Commandline Parameters

Commandline parameters for DiffSearch

Introduction

To simplify the execution, use the command

dfs() { mvn exec:java -Dexec.mainClass=research.diffsearch.main.App -Dexec.args="$1" ; }

Then, DiffSearch can be run using

dfs "args"

Setup

To create an index, perform the following steps:

1. Clone all repositories using `dfs "-clone ". The path must be a path to a text file with all links to the GitHub repositories (one link per line).
2. Parse a corpus of code changes using dfs "-d -lang <language>", where language is Java, JavaScript, or Python.
3. Extract feature vectors and create an index using dfs "-fe -lang <language>". Further below are additional parameters given that modify how features are extracted.
4. DiffSearch can now accept queries.

DiffSearch Modes

Parameter	Usage	Description
n	-n	DiffSearch in console mode. Queries can be entered in the console, where also the results are shown.
g	-g	DiffSearch serves as webserver for the DiffSearch UI
w	-w	DiffSearch serves as webserver for the old DiffSearch UI (deprecated)
q	-q query	performs a search on the given query
b	-b inputpath outputpath	processes all queries of a text file at the given path and saves the result in the given output file.
fe	-fe	Feature extraction mode, creates feature vectors from code changes and indexes them
clone	-clone repository-list	Clones the list of git repositories. The parameter should be a path to a txt files with all links to GitHub repositories.
d	-d	Extracts and parses code changes of the cloned git repositories.

Other parameters

Parameter	Usage	Description
pyc	-pyc pycommand	sets the command to start the python environment. For windows systems, the windows subsystem for linux is required. With WSL this parameter usually should be "wsl python3".
lang	-lang (java|javascript|python)	sets the target programming language
l	-l	saves the server log to a log file. (WIP)
oj	-oj	runs only the java server and does not invoke python. Only matters in "-fe", "-q", "-n", "-w", "-g", and "-b" mode.
p	-p port	sets the port of the web interface
r	-r	enables recall measurement. This has requires a lot of time.
silent	--silent	if given, DiffSearch omits large console outputs, like results of a query.
py_port	-py_port port	sets the port of the python server
k	-k number	sets value for k, which is the number of candidate changes.
vl	-vl number	size a each partition of the feature vectors.
cb	-cb number	number of count bits (Default 1)
t	-t number	number of threads to use
extractors	-extractors (name(:length)?;)+	defines the extractors DiffSearch uses, e.g. -extractors parentchild:2000;triangle:2000. Valid extractors are node, triangle, parentchild, sibling, rulecount, editscript.
mt	-mt seconds	maximum matching time for a single code change
extract-query-placeholders	--extract-query-placeholders	extract query placeholders like EXPR, default is false
tfidf	-tfidf	tfidf weights are used in the feature vectors
noquerymultiplication	-noquerymultiplication	query vectors do not get multiplied
nondividedextraction	-nondividedextraction	feature extraction is not divided in the old and new part