Deep Just-In-Time Inconsistency Detection Between Comments and Source Code

Code and datasets for our AAAI-2021 paper "Deep Just-In-Time Inconsistency Detection Between Comments and Source Code" which can be found here.

If you find this work useful, please consider citing our paper:

@inproceedings{PanthaplackelETAL21DeepJITInconsistency,
  author = {Panthaplackel, Sheena and Li, Junyi Jessy and Gligoric, Milos and Mooney, Raymond J.},
  title = {Deep Just-In-Time Inconsistency Detection Between Comments and Source Code},
  booktitle = {AAAI},
  pages = {427--435},
  year = {2021},
}

The code base shares components with our prior work called Learning to Update Natural Language Comments Based on Code Changes.

Download data from here. Download additional model resources from here. Edit configurations in constants.py to specify data, resource, and output locations.

Inconsistency Detection:

SEQ(C, M_edit) + features

python3 run_comment_model.py --task=detect --attend_code_sequence_states --features --model_path=detect_attend_code_sequence_states_features.pkl.gz --model_name=detect_attend_code_sequence_states_features

GRAPH(C, T_edit) + features (The GGNN used for this approach is derived from here.)

python3 run_comment_model.py --task=detect --attend_code_graph_states --features --model_path=detect_attend_code_graph_states_features.pkl.gz --model_name=detect_attend_code_graph_states_features

HYBRID(C, M_edit, T_edit) + features

python3 run_comment_model.py --task=detect --attend_code_sequence_states --attend_code_graph_states --features --model_path=detect_attend_code_sequence_states_attend_code_graph_states_features.pkl.gz --model_name=detect_attend_code_sequence_states_attend_code_graph_states_features

To run inference on a detection model, add --test_mode to the command used to train the model.

Combined Detection + Update:

Update w/ implicit detection

python3 run_comment_model.py --task=update --features --model_path=update_features.pkl.gz --model_name=update_features

To run inference, add --test_mode --rerank to the command used to train the model.

Pretrained update + detection

python3 run_comment_model.py --task=update --features --positive_only --model_path=update_features_positive_only.pkl.gz --model_name=update_features_positive_only

One of the detection models should also be trained, following instructions provided in the "Inconsistency Detection" section above. To run inference on the update model, add --test_mode --rerank to the command used to train the model. Inference on the detection model should also be done as instructed in the "Inconsistency Detection" section.

Jointly trained update + detection

To train, simply replace --task=detect with --task=dual in the configurations given for "Inconsistency Detection." For inference, additionally include --test_mode --rerank.

Displaying metrics:

To display metrics for the full test set as well as the cleaned test sample, run:

python3 display_scores.py --detection_output_file=[PATH TO DETECTION PREDICTIONS] --update_output_file=[PATH TO UPDATE PREDICTIONS]

For evaluating in the pretrained update + detection setting, both filepaths are required. For all other settings, only one should be specified.

AST Diffing:

The AST diffs were built using Java files provided by Pengyu Nie. First, download ast-diffing-1.6-jar-with-dependencies.jar from here. Then, go to data_processing/ast_diffing/python and run:

python3 xml_diff_parser.py --old_sample_path=[PATH TO OLD VERSION OF CODE] --new_sample_path=[PATH TO NEW VERSION OF CODE] --jar_path=[PATH TO DOWNLOADED JAR FILE]

You can see an example by running:

python3 xml_diff_parser.py --old_sample_path=../code_samples/old.java  --new_sample_path=../code_samples/new.java --jar_path=[PATH TO DOWNLOADED JAR FILE]

Logende/deep-jit-inconsistency-detection

Deep Just-In-Time Inconsistency Detection Between Comments and Source Code