def _test(self):
y_cur = deepcopy(self.y)
for i, yi in zip(y_indexes, y_pred):
y_cur[i] = yi
code_generator = CodeGenerator(self.feature_extractor)
pred_vnodes = code_generator.apply_predicted_y(
self.vnodes, self.vnodes_y, list(range(len(self.vnodes_y))),
FakeRules(y_cur))
generated_file = code_generator.generate(pred_vnodes, "local")
self.assertEqual(generated_file, result_local)
def _generate_token_fixes(
self,
file: File,
fe: FeatureExtractor,
feature_extractor_output,
bblfsh_stub: "bblfsh.aliases.ProtocolServiceStub",
rules: Rules,
) -> Tuple[List[LineFix], List[VirtualNode], numpy.ndarray, numpy.ndarray]:
X, y, (vnodes_y, vnodes, vnode_parents,
node_parents) = feature_extractor_output
y_pred_pure, rule_winners, new_rules, grouped_quote_predictions = rules.predict(
X=X, vnodes_y=vnodes_y, vnodes=vnodes, feature_extractor=fe)
y_pred = rules.fill_missing_predictions(y_pred_pure, y)
if self.config["uast_break_check"]:
y, y_pred, vnodes_y, rule_winners, safe_preds = filter_uast_breaking_preds(
y=y,
y_pred=y_pred,
vnodes_y=vnodes_y,
vnodes=vnodes,
files={file.path: file},
feature_extractor=fe,
stub=bblfsh_stub,
vnode_parents=vnode_parents,
node_parents=node_parents,
rule_winners=rule_winners,
grouped_quote_predictions=grouped_quote_predictions)
y_pred_pure = y_pred_pure[safe_preds]
assert len(y) == len(y_pred)
assert len(y) == len(rule_winners)
code_generator = CodeGenerator(fe, skip_errors=True)
new_vnodes = code_generator.apply_predicted_y(vnodes, vnodes_y,
rule_winners, new_rules)
token_fixes = []
for line_number, line in self._group_line_nodes(
y, y_pred, vnodes_y, new_vnodes, rule_winners):
line_ys, line_ys_pred, line_vnodes_y, new_line_vnodes, line_winners = line
new_code_line = code_generator.generate(
new_line_vnodes, "local").lstrip("\n").splitlines()[0]
confidence = self._get_comment_confidence(line_ys, line_ys_pred,
line_winners, new_rules)
fixed_vnodes = [
vnode for vnode in new_line_vnodes
if hasattr(vnode, "y_old") and vnode.y_old != vnode.y
]
token_fixes.append(
LineFix(
line_number=line_number, # line number for the comment
suggested_code=
new_code_line, # code line suggested by our model
fixed_vnodes=fixed_vnodes, # VirtualNode-s with changed y
confidence=
confidence, # overall confidence in the prediction, 0-100
))
return token_fixes, new_vnodes, y_pred_pure, y
def _test(self):
y_cur = deepcopy(self.y)
for offset, yi in zip(offsets, y_pred):
i = None
for i, vnode in enumerate(vnodes_y): # noqa: B007
if offset == vnode.start.offset:
break
y_cur[i] = yi
code_generator = CodeGenerator(self.feature_extractor)
pred_vnodes = code_generator.apply_predicted_y(
self.vnodes, self.vnodes_y, list(range(len(self.vnodes_y))),
FakeRules(y_cur))
generated_file = code_generator.generate(pred_vnodes)
self.assertEqual(generated_file, result)
def test_reproduction(self):
for indent in ("local", "global"):
code_generator = CodeGenerator(self.feature_extractor)
generated_file = code_generator.generate(self.vnodes, indent)
self.assertEqual(generated_file, self.file.content.decode("utf-8"))
def test_reproduction(self):
code_generator = CodeGenerator(self.feature_extractor)
generated_file = code_generator.generate(self.vnodes)
self.assertEqual(generated_file, self.file.content)