Towards Automating Code Review Activities

Tufano, Rosalia; Pascarella, Luca; Tufano, Michele; Poshyvanyk, Denys; Bavota, Gabriele

doi:10.1109/icse43902.2021.00027

Cited by 71 publications

(70 citation statements)

References 34 publications

(34 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This resulted in a list of 4,901 projects. We also mined the six Gerrit [1] installations used in [46] containing code review data about 6,388 projects.…”

Section: Fine-tuning Datasetsmentioning

confidence: 99%

“…Overall, we mined 382,955 valid triplets from GitHub and Gerrit using the pipeline from [46] that we summarize in the following (see [46] for additional details). We target triplets in which a comment 𝑐 𝑛𝑙 has been posted by a reviewer on a method 𝑚 𝑠 .…”

Section: Fine-tuning Datasetsmentioning

confidence: 99%

“…In this paper, we build on top of our previous work [46] experimenting with DL models for code review automation in more realistic and challenging scenarios. We start by training the recently proposed Text-To-Text-Transfer Transformer (T5) model [35] on a dataset similar to the one used in [46]. However, we adopt a tokenizer (i.e., SentencePiece [26]) that allows us to work with raw source code, without the need for code abstraction.…”

Section: Introductionmentioning

confidence: 99%

“…Also, we increase the maximum length of the considered code components from 100 "abstracted" tokens to 512 "SentencePiece" tokens (i.e., ∼390 "abstracted" tokens). The absence of an abstraction mechanism and the increased upper bound for input/output length allowed us to build a substantially larger dataset as compared to the one used in [46] (168k instances vs. 17k) and, more importantly, to feature in such a dataset a wider variety of code transformations implemented in the code review process, including quite challenging instances such as those requiring the introduction of new identifiers and literals (accounting for 63% of the new dataset we built). Also, we experimented with the automation of a third task related to the code review process: Given the code submitted for review (𝐶 𝑠 ), generating a natural language comment 𝑅 𝑛𝑙 requesting to the contributor code changes as a reviewer would do (i.e., simulating a reviewer commenting on the submitted code).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Using Pre-Trained Models to Boost Code Review Automation

Tufano¹,

Masiero²,

Mastropaolo³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Code review is a practice widely adopted in open source and industrial projects. Given the non-negligible cost of such a process, researchers started investigating the possibility of automating specific code review tasks. We recently proposed Deep Learning (DL) models targeting the automation of two tasks: the first model takes as input a code submitted for review and implements in it changes likely to be recommended by a reviewer; the second takes as input the submitted code and a reviewer comment posted in natural language and automatically implements the change required by the reviewer. While the preliminary results we achieved are encouraging, both models had been tested in rather simple code review scenarios, substantially simplifying the targeted problem. This was also due to the choices we made when designing both the technique and the experiments. In this paper, we build on top of that work by demonstrating that a pre-trained Text-To-Text Transfer Transformer (T5) model can outperform previous DL models for automating code review tasks. Also, we conducted our experiments on a larger and more realistic (and challenging) dataset of code review activities.

show abstract

“…This resulted in a list of 4,901 projects. We also mined the six Gerrit [1] installations used in [46] containing code review data about 6,388 projects.…”

Section: Fine-tuning Datasetsmentioning

confidence: 99%

Section: Fine-tuning Datasetsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Using Pre-Trained Models to Boost Code Review Automation

Tufano¹,

Masiero²,

Mastropaolo³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the understanding of automatic code review is not the only one, Tufan [38] proposes a method, learning the code changes recommended by reviewer, to implement them in the original code automatically. In other words, they are trying to make a map from the original code file to the revised code file, which is totally different from the opinion that Shi [9] hold.…”

Section: Automatic Code Reviewmentioning

confidence: 99%

Turn Tree into Graph: Automatic Code Review via Simplified AST Driven Graph Convolutional Network

Wu¹,

Liang²,

Zhang³

2022

Preprint

View full text Add to dashboard Cite

Automatic code review (ACR), which can relieve the costs of manual inspection, is an indispensable and essential task in software engineering. To deal with ACR, existing work is to serialize the abstract syntax tree (AST). However, making sense of the whole AST with sequence encoding approach is a daunting task, mostly due to some redundant nodes in AST hinder the transmission of node information. Not to mention that the serialized representation is inadequate to grasp the information of tree structure in AST. In this paper, we first present a new large-scale Apache Automatic Code Review (AACR) dataset for ACR task since there is still no publicly available dataset in this task. The release of this dataset would push forward the research in this field. Based on it, we propose a novel Simplified AST based Graph Convolutional Network (SimAST-GCN) to deal with ACR task. Concretely, to improve the efficiency of node information dissemination, we first simplify the AST of code by deleting the redundant nodes that do not contain connection attributes, and thus deriving a Simplified AST. Then, we construct a relation graph for each code based on the Simplified AST to properly embody the relations among code fragments of the tree structure into the graph. Subsequently, in the light of the merit of graph structure, we explore a graph convolution networks architecture that follows an attention mechanism to leverage the crucial implications of code fragments to derive code representations. Finally, we exploit a simple but effective subtraction operation in the representations between the original and revised code, enabling the revised difference to be preferably learned for deciding the results of ACR. Experimental results on the AACR dataset illustrate that our proposed model outperforms the state-of-the-art methods.

show abstract