Recent advancements in developing Pre-trained Language Models for Code (Code-PLMs) have urged many areas of Software Engineering (SE) and brought breakthrough results for many SE tasks. Though these models have achieved the state-of-the-art performance for SE tasks for many popular programming languages, such as Java and Python, the Scientific Software and its related languages like R programming language have rarely benefited or even been evaluated with the Code-PLMs. Research has shown that R has many differences with other programming languages and requires specific techniques. In this study, we provide the first insights for code intelligence for R. For this purpose, we collect and open source an R dataset, and evaluate Code-PLMs for the two tasks of code summarization and method name prediction using several settings and strategies, including the differences in two R styles, Tidy-verse and Base R. Our results demonstrate that the studied models have experienced varying degrees of performance degradation when processing R programming language code, which is supported by human evaluation. Additionally, not all models show performance improvement in R-specific tasks even after multi-language fine-tuning. The dual syntax paradigms in R significantly impact the models' performance, particularly in code summarization tasks. Furthermore, the project-specific context inherent in R codebases significantly impacts the performance when attempting cross-project training.
Large Language Models for Code (code LLMs) have demonstrated remarkable performance across various software engineering (SE) tasks, increasing the application of code LLMs in software development. Despite the success of code LLMs, there remain significant concerns about the actual capabilities and reliability of these models, "whether these models really learn the semantics of code from the training data and leverage the learned knowledge to perform the SE tasks". In this paper, we introduce EMPICA, a comprehensive framework designed to systematically and empirically evaluate the capabilities of code LLMs in understanding code semantics. Specifically, EMPICA systematically introduces controlled modifications/transformations into the input code and examines the models' responses. Generally, code LLMs must be robust to semantically equivalent code inputs and be sensitive to non-equivalent ones for all SE tasks. Specifically, for every SE task, given an input code snippet c and its semantic equivalent variants, code LLMs must robustly produce consistent/equivalent outputs while they are expected to generate different outputs for c and its semantic non-equivalent variants. Our experimental results on three representative code understanding tasks, including code summarization, method name prediction, and output prediction, reveal that the robustness and sensitivity of the state-of-the-art code LLMs to code transformations vary significantly across tasks and transformation operators. In addition, the code LLMs exhibit better robustness to the semantic preserving transformations than their sensitivity to the semantic non-preserving transformations. These results highlight a need to enhance the model's capabilities of understanding code semantics, especially the sensitivity property.
Evaluating natural language generation models, particularly for method name prediction, poses significant challenges. A robust metric must account for the versatility of method naming, considering both semantic and syntactic variations. Traditional overlap-based metrics, such as ROUGE, fail to capture these nuances. Existing embedding-based metrics often suffer from imbalanced precision and recall, lack normalized scores, or make unrealistic assumptions about sequences. To address these limitations, we leverage the theory of optimal transport and construct WRDScore, a novel metric that strikes a balance between simplicity and effectiveness. In the WRDScore framework, we define precision as the maximum degree to which the predicted sequence's tokens are included in the reference sequence, token by token. Recall is calculated as the total cost of the optimal transport plan that maps the reference sequence to the predicted one. Finally, WRDScore is computed as the harmonic mean of precision and recall, balancing these two complementary metrics. Our metric is lightweight, normalized, and precision-recall-oriented, avoiding unrealistic assumptions while aligning well with human judgments. Experiments on a human-curated dataset confirm the superiority of WRDScore over other available text metrics.
Pre-trained Code Language Models (Code-PLMs) have shown many advancements and achieved state-of-the-art results for many software engineering tasks in the past few years. These models are mainly targeted for popular programming languages such as Java and Python, leaving out many other ones like R. Though R has a wide community of developers and users, there is little known about the applicability of Code-PLMs for R. In this preliminary study, we aim to investigate the vulnerability of Code-PLMs for code entities in R. For this purpose, we use an R dataset of code and comment pairs and then apply CodeAttack, a black-box attack model that uses the structure of code to generate adversarial code samples. We investigate how the model can attack different entities in R. This is the first step towards understanding the importance of R token types, compared to popular programming languages (e.g., Java). We limit our study to code summarization. Our results show that the most vulnerable code entity is the identifier, followed by some syntax tokens specific to R. The results can shed light on the importance of token types and help in developing models for code summarization and method name prediction for the R language.
Artificial intelligence (AI) has revolutionized software engineering (SE) by enhancing software development efficiency. The advent of pre-trained models (PTMs) leveraging transfer learning has significantly advanced AI for SE. However, existing PTMs that operate on individual code tokens suffer from several limitations: They are costly to train and fine-tune; and they rely heavily on labeled data for fine-tuning on task-specific datasets. In this paper, we present TransformCode, a novel framework that learns code embeddings in a contrastive learning manner. Our framework is encoder-agnostic and language-agnostic, which means that it can leverage any encoder model and handle any programming language. We also propose a novel data-augmentation technique called abstract syntax tree (AST) transformation, which applies syntactic and semantic transformations to the original code snippets, to generate more diverse and robust samples for contrastive learning. Our framework has several advantages over existing methods: (1) It is flexible and adaptable, because it can easily be extended to other downstream tasks that require code representation (such as code-clone detection and classification); (2) it is efficient and scalable, because it does not require a large model or a large amount of training data, and it can support any programming language; (3) it is not limited to unsupervised learning, but can also be applied to some supervised learning tasks by incorporating task-specific labels or objectives; and (4) it can also adjust the number of encoder parameters based on computing resources. We evaluate our framework on several code-related tasks, and demonstrate its effectiveness and superiority over the state-of-the-art methods such as SourcererCC, Code2vec, and InferCode.
As researchers and practitioners apply Machine Learning to increasingly more software engineering problems, the approaches they use become more sophisticated. A lot of modern approaches utilize internal code structure in the form of an abstract syntax tree (AST) or its extensions: path-based representation, complex graph combining AST with additional edges. Even though the process of extracting ASTs from code can be done with different parsers, the impact of choosing a parser on the final model quality remains unstudied. Moreover, researchers often omit the exact details of extracting particular code representations. In this work, we evaluate two models, namely Code2Seq and TreeLSTM, in the method name prediction task backed by eight different parsers for the Java language. To unify the process of data preparation with different parsers, we develop SuperParser, a multi-language parser-agnostic library based on PathMiner. SuperParser facilitates the end-to-end creation of datasets suitable for training and evaluation of ML models that work with structural information from source code. Our results demonstrate that trees built by different parsers vary in their structure and content. We then analyze how this diversity affects the models' quality and show that the quality gap between the most and least suitable parsers for both models turns out to be significant. Finally, we discuss other features of the parsers that researchers and practitioners should take into account when selecting a parser along with the impact on the models' quality. The code of SuperParser is publicly available at https://doi.org/10.5281/zenodo.6366591. We also publish Java-norm, the dataset we use to evaluate the models: https://doi.org/10.5281/zenodo.6366599.
Machine Learning for Software Engineering (ML4SE) is an actively growing research area that focuses on methods that help programmers in their work. In order to apply the developed methods in practice, they need to achieve reasonable quality in order to help rather than distract developers. While the development of new approaches to code representation and data collection improves the overall quality of the models, it does not take into account the information that we can get from the project at hand. In this work, we investigate how the model's quality can be improved if we target a specific project. We develop a framework to assess quality improvements that models can get after fine-tuning for the method name prediction task on a particular project. We evaluate three models of different complexity and compare their quality in three settings: trained on a large dataset of Java projects, further fine-tuned on the data from a particular project, and trained from scratch on this data. We show that per-project fine-tuning can greatly improve the models' quality as they capture the project's domain and naming conventions. We open-source the tool we used for data collection, as well as the code to run the experiments: https://zenodo.org/record/6040745.
Neural code intelligence (CI) models are opaque black-boxes and offer little insight on the features they use in making predictions. This opacity may lead to distrust in their prediction and hamper their wider adoption in safety-critical applications. Recently, input program reduction techniques have been proposed to identify key features in the input programs to improve the transparency of CI models. However, this approach is syntax-unaware and does not consider the grammar of the programming language. In this paper, we apply a syntax-guided program reduction technique that considers the grammar of the input programs during reduction. Our experiments on multiple models across different types of input programs show that the syntax-guided program reduction technique is faster and provides smaller sets of key tokens in reduced programs. We also show that the key tokens could be used in generating adversarial examples for up to 65% of the input programs.
The code intelligence (CI) models are often black-box and do not offer any insights on the input features that they learn for making correct predictions. This opacity may lead to distrust in their prediction and hamper their wider adoption in safety-critical applications. In recent, the program reduction technique is widely being used to identify key input features in order to explain the prediction of CI models. The approach removes irrelevant parts from an input program and keeps the minimal snippets that a CI model needs to maintain its prediction. However, the state-of-the-art approaches mainly use a syntax-unaware program reduction technique that does not follow the syntax of programs, which adds significant overhead to the reduction of input programs and explainability of models. In this paper, we apply a syntax-guided program reduction technique that follows the syntax of input programs during reduction. Our experiments on multiple models across different types of input programs show that the syntax-guided program reduction technique significantly outperforms the syntax-unaware program reduction technique in reducing the size of input programs. Extracting key input features from reduced programs reveals that the syntax-guided reduced programs contain more label-specific key input features and are more vulnerable to adversarial transformation when renaming the key tokens in programs. These label-specific key input features may help to understand the reasoning of models' prediction from different perspectives and increase the trustworthiness to correct classification given by CI models.
There are several approaches for encoding source code in the input vectors of neural models. These approaches attempt to include various syntactic and semantic features of input programs in their encoding. In this paper, we investigate Code2Snapshot, a novel representation of the source code that is based on the snapshots of input programs. We evaluate several variations of this representation and compare its performance with state-of-the-art representations that utilize the rich syntactic and semantic features of input programs. Our preliminary study on the utility of Code2Snapshot in the code summarization and code classification tasks suggests that simple snapshots of input programs have comparable performance to state-of-the-art representations. Interestingly, obscuring input programs have insignificant impacts on the Code2Snapshot performance, suggesting that, for some tasks, neural models may provide high performance by relying merely on the structure of input programs.
Deep Neural Networks (DNNs) are increasingly being used in software engineering and code intelligence tasks. These are powerful tools that are capable of learning highly generalizable patterns from large datasets through millions of parameters. At the same time, their large capacity can render them prone to memorizing data points. Recent work suggests that the memorization risk manifests especially strongly when the training dataset is noisy, involving many ambiguous or questionable samples, and memorization is the only recourse. The goal of this paper is to evaluate and compare the extent of memorization and generalization in neural code intelligence models. It aims to provide insights on how memorization may impact the learning behavior of neural models in code intelligence systems. To observe the extent of memorization in models, we add random noise to the original training dataset and use various metrics to quantify the impact of noise on various aspects of training and testing. We evaluate several state-of-the-art neural code intelligence models and benchmarks based on Java, Python, and Ruby codebases. Our results highlight important risks: millions of trainable parameters allow the neural networks to memorize anything, including noisy data, and provide a false sense of generalization. We observed all models manifest some forms of memorization. This can be potentially troublesome in most code intelligence tasks where they rely on rather noise-prone and repetitive data sources, such as code from GitHub. To the best of our knowledge, we provide the first study to quantify memorization effects in the domain of software engineering and code intelligence systems. This work raises awareness and provides new insights into important issues of training neural models in code intelligence systems that are usually overlooked by software engineering researchers.
A wide range of code intelligence (CI) tools, powered by deep neural networks, have been developed recently to improve programming productivity and perform program analysis. To reliably use such tools, developers often need to reason about the behavior of the underlying models and the factors that affect them. This is especially challenging for tools backed by deep neural networks. Various methods have tried to reduce this opacity in the vein of "transparent/interpretable-AI". However, these approaches are often specific to a particular set of network architectures, even requiring access to the network's parameters. This makes them difficult to use for the average programmer, which hinders the reliable adoption of neural CI systems. In this paper, we propose a simple, model-agnostic approach to identify critical input features for models in CI systems, by drawing on software debugging research, specifically delta debugging. Our approach, SIVAND, uses simplification techniques that reduce the size of input programs of a CI model while preserving the predictions of the model. We show that this approach yields remarkably small outputs and is broadly applicable across many model architectures and problem domains. We find that the models in our experiments often rely heavily on just a few syntactic features in input programs. We believe that SIVAND's extracted features may help understand neural CI systems' predictions and learned behavior.
The application of machine learning algorithms to source code has grown in the past years. Since these algorithms are quite sensitive to input data, it is not surprising that researchers experiment with input representations. Nowadays, a popular starting point to represent code is abstract syntax trees (ASTs). Abstract syntax trees have been used for a long time in various software engineering domains, and in particular in IDEs. The API of modern IDEs allows to manipulate and traverse ASTs, resolve references between code elements, etc. Such algorithms can enrich ASTs with new data and therefore may be useful in ML-based code analysis. In this work, we present PSIMiner - a tool for processing PSI trees from the IntelliJ Platform. PSI trees contain code syntax trees as well as functions to work with them, and therefore can be used to enrich code representation using static analysis algorithms of modern IDEs. To showcase this idea, we use our tool to infer types of identifiers in Java ASTs and extend the code2seq model for the method name prediction problem.
Learning from source code usually requires a large amount of labeled data. Despite the possible scarcity of labeled data, the trained model is highly task-specific and lacks transferability to different tasks. In this work, we present effective pre-training strategies on top of a novel graph-based code representation, to produce universal representations for code. Specifically, our graph-based representation captures important semantics between code elements (e.g., control flow and data flow). We pre-train graph neural networks on the representation to extract universal code properties. The pre-trained model then enables the possibility of fine-tuning to support various downstream applications. We evaluate our model on two real-world datasets -- spanning over 30M Java methods and 770K Python methods. Through visualization, we reveal discriminative properties in our universal code representation. By comparing multiple benchmarks, we demonstrate that the proposed framework achieves state-of-the-art results on method name prediction and code graph link prediction.
Building deep learning models on source code has found many successful software engineering applications, such as code search, code comment generation, bug detection, code migration, and so on. Current learning techniques, however, have a major drawback that these models are mostly trained on datasets labeled for particular downstream tasks, and code representations may not be suitable for other tasks. While some techniques produce representations from unlabeled code, they are far from satisfactory when applied to downstream tasks. Although certain techniques generate representations from unlabeled code when applied to downstream tasks they are far from satisfactory. This paper proposes InferCode to overcome the limitation by adapting the self-supervised learning mechanism to build source code model. The key novelty lies in training code representations by predicting automatically identified subtrees from the context of the ASTs. Subtrees in ASTs are treated with InferCode as the labels for training code representations without any human labeling effort or the overhead of expensive graph construction, and the trained representations are no longer tied to any specific downstream tasks or code units. We trained an InferCode model instance using the Tree-based CNN as the encoder of a large set of Java code and applied it to downstream unsupervised tasks such as code clustering, code clone detection, cross-language code search or reused under a transfer learning scheme to continue training the model weights for supervised tasks such as code classification and method name prediction. Compared to previous code learning techniques applied to the same downstream tasks, such as Code2Vec, Code2Seq, ASTNN, higher performance results are achieved using our pre-trained InferCode model with a significant margin for most tasks including those involving different programming languages.
We propose Corder, a self-supervised contrastive learning framework for source code model. Corder is designed to alleviate the need of labeled data for code retrieval and code summarization tasks. The pre-trained model of Corder can be used in two ways: (1) it can produce vector representation of code which can be applied to code retrieval tasks that do not have labeled data; (2) it can be used in a fine-tuning process for tasks that might still require label data such as code summarization. The key innovation is that we train the source code model by asking it to recognize similar and dissimilar code snippets through a contrastive learning objective. To do so, we use a set of semantic-preserving transformation operators to generate code snippets that are syntactically diverse but semantically equivalent. Through extensive experiments, we have shown that the code models pretrained by Corder substantially outperform the other baselines for code-to-code retrieval, text-to-code retrieval, and code-to-text summarization tasks.
Source code representations are key in applying machine learning techniques for processing and analyzing programs. A popular approach in representing source code is neural source code embeddings that represents programs with high-dimensional vectors computed by training deep neural networks on a large volume of programs. Although successful, there is little known about the contents of these vectors and their characteristics. In this paper, we present our preliminary results towards better understanding the contents of code2vec neural source code embeddings. In particular, in a small case study, we use the code2vec embeddings to create binary SVM classifiers and compare their performance with the handcrafted features. Our results suggest that the handcrafted features can perform very close to the highly-dimensional code2vec embeddings, and the information gains are more evenly distributed in the code2vec embeddings compared to the handcrafted features. We also find that the code2vec embeddings are more resilient to the removal of dimensions with low information gains than the handcrafted features. We hope our results serve a stepping stone toward principled analysis and evaluation of these code representations.
With the prevalence of publicly available source code repositories to train deep neural network models, neural program models can do well in source code analysis tasks such as predicting method names in given programs that cannot be easily done by traditional program analysis techniques. Although such neural program models have been tested on various existing datasets, the extent to which they generalize to unforeseen source code is largely unknown. Since it is very challenging to test neural program models on all unforeseen programs, in this paper, we propose to evaluate the generalizability of neural program models with respect to semantic-preserving transformations: a generalizable neural program model should perform equally well on programs that are of the same semantics but of different lexical appearances and syntactical structures. We compare the results of various neural program models for the method name prediction task on programs before and after automated semantic-preserving transformations. We use three Java datasets of different sizes and three state-of-the-art neural network models for code, namely code2vec, code2seq, and GGNN, to build nine such neural program models for evaluation. Our results show that even with small semantically preserving changes to the programs, these neural program models often fail to generalize their performance. Our results also suggest that neural program models based on data and control dependencies in programs generalize better than neural program models based only on abstract syntax trees. On the positive side, we observe that as the size of the training dataset grows and diversifies the generalizability of correct predictions produced by the neural program models can be improved too. Our results on the generalizability of neural program models provide insights to measure their limitations and provide a stepping stone for their improvement.
Recent work learns contextual representations of source code by reconstructing tokens from their context. For downstream semantic understanding tasks like summarizing code in English, these representations should ideally capture program functionality. However, we show that the popular reconstruction-based BERT model is sensitive to source code edits, even when the edits preserve semantics. We propose ContraCode: a contrastive pre-training task that learns code functionality, not form. ContraCode pre-trains a neural network to identify functionally similar variants of a program among many non-equivalent distractors. We scalably generate these variants using an automated source-to-source compiler as a form of data augmentation. Contrastive pre-training improves JavaScript summarization and TypeScript type inference accuracy by 2% to 13%. We also propose a new zero-shot JavaScript code clone detection dataset, showing that ContraCode is both more robust and semantically meaningful. On it, we outperform RoBERTa by 39% AUROC in an adversarial setting and up to 5% on natural code.
Learning distributed representations of source code has been a challenging task for machine learning models. Earlier works treated programs as text so that natural language methods can be readily applied. Unfortunately, such approaches do not capitalize on the rich structural information possessed by source code. Of late, Graph Neural Network (GNN) was proposed to learn embeddings of programs from their graph representations. Due to the homogeneous and expensive message-passing procedure, GNN can suffer from precision issues, especially when dealing with programs rendered into large graphs. In this paper, we present a new graph neural architecture, called Graph Interval Neural Network (GINN), to tackle the weaknesses of the existing GNN. Unlike the standard GNN, GINN generalizes from a curated graph representation obtained through an abstraction method designed to aid models to learn. In particular, GINN focuses exclusively on intervals for mining the feature representation of a program, furthermore, GINN operates on a hierarchy of intervals for scaling the learning to large graphs. We evaluate GINN for two popular downstream applications: variable misuse prediction and method name prediction. Results show in both cases GINN outperforms the state-of-the-art models by a comfortable margin. We have also created a neural bug detector based on GINN to catch null pointer deference bugs in Java code. While learning from the same 9,000 methods extracted from 64 projects, GINN-based bug detector significantly outperforms GNN-based bug detector on 13 unseen test projects. Next, we deploy our trained GINN-based bug detector and Facebook Infer to scan the codebase of 20 highly starred projects on GitHub. Through our manual inspection, we confirm 38 bugs out of 102 warnings raised by GINN-based bug detector compared to 34 bugs out of 129 warnings for Facebook Infer.