def which_fix_goes_first(program, fix1, fix2):
try:
fix1_location = extract_line_number(' '.join(fix1.split()[1:]))
fix2_location = extract_line_number(' '.join(fix2.split()[1:]))
except Exception:
#print fix1
#print fix2
raise
if not fix_ids_are_in_program(recompose_program(get_lines(program)[fix2_location:]), fix2) and fix_ids_are_in_program(recompose_program(get_lines(program)[fix1_location:]), fix1):
return fix1
if not fix_ids_are_in_program(recompose_program(get_lines(program)[fix1_location:]), fix1) and fix_ids_are_in_program(recompose_program(get_lines(program)[fix2_location:]), fix2):
return fix2
if not fix_ids_are_in_program(recompose_program(get_lines(program)[fix1_location:]), fix1) and not fix_ids_are_in_program(recompose_program(get_lines(program)[fix2_location:]), fix2):
raise CouldNotFindUsesForEitherException
if fix1_location < fix2_location:
return fix1
elif fix2_location < fix1_location:
return fix2
prog_lines = get_lines(program)
id_in_fix1 = None
id_in_fix2 = None
for token in fix1.split():
if '_<id>_' in token:
assert id_in_fix1 == None, fix1
id_in_fix1 = token
elif token == '_<op>_[':
break
for token in fix2.split():
if '_<id>_' in token:
assert id_in_fix2 == None, fix2
id_in_fix2 = token
elif token == '_<op>_[':
break
assert id_in_fix1 != id_in_fix2, fix1 + ' & ' + fix2
assert fix1_location == fix2_location
for i in range(fix1_location, len(prog_lines)):
for token in prog_lines[i].split():
if token == id_in_fix1:
return fix1
elif token == id_in_fix2:
return fix2
assert False, 'unreachable code'
raise CouldNotFindUsesForEitherException
def id_mutate(rng,
prog,
max_num_mutations,
num_mutated_progs,
exact=False,
name_dict=None):
assert max_num_mutations > 0 and num_mutated_progs > 0, "Invalid argument(s) supplied to the function token_mutate"
corrupted = []
fixes = []
for _ in range(num_mutated_progs):
tokens = prog
if exact:
num_mutations = max_num_mutations
else:
num_mutations = rng.choice(range(max_num_mutations)) + 1
mutation_count = 0
fix_line = None
for _ in range(num_mutations):
# Step 2: Induce _[ONE]_ mutation, removing empty lines and shifting program if required
try:
mutated, this_fix, _ = undeclare_variable(rng, tokens, tokens)
mutation_count += 1
# Couldn't delete anything
except NothingToMutateException:
break
# Insertion or something failed
except FailedToMutateException:
raise
# Deleted something that can't be fixed (all uses gone from the program)
else:
# REPLACE! program with mutated version
tokens = mutated
if not fix_ids_are_in_program(mutated, this_fix):
# Discarding previous fix: all uses are gone
continue
# Update fix line
if fix_line is not None:
fix_line = which_fix_goes_first(mutated, fix_line,
this_fix)
else:
fix_line = this_fix
if fix_line is not None:
corrupted.append(tokens)
fixes.append(fix_line)
for fix in fixes:
assert fix is not None
return zip(corrupted, fixes)
def token_mutate(prog, max_num_mutations, num_mutated_progs, exact=False, name_dict=None):
assert max_num_mutations > 0 and num_mutated_progs > 0, "Invalid argument(s) supplied to the function token_mutate"
corrupted = []
fixes = []
for _ in range(num_mutated_progs):
tokens = prog
if exact:
num_mutations = max_num_mutations
else:
num_mutations = random.choice(range(max_num_mutations)) + 1
# num_undeclared = np.min([np.random.poisson(num_mutations/5), num_mutations])
deleted_vars = []
mutation_count = 0
# loop_counter = 0
# loop_count_threshold = 50
fix = None
fix_line = None
# last_id = get_last_id(tokens)
# extra_ids = {}
#print tokens_to_source(tokens, name_dict, clang_format=True)
for _ in range(num_mutations):
# Step 2: Induce _[ONE]_ mutation, removing empty lines and shifting program if required
try:
mutated, this_fix, _ = undeclare_variable(tokens, tokens, deleted_vars, name_dict)
mutation_count += 1
#print tokens_to_source(mutated, name_dict, clang_format=True)
#print pretty_fix(this_fix, name_dict)
#print
#print "--------"
#print
# Couldn't delete anything
except NothingToMutateException:
break
# Insertion or something failed
except FailedToMutateException:
#print '{insertion position} or {declaration} localization failed'
#print prog
raise
continue
# Deleted something that can't be fixed (all uses gone from the program)
else:
# REPLACE! program with mutated version
tokens = mutated
if not fix_ids_are_in_program(mutated, this_fix):
#print 'Discarding previous fix: all uses are gone'
continue
# Update fix line
if fix_line is not None:
fix_line = which_fix_goes_first(mutated, fix_line, this_fix)
else:
fix_line = this_fix
if fix_line != None:
corrupted.append(tokens)
fixes.append(fix_line)
for fix in fixes:
assert (fix is not None)
return zip(corrupted, fixes)