A PyTorch Implementation of "Tare: Type-Aware Neural Program Repair".

Automated program repair (APR) aims to reduce the effort for software development. With the development of deep learning, lots of DL-based APR approaches have been proposed using an encoder-decoder architecture. Despite the promising performance, these models share one same limitation: generating lots of untypable patches. The main reason for this phenomenon is that the existing models do not consider the constraints of code captured by a set of typing rules.

In this paper, we propose, Tare, a type-aware model for neural program repair to learn the typing rules. To encode an individual typing rule, we introduces three novel components: (1) a novel type of grammars, T-Grammar that integrates the type information into a standard grammar, (2) a novel representation of code, T-Graph that integrates the key information needed for type checking in an AST, and (3) a novel type-aware neural program repair approach, Tare that encodes the T-Graph and generates the patches guided by T-Grammar.

The experiment was conducted on three benchmarks, 393 bugs from Defects4J v1.2, 444 additional bugs from Defects4J v2.0, and 40 bugs from QuixBugs. Our results show that Tare repairs 64, 32, and 27 bugs on these benchmarks respectively and outperforms the existing APR approaches on all benchmarks. The further analysis also shows that Tare tends to generate more compilable patches than the existing DL-based APR approaches with the typing rule information.

.
├── generation: code for patch generation
│   ├── bugs-QuixBugs
│ 	├── location
│ 	├── location2
│ 	└── tesetDefect4j.py
├── model: code for neural model
│   ├── data
│   ├── Model.py
│   └── relAttention.py
├── train: code for training the model
│   ├── run.py
│   ├── Dataset.py
│   └── SearchNode.py
├── validation: code for patch validation
│   ├── patches-all: plausible patches
│   ├── Dataset.py
│   └── SearchNode.py

The raw data https://drive.google.com/drive/folders/1ECNX98qj9FMdRT2MXOUY6aQ6-sNT0b_a?usp=sharing from Recoder.

❗❗❗ We test Tare with the perfect fault localization on Defects4J 2.0.

The left of table below is copied the original Table V in the paper, showing the results using Ochiai for fault localization, where "x/y" indicates the approach correctly repairs x bugs and produces plausible patches for y bugs (a plausible patch passes all the tests). The last column is newly added, showing the number of bugs Tare correctly repairs with perfect fault localization. Only the first plausible patch is considered in the experiment.

CUDA_VISIBLE_DEVICES=0,1 python3 train/run.py train

The saved model is checkModel/best_model.cpkt.

CUDA_VISIBLE_DEVICES=0 python3 generation/testDefect4j.py bugid

The generated patches are in folder generation/patch-all/patch/ in json.

CUDA_VISIBLE_DEVICES=0 python3 generation/testDefect4j1.py bugid

The generated patches are in folder generation/patch-all/patchground/ in json.

CUDA_VISIBLE_DEVICES=0 python3 generation/testDefects4j2.py bugid

The generated patches are in folder generation/patch-all/patch2 in json.

CUDA_VISIBLE_DEVICES=0 python3 generation/testQuixbug.py bugid

The generated patches are in folder generation/patch-all/patchQuix in json.

python3 validation/repair.py bugid

The results are in folder validation/patch-all/patches/ in json.

python3 validation/repair1.py bugid

The results are in folder validation/patch-all/patchesground/ in json.

python3 validation/repair2.py bugid

The results are in folder validation/patch-all/patches2/ in json.

python3 validation/repairqfix.py bugid

The results are in folder validation/patch-all/patchesqfix/ in json.

The generated patches are in the folder patches-all.

• Python 3.7
• PyTorch 1.3
• Defects4J
• Java 8
• docker
• nvidia-docker