Code completion aims at speeding up code writing by recommending to developers the next tokens they are likely to type. Deep Learning (DL) models pushed the boundaries of code completion by redefining what these coding assistants can do: We moved from predicting few code tokens to automatically generating entire functions. One important factor impacting the performance of DL-based code completion techniques is the context provided them as input. With “context” we refer to what the model knows about the code to complete. In a simple scenario, the DL model might be fed with a partially implemented function to complete. In this case, the context is represented by the incomplete function and, based on it, the model must generate a prediction. It is however possible to expand such a context to include additional information, like the whole source code file containing the function to complete, which could be useful to boost the prediction performance. In this work, we present an empirical study investigating how the performance of a DL-based code completion technique is affected by different contexts. We experiment with 8 types of contexts and their combinations. These contexts include: (i) coding contexts, featuring information extracted from the code base in which the code completion is invoked (e.g., code components structurally related to the one to “complete”); (ii) process context, with information aimed at depicting the current status of the project in which a code completion task is triggered (e.g., a textual representation of open issues relevant for the code to complete); and (iii) developer contexts, capturing information about the developer invoking the code completion (e.g., the APIs they frequently use). Our results show that additional contextual information can benefit the performance of DL-based code completion, with relative improvements up to +22% in terms of correct predictions.

When comprehending code, a helping hand may come from the natural language comments documenting it that, unfortunately, are not always there. To support developers in such a scenario, several techniques have been presented to automatically generate natural language summaries for a given code. Most recent approaches exploit deep learning (DL) to automatically document classes or functions, while little effort has been devoted to more fine-grained documentation (e.g., documenting code snippets or even a single statement). Such a design choice is dictated by the availability of training data: For example, in the case of Java, it is easy to create datasets composed of pairs that can be fed to DL models to teach them how to summarize a method. Such a comment-to-code linking is instead non-trivial when it comes to inner comments documenting a few statements. In this work, we take all the steps needed to train a DL model to automatically document code snippets. First, we manually built a dataset featuring 6.6k comments that have been (i) classified based on their type (e.g., code summary, TODO), and (ii) linked to the code statements they document. Second, we used such a dataset to train a multi-task DL model taking as input a comment and being able to (i) classify whether it represents a "code summary" or not, and (ii) link it to the code statements it documents. Our model identifies code summaries with 84% accuracy and is able to link them to the documented lines of code with recall and precision higher than 80%. Third, we run this model on 10k projects, identifying and linking code summaries to the documented code. This unlocked the possibility of building a large-scale dataset of documented code snippets that have then been used to train a new DL model able to automatically document code snippets. A comparison with state-of-the-art baselines shows the superiority of the proposed approach, which however, is still far from representing an accurate solution for snippet summarization.

Logging assists in monitoring events that transpire during the execution of software. Previous research has highlighted the challenges confronted by developers when it comes to logging, including dilemmas such as where to log, what data to record, and which log level to employ (e.g., info, fatal). In this context, we introduced LANCE, an approach rooted in deep learning (DL) that has demonstrated the ability to correctly inject a log statement into Java methods in ∼15% of cases. Nevertheless, LANCE grapples with two primary constraints: (i) it presumes that a method necessitates the inclusion of logging statements and; (ii) it allows the injection of only a single (new) log statement, even in situations where the injection of multiple log statements might be essential. To address these limitations, we present LEONID, a DL-based technique that can distinguish between methods that do and do not require the inclusion of log statements. Furthermore, LEONID supports the injection of multiple log statements within a given method when necessary, and it also enhances LANCE’s proficiency in generating meaningful log messages through the combination of DL and Information Retrieval (IR).

The automatic generation of source code is one of the long-lasting dreams in software engineering research. Several techniques have been proposed to speed up the writing of new code. For example, code completion techniques can recommend to developers the next few tokens they are likely to type, while retrieval-based approaches can suggest code snippets relevant for the task at hand. Also, deep learning has been used to automatically generate code statements starting from a natural language description. While research in this field is very active, there is no study investigating what the users of code recommender systems (i.e., software practitioners) actually need from these tools. We present a study involving 80 software developers to investigate the characteristics of code recommender systems they consider important. The output of our study is a taxonomy of 70 “requirements” that should be considered when designing code recommender systems. For example, developers would like the recommended code to use the same coding style of the code under development. Also, code recommenders being “aware” of the developers' knowledge (e.g., what are the framework/libraries they already used in the past) and able to customize the recommendations based on this knowledge would be appreciated by practitioners. The taxonomy output of our study points to a wide set of future research directions for code recommenders.

Software engineering research has always being concerned with the improvement of code completion approaches, which suggest the next tokens a developer will likely type while coding. The release of GitHub Copilot constitutes a big step forward, also because of its unprecedented ability to automatically generate even entire functions from their natural language description. While the usefulness of Copilot is evident, it is still unclear to what extent it is robust. Specifically, we do not know the extent to which semantic-preserving changes in the natural language description provided to the model have an effect on the generated code function. In this paper we present an empirical study in which we aim at understanding whether different but semantically equivalent natural language descriptions result in the same recommended function. A negative answer would pose questions on the robustness of deep learning (DL)-based code generators since it would imply that developers using different wordings to describe the same code would obtain different recommendations. We asked Copilot to automatically generate 892 Java methods starting from their original Javadoc description. Then, we generated different semantically equivalent descriptions for each method both manually and automatically, and we analyzed the extent to which predictions generated by Copilot changed. Our results show that modifying the description results in different code recommendations in ∼46% of cases. Also, differences in the semantically equivalent descriptions might impact the correctness of the generated code (±28%).

Transformers have gained popularity in the software engineering (SE) literature. These deep learning models are usually pre-trained through a self-supervised objective, meant to provide the model with basic knowledge about a language of interest (e.g., Java). A classic pre-training objective is the masked language model (MLM), in which a percentage of tokens from the input (e.g., a Java method) is masked, with the model in charge of predicting them. Once pre-trained, the model is then fine-tuned to support the specific downstream task of interest (e.g., code summarization). While there is evidence suggesting the boost in performance provided by pre-training, little is known about the impact of the specific pre-training objective(s) used. Indeed, MLM is just one of the possible pre-training objectives and recent work from the natural language processing field suggest that pre-training objectives tailored for the specific downstream task of interest may substantially boost the model's performance. For example, in the case of code summarization, a tailored pre-training objective could be the identification of an appropriate name for a given method, considering the method name to generate as an extreme summary. In this study, we focus on the impact of pre-training objectives on the performance of trans-formers when automating code-related tasks. We start with a systematic literature review aimed at identifying the pre-training objectives used in SE. Then, we pre-train 32 transformers using both (i) generic pre-training objectives usually adopted in SE; and (ii) pre-training objectives tailored to specific code-related tasks subject of our experimentation, namely bug-fixing, code summarization, and code completion. We also compare the pre-trained models with non pre-trained ones and show the advantage brought by pre-training in different scenarios, in which more or less fine-tuning data are available. Our results show that: (i) pre-training helps in boosting performance only if the amount of fine-tuning data available is small; (ii) the MLM objective is usually sufficient to maximize the prediction performance of the model, even when comparing it with pre-training objectives specialized for the downstream task at hand.

Abstract Identifiers, such as method and variable names, form a large portion of source code. Therefore, low-quality identifiers can substantially hinder code comprehension. To support developers in using meaningful identifiers, several (semi-)automatic techniques have been proposed, mostly being data-driven (e.g., statistical language models, deep learning models) or relying on static code analysis. Still, limited empirical investigations have been performed on the effectiveness of such techniques for recommending developers with meaningful identifiers, possibly resulting in rename refactoring operations. We present a large-scale study investigating the potential of data-driven approaches to support automated variable renaming. We experiment with three state-of-the-art techniques: a statistical language model and two DL-based models. The three approaches have been trained and tested on three datasets we built with the goal of evaluating their ability to recommend meaningful variable identifiers. Our quantitative and qualitative analyses show the potential of such techniques that, under specific conditions, can provide valuable recommendations and are ready to be integrated in rename refactoring tools. Nonetheless, our results also highlight limitations of the experimented approaches that call for further research in this field.

Automatically linking bug-fixing changes to bug-inducing ones (BICs) is one of the key data-extraction steps behind several empirical studies in software engineering. The SZZ algorithm is the de facto standard to achieve this goal, with several improvements proposed over time. Evaluating the performance of SZZ implementations is, however, far from trivial. In previous works, researchers (i) manually assessed whether the BICs identified by the SZZ implementation were correct or not, or (ii) defined oracles in which they manually determined BICs from bug-fixing commits. However, ideally, the original developers should be involved in defining a labeled dataset to evaluate SZZ implementations. We propose a methodology to define a “developer-informed” oracle for evaluating SZZ implementations, without requiring a manual inspection from the original developers. We use Natural Language Processing (NLP) to identify bug-fixing commits in which developers explicitly reference the commit(s) that introduced the fixed bug. We use the built oracle to extensively evaluate existing SZZ variants defined in the literature. We also introduce and evaluate two new variants aimed at addressing two weaknesses we observed in state-of-the-art implementations (i.e., processing added lines and handling of revert commits).

Fine-grained just-in-time defect prediction aims at identifying likely defective files within new commits. Popular techniques are based on supervised learning, where machine learning algorithms are fed with historical data. One of the limitations of these techniques is concerned with the use of imbalanced data that only contain a few defective samples to enable a proper learning phase. To overcome this problem, recent work has shown that anomaly detection can be used as an alternative. With our study, we aim at assessing how anomaly detection can be employed for the problem of fine-grained just-in-time defect prediction. We conduct an empirical investigation on 32 open-source projects, designing and evaluating three anomaly detection methods for fine-grained just-in-time defect prediction. Our results do not show significant advantages that justify the benefit of anomaly detection over machine learning approaches.

Logging is a practice widely adopted in several phases of the software lifecycle. For example, during software development log statements allow engineers to verify and debug the system by exposing fine-grained information of the running software. While the benefits of logging are undisputed, taking proper decisions about where to inject log statements, what information to log, and at which log level (e.g., error, warning) is crucial for the logging effectiveness. In this paper, we present LANCE (Log stAtemeNt reCommEnder), the first approach supporting developers in all these decisions. LANCE features a Text-To-Text-Transfer-Transformer (T5) model that has been trained on 6,894,456 Java methods. LANCE takes as input a Java method and injects in it a full log statement, including a human-comprehensible logging message and properly choosing the needed log level and the statement location. Our results show that LANCE is able to (i) properly identify the location in the code where to inject the statement in 65.9% of Java methods requiring it; (ii) selecting the proper log level in 66.2% of cases; and (iii) generate a completely correct log statement including a meaningful logging message in 15.2% of cases.

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.

Different from what happens for most types of software systems, testing video games has largely remained a manual activity performed by human testers. This is mostly due to the continuous and intelligent user interaction video games require. Recently, reinforcement learning (RL) has been exploited to partially automate functional testing. RL enables training smart agents that can even achieve super-human performance in playing games, thus being suitable to explore them looking for bugs. We investigate the possibility of using RL for load testing video games. Indeed, the goal of game testing is not only to identify functional bugs, but also to examine the game's performance, such as its ability to avoid lags and keep a minimum number of frames per second (FPS) when high-demanding 3D scenes are shown on screen. We define a methodology employing RL to train an agent able to play the game as a human while also trying to identify areas of the game resulting in a drop of FPS. We demonstrate the feasibility of our approach on three games. Two of them are used as proof-of-concept, by injecting artificial performance bugs. The third one is an open-source 3D game that we load test using the trained agent showing its potential to identify areas of the game resulting in lower FPS.

Deep Learning (DL) models have been widely used to support code completion. These models, once properly trained, can take as input an incomplete code component (e.g., an incomplete function) and predict the missing tokens to finalize it. GitHub Copilot is an example of code recommender built by training a DL model on millions of open source repositories: The source code of these repositories acts as training data, allowing the model to learn "how to program". The usage of such a code is usually regulated by Free and Open Source Software (FOSS) licenses, that establish under which conditions the licensed code can be redistributed or modified. As of Today, it is unclear whether the code generated by DL models trained on open source code should be considered as "new" or as "derivative" work, with possible implications on license infringements. In this work, we run a large-scale study investigating the extent to which DL models tend to clone code from their training set when recommending code completions. Such an exploratory study can help in assessing the magnitude of the potential licensing issues mentioned before: If these models tend to generate new code that is unseen in the training set, then licensing issues are unlikely to occur. Otherwise, a revision of these licenses urges to regulate how the code generated by these models should be treated when used, for example, in a commercial setting. Highlights from our results show that ~10% to ~0.1% of the predictions generated by a state-of-the-art DL-based code completion tool are Type-1 clones of instances in the training set, depending on the size of the predicted code. Long predictions are unlikely to be cloned.

Code completion aims at speeding up code writing by predicting the next code token(s) the developer is likely to write. Works in this field focused on improving the accuracy of the generated predictions, with substantial leaps forward made possible by deep learning (DL) models. However, code completion techniques are mostly evaluated in the scenario of predicting the next token to type, with few exceptions pushing the boundaries to the prediction of an entire code statement. Thus, little is known about the performance of state-of-the-art code completion approaches in more challenging scenarios in which, for example, an entire code block must be generated. We present a large-scale study exploring the capabilities of state-of-the-art Transformer-based models in supporting code completion at different granularity levels, including single tokens, one or multiple entire statements, up to entire code blocks (e.g., the iterated block of a for loop). We experimented with several variants of two recently proposed Transformer-based models, namely RoBERTa and the Text-To-Text Transfer Transformer (T5), for the task of code completion. The achieved results show that Transformer-based models, and in particular the T5, represent a viable solution for code completion, with perfect predictions ranging from ~29%, obtained when asking the model to guess entire blocks, up to ~69%, reached in the simpler scenario of few tokens masked from the same code statement.

The SZZ algorithm for identifying bug-inducing changes has been widely used to evaluate defect prediction techniques and to empirically investigate when, how, and by whom bugs are introduced. Over the years, researchers have proposed several heuristics to improve the SZZ accuracy, providing various implementations of SZZ. However, fairly evaluating those implementations on a reliable oracle is an open problem: SZZ evaluations usually rely on (i) the manual analysis of the SZZ output to classify the identified bug-inducing commits as true or false positives; or (ii) a golden set linking bug-fixing and bug-inducing commits. In both cases, these manual evaluations are performed by researchers with limited knowledge of the studied subject systems. Ideally, there should be a golden set created by the original developers of the studied systems. We propose a methodology to build a "developer-informed" oracle for the evaluation of SZZ variants. We use Natural Language Processing (NLP) to identify bug-fixing commits in which developers explicitly reference the commit(s) that introduced a fixed bug. This was followed by a manual filtering step aimed at ensuring the quality and accuracy of the oracle. Once built, we used the oracle to evaluate several variants of the SZZ algorithm in terms of their accuracy. Our evaluation helped us to distill a set of lessons learned to further improve the SZZ algorithm.

Code reviews are popular in both industrial and open source projects. The benefits of code reviews are widely recognized and include better code quality and lower likelihood of introducing bugs. However, since code review is a manual activity it comes at the cost of spending developers' time on reviewing their teammates' code. Our goal is to make the first step towards partially automating the code review process, thus, possibly reducing the manual costs associated with it. We focus on both thecontributorand thereviewersides of the process, by training two different Deep Learning architectures. The first one learns code changes performed by developers during real code review activities, thus providing thecontributorwith a revised version of her code implementing code transformations usually recommended during code review before the code is even submitted for review. The second one automatically provides thereviewercommenting on a submitted code with the revised code implementing her comments expressed in natural language. The empirical evaluation of the two models shows that, on the contributor side, the trained model succeeds in replicating the code transformations applied during code reviews in up to 16% of cases. On the reviewer side, the model can correctly implement a comment provided in natural language in up to 31% of cases. While these results are encouraging, more research is needed to make these models usable by developers.

Code comments play a prominent role in program comprehension activities. However, source code is not always documented and code and comments not always co-evolve. To deal with these issues, researchers have proposed techniques to automatically generate comments documenting a given code at hand. The most recent works in the area applied deep learning (DL) techniques to support such a task. Despite the achieved advances, the empirical evaluations of these approaches show that they are still far from a performance level that would make them valuable for developers. We tackle a simpler and related problem: Code comment completion. Instead of generating a comment for a given code from scratch, we investigate the extent to which state-of-the-art techniques can help developers in writing comments faster. We present a large-scale study in which we empirically assess how a simple n-gram model and the recently proposed Text-To-Text Transfer Transformer (T5) architecture can perform in autocompleting a code comment the developer is typing. The achieved results show the superiority of the T5 model, despite the n-gram model being a competitive solution.

Code completion is one of the main features of modern Integrated Development Environments (IDEs). Its objective is to speed up code writing by predicting the next code token(s) the developer is likely to write. Research in this area has substantially bolstered the predictive performance of these techniques. However, the support to developers is still limited to the prediction of the next few tokens to type. In this work, we take a step further in this direction by presenting a large-scale empirical study aimed at exploring the capabilities of state-of-the-art deep learning (DL) models in supporting code completion at different granularity levels, including single tokens, one or multiple entire statements, up to entire code blocks (e.g., the iterated block of a for loop). To this aim, we train and test several adapted variants of the recently proposed RoBERTa model, and evaluate its predictions from several perspectives, including: (i) metrics usually adopted when assessing DL generative models (i.e., BLEU score and Levenshtein distance); (ii) the percentage of perfect predictions (i.e., the predicted code snippets that match those written by developers); and (iii) the "semantic" equivalence of the generated code as compared to the one written by developers. The achieved results show that BERT models represent a viable solution for code completion, with perfect predictions ranging from ~7%, obtained when asking the model to guess entire blocks, up to ~58%, reached in the simpler scenario of few tokens masked from the same code statement.

Code comments are a key software component containing information about the underlying implementation. Several studies have shown that code comments enhance the readability of the code. Nevertheless, not all the comments have the same goal and target audience. In this paper, we investigate how 14 diverse Java open and closed source software projects use code comments, with the aim of understanding their purpose. Through our analysis, we produce a taxonomy of source code comments; subsequently, we investigate how often each category occur by manually classifying more than 40,000 lines of code comments from the aforementioned projects. In addition, we investigate how to automatically classify code comments at line level into our taxonomy using machine learning; initial results are promising and suggest that an accurate classification is within reach, even when training the machine learner on projects different than the target one. Data and Materials [ https://doi.org/10.5281/zenodo.2628361 ].

Obtaining a good dataset to conduct empirical studies on the engineering of Android apps is an open challenge. To start tackling this challenge, we present AndroidTimeMachine, the first, self-contained, publicly available dataset weaving spread-out data sources about real-world, open-source Android apps. Encoded as a graph-based database, AndroidTimeMachine concerns 8,431 real open-source Android apps and contains: (i) metadata about the apps' GitHub projects, (ii) Git repositories with full commit history and (iii) metadata extracted from the Google Play store, such as app ratings and permissions.

Recent research has provided evidence that, in the industrial context, developing video games diverges from developing software systems in other domains, such as office suites and system utilities.

Bug prediction is aimed at supporting developers in the identification of code artifacts more likely to be defective. Researchers have proposed prediction models to identify bug prone methods and provided promising evidence that it is possible to operate at this level of granularity. Particularly, models based on a mixture of product and process metrics, used as independent variables, led to the best results. In this study, we first replicate previous research on method-level bug prediction on different systems/timespans. Afterwards, we reflect on the evaluation strategy and propose a more realistic one. Key results of our study show that the performance of the method-level bug prediction model is similar to what previously reported also for different systems/timespans, when evaluated with the same strategy. However-when evaluated with a more realistic strategy-all the models show a dramatic drop in performance exhibiting results close to that of a random classifier. Our replication and negative results indicate that method-level bug prediction is still an open challenge.

Contemporary code review is a widespread practice used by software engineers to maintain high software quality and share project knowledge. However, conducting proper code review takes time and developers often have limited time for review. In this paper, we aim at investigating the information that reviewers need to conduct a proper code review, to better understand this process and how research and tool support can make developers become more effective and efficient reviewers. Previous work has provided evidence that a successful code review process is one in which reviewers and authors actively participate and collaborate. In these cases, the threads of discussions that are saved by code review tools are a precious source of information that can be later exploited for research and practice. In this paper, we focus on this source of information as a way to gather reliable data on the aforementioned reviewers' needs. We manually analyze 900 code review comments from three large open-source projects and organize them in categories by means of a card sort. Our results highlight the presence of seven high-level information needs, such as knowing the uses of methods and variables declared/modified in the code under review. Based on these results we suggest ways in which future code review tools can better support collaboration and the reviewing task.

Code comments are a key software component containing information about the underlying implementation. Several studies have shown that code comments enhance the readability of the code. Nevertheless, not all the comments have the same goal and target audience. In this paper, we investigate how six diverse Java OSS projects use code comments, with the aim of understanding their purpose. Through our analysis, we produce a taxonomy of source code comments, subsequently, we investigate how often each category occur by manually classifying more than 2,000 code comments from the aforementioned projects. In addition, we conduct an initial evaluation on how to automatically classify code comments at line level into our taxonomy using machine learning, initial results are promising and suggest that an accurate classification is within reach.