start off with one hole and fill that with the assumption that it will pass all tests.
if one candidate passes n tests and fails m tests, keep the program corresponding to n tests together (along with branch condition) and recursively apply this procedure.
Then apply the merge procedure as in the paper.
This method will eliminate individual cost for generating candidates for each test.
n
tests and failsm
tests, keep the program corresponding ton
tests together (along with branch condition) and recursively apply this procedure.This method will eliminate individual cost for generating candidates for each test.