MintHint is available under the Apache License, version 2.0. Please see the
LICENSE file for details.


This directory contains the prototype implementation of MintHint. MintHint is a
novel program repair tool. It generates repair hints for the developer, as
opposed to finding a complete repair. See our ICSE'14 paper for details.

Requirements (This prototype was tested with the following environment)
* MATLAB R2012a
* Java 1.7
* Antlr 3.5.2 - download the complete jar from 
  Note: this prototype is not compatible with antlr 4.
* Guava 17.0 - download jar from
* A Linux environment that supports bash scripting.

Run MintHint
This prototype implements the statistical analysis and hint generation part of
MintHint. Fault Localization and derivation of the state transformer from the
test suite is to be done offline. This prototype generates the repair hints for
a single statement. It requires the following input files:
	- a csv file recording the data obtained from state transformer. In
	  other words, this file records the values of expressions from the
	  repair space (including the LHS variable). Each column is an
	  expression and each row is a run of the program.
	- attrmap:- list of expressions in the repair space. Each line in this file is a
	  candidate expression from the repair space. Note that this list must
	  also include subexpressions from the faulty RHS as well as the faulty
	  RHS itself. (See example/attrmap.75). Each line in this file is
	  formatted as "linenumber: expression".
	- rhsfile: list of subexpressions in the faulty RHS. This list needs to record
	  the line numbers of these expressions in the attrmap file. Make sure to
	  put the complete faulty RHS as the first expression in this file.
	- typefile: type declarations for all variables used in the repair
	  space. (See example/type.75). All declarations needs to end in ';'

Let us run the example program in "example" folder. You can find the following
files in this folder:
	- tcas.c : the source code of the program. Line 75 has the fault.
	- v75.complete.csv: the csv file
	- attrmap.75: expressions in the repair space
	- rhsfile.75: subexpressions in the faulty RHS
	- type.75: type declarations of all variables used in the repair space

Step 1: Run MATLAB from the minthint-src directory. Add the subdirectory
"matlab" to the MATLAB path.

Step 2: Put antlr-3.5.2-complete.jar, guava-17.0.jar and minthint.jar to your
classpath. You can find minthint.jar in the subdirectory "java".

Step 3: Type the following at the matlab prompt.
	minthint('example/v75.complete.csv', 1, 'example/attrmap.75', 'example/rhsfile.75', 'example/type.75');
The parameters of this call are:
	- the csv file
	- the line number of the LHS variable in the attrmap file
	- the attrmap file
	- the subexpressions of faulty RHS
	- the type declarations file

Step 4: You can find the generated hints in the file "final.hints.file" in the
top-level directory. Each line in this file records a hint and its associated

This prototype uses MATLAB for the statistical analysis and Java+Antlr for edit
distance computation. The MATLAB sources can be found in "matlab" folder. The
java sources are in "java/minthint-src-java.jar". Antlr is used to generate
syntax trees of expressions and use tree edit distance for the edit distance
computation. (See Reference [41] in the ICSE paper for details on algorithm used
for tree edit distance). Currently, the expressions supported are over char,
bool and int variable/arrays. Modify the antlr grammar to support more
datatypes. There are some helper shell scripts in "shell-scripts" folder. These
were used as a way of data flow between the matlab and java code.

There could obviously be bugs in this prototype. Feel free to contact in this case.