Refazer

Refazer (rebuild, in Portuguese) is technique for automatically learning program transformations. It builds on the observation that code edits performed by developers can be used as input-output examples for learning program transformations. Example edits may share the same structure but involve different variables and subexpressions, which must be generalized in a transformation at the right level of abstraction.

Usage

Let's say that we want to learn a transformation that adds a missing call to a function. The following code illustrates how we can invoke Refazer passing an input-output example and Refazer learns a list o transformations that satisfy the example.

var input = 
@"def product(n, term):
  return term(n)*product(n-1)";
var output = 
@"def product(n, term):
  if n == 1: 
    return term(1)
  return term(n)*product(n-1)";
        
var examples = new List<Tuple<string, string>>() { Tuple.Create(input, output) };
var refazer = new Refazer4Python();
var transformations = refazer.LearnTransformations(examples);

After learning the transformations, we can get the top learned transformation and applied it to a program to add the base case.

if (!transformations.IsNullOrEmpty()) {
  var result = refazer.Apply(transformations.First(), input);
  if (!result.IsNullOrEmpty()) {
    //print the transformed program
    Console.Out.WriteLine(result.First());    
  } 
}