Skip to content

Latest commit

 

History

History
47 lines (36 loc) · 3.08 KB

ARCHITECTURE.md

File metadata and controls

47 lines (36 loc) · 3.08 KB
Raw

Architecture

This document contains a high-level architectural overview of SynTest, and serves as a good place to get to know the codebase.

The SynTest Core repository serves as common core to automatically generate test cases based on JavaScript. It is important to understand that this repository should not be used directly, but that its architecture defines the underlying structure of extensions based on a language (e.g., Solidity and JavaScript/Typescript).

Code organization

Below is a brief view of SynTest's structure:

  • src/criterion: functions to calculate fitness based on (e.g. branch coverage or function coverage).
  • src/graph: representation of control flow graph (CFG).
  • src/search: implementation of search algorithms.
    • budget: classes to keep track of the available budget using the budget manager.
    • comparators: functions that allow the comparison of two or more testcases.
    • factories: factory functions to instantiate objects based on the configuration file.
    • metaheuristics: classes of search algorithms, e.g. RandomSearch, NSGAII and DynaMOSA.
    • objective: functions to calculate fitness values for TestCases using for example branch distance.
    • operators: mutation operators used to traverse the search space.
  • src/statistics: keeps track of test statistics.
  • src/testcase: representation of test cases.
    • decoder: decodes encoded test case to concrete test case.
    • execution: runs concrete test case.
    • sampling: functions that allow sampling of various statements from test cases based on types.
    • statements: action statements (constructor/function call) or primitive statements (bool/numeric assignment).
  • src/util: several util files, which include helper classes and the logger.

Understanding internals

To get a better idea of how the SynTest Core operates, let's take a look at the flow of generating a test case and executing it. When testing for a specific target, this can be separated into the following steps:

  1. The search algorithm with objective manager is initialized with the user supplied configuration.
  2. The budget mangers are initialized responsible for iterations, evaluations etc.
  3. The algorithm searches given the budgets.
  4. Statistics are collected regarding the budgets and coverage.
  5. Finally, the statistics are logged and the execution terminates.

Search algorithms

There are several search algorithms available, each relying on a different objective manager. These are shown in the table below.

Search algorithm Objective Manager
RandomSearch SimpleObjectiveManager
NSGAII [1] SimpleObjectiveManager
MOSA [2] UncoveredObjectiveManager
DynaMOSA [3] StructuralObjectiveManager