Skip to content

Latest commit

 

History

History
158 lines (123 loc) · 5.99 KB

README.md

File metadata and controls

158 lines (123 loc) · 5.99 KB
Raw

reaper

Reaper is a tool used to assess a GitHub repository in the form of a score. It considers a number of different attributes in order to perform a thorough assessment.

Together with a database of metadata provided by the GHTorrent project, reaper considers both contextual information such as commit history as well as the contents of the repository itself.

Installation

The projects runs on systems with python3. There are a number of python libraries that the code needs in order to execute. To install them, simply run pip install -r requirements.txt (or pip3 if your system does not have python3 set as the default.)

Interface

The main interface that should be used to run reaper is the Python script called batch_score.py. This script should be called with a set of parameters that specify where the datasource can be found, what projects need to be analyzed, etc.

Additionally there is a script called score_repo.py, however at the moment it is outdated and cannot be used to score repos.

Usage

batch_score.py can be called as follows:

batch_score.py -c <config> -r <repos_path> -m <manifest> -s <sample_file>

Where:

  • <config>: Is an instance of config.json.
  • <repos_path>: Is the path to a directory where reaper can check out the source files of a project.
  • <manifest>: Is an instance of manifest.json (which can be found in this repository) containing information on what attributes should be executed.
  • <sample_file>: A list of GHTorrent project ids that should be analyzed, newline seperated.

config.json

This file is responsible for controlling various aspects of the system. There are two high level keys that can be altered, options and attributes.

options

Key Values Description
threshold Positive Numbers Defines the threshold by which the system considers a repository to contain a software project.
persistResult true or false Whether the granular results should be saved to the specified datasource.
datasource object Settings for connecting to the GHTorrent database, see description below.
github_tokens list List of GitHub OAuth tokens to be used for authentication for rate limiting purposes.
datasource

GHTorrent is a research project that aims to collect and store information produced by the public GitHub events feed. The initial implementation of score_repo.py relies on this information being accessible.

Load a dump of the GHTorrent data set into MySQL or MariaDB (binary compatible as of this writing), copy the config.json.sample file to config.json and edit the appropriate parameters under the options => datasource key.

peristResult

If persist results is enabled a database table needs to exist to which reaper can write results. This table should be named reaper_results and should contain at least a column for project ids named project_id, and a column to store the score for a repository named score. Additionally, there should be a column for every attribute that you want to store.

For instance, to create this table in MySQL the following table create statement can be used:

CREATE TABLE `reaper_results` (
  `project_id` int(11) NOT NULL,
  `architecture` double DEFAULT NULL,
  `community` double DEFAULT NULL,
  `continuous_integration` double DEFAULT NULL,
  `documentation` double DEFAULT NULL,
  `history` double DEFAULT NULL,
  `license` double DEFAULT NULL,
  `management` double DEFAULT NULL,
  `project_size` double DEFAULT NULL,
  `repository_size` double DEFAULT NULL,
  `state` varchar(255) DEFAULT NULL,
  `stars` double DEFAULT NULL,
  `unit_test` double DEFAULT NULL,
  `score` double DEFAULT NULL,
  PRIMARY KEY (`project_id`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8;

attributes

The system is designed as a number of plugins that all have a chance to analyze a given repository (both metadata as well as contents). A number of provided attributes are apart of the base distribution of the system, but more can easily be added.

In order for an attribute to be executed, it must be listed under the attributes key in the configuration file.

An example entry looks like the following:

{
  "name": "architecture",
  "dependencies": [
    "ctags"
  ],
  "enabled": true,
  "weight": 50,
  "options": {
  }
}

name refers to the name of the attribute as it appears under the attribute/ directory. dependencies is a list of system utilities that the attribute implementation relies on in order to function. enabled controls whether the attribute will be considered during the scoring of the repository. weight allows the bias of the attribute to be fine tuned in order to adjust its effect on the final score. Finally, options are specific options for each particular attribute implementation.

Attribute Development

In order to add your own attribute plugin to the system, there are few things that must be done. First, add an attribute entry as described in the above section that refers to your specific attribute.

Secondly, create the appropriately named directory under attributes/ along with a main.py. Inside of this, the following function signature should be used to kickoff the execution of the plugin:

def run(project_id, repo_path, cursor, **options):
  # Implementation goes here.

Check the doc block for details on what each parameter provides in terms of functionality. Attribute implementations should return a tuple of two values: the binary result of execution and the raw result of execution. The binary result should be True or False and the raw result should be a real number that is the raw calculation made by the plugin. In the case of purely binary results, do something like return result, int(result).

Additionally, there is the option of initializing the plugin. To take advantage of initialization, add the following function signature to main.py:

def init(cursor, **options):
  # Implementation goes here.