        Bytecodes bytecodes = record.load(benchmark);
        byte[] bytes1 = bytecodes.bytes1();
        byte[] bytes2 = bytecodes.bytes2();
        
        // now compare bytecodes, e.g. check for byte-by-byte equality, whether certain hashes match, etc !
        // example:
        assert Arrays.equals(bytes1,bytes2);
    });

Records can also be filtered to create new oracles, used to study particular aspects of equivalence relations. The benchmark API contains a number of pre-defined oracles defined by filters. Custom filters can be easily created. For instance, the following will only iterate over records where different compilers were used to compile the two classes being compared (there is actually a built-in oracle EQDifferentCompiler for this purpose).

Path benchmark = ...;  // the benchmark has been downloaded into this folder
new EQ()
    .records(benchmark)
    .filter(record -> !Objects.equals(record.getCompiler_name_1(),record.getCompiler_name_2()))
    .forEach (record -> {
        ..
    });