Incorporating Code Structure and Quality in Deep Code Search

JavaScript frameworks, such as jQuery, are widely used for developing web applications. To facilitate using these JavaScript frameworks to implement a feature (e.g., functionality), a large number of programmers often search for code snippets that implement the same or similar feature. However, existing code search approaches tend to be ineffective, without taking into account the fact that JavaScript code snippets often implement a feature based on various relationships (e.g., sequencing, condition, and callback relationships) among the invoked framework API methods. To address this issue, we present a novel Relationship-Aware Code Search (RACS) approach for finding code snippets that use JavaScript frameworks to implement a specific feature. In advance, RACS collects a large number of code snippets that use some JavaScript frameworks, mines API usage patterns from the collected code snippets, and represents the mined patterns with method call relationship (MCR) graphs, which capture framework API methods' signatures and their relationships. Given a natural language (NL) search query issued by a programmer, RACS conducts NL processing to automatically extract an action relationship (AR) graph, which consists of actions and their relationships inferred from the query. In this way, RACS reduces code search to the problem of graph search: finding similar MCR graphs for a given AR graph. We conduct evaluations against representative real-world jQuery questions posted on Stack Overflow, based on 308,294 code snippets collected from over 81,540 files on the Internet. The evaluation results show the effectiveness of RACS: the top 1 snippet produced by RACS matches the target code snippet for 46% questions, compared to only 4% achieved by a relationship-oblivious approach.

Recently, deep learning methods have become mainstream in code search since they do better at capturing semantic correlations between code snippets and search queries and have promising performance. However, code snippets have diverse information from different dimensions, such as business logic, specific algorithm, and hardware communication, so it is hard for a single code representation module to cover all the perspectives. On the other hand, as a specific query may focus on one or several perspectives, it is difficult for a single query representation module to represent different user intents. In this paper, we propose MuCoS, a multi-model ensemble learning architecture for semantic code search. It combines several individual learners, each of which emphasizes a specific perspective of code snippets. We train the individual learners on different datasets which contain different perspectives of code information, and we use a data augmentation strategy to get these different datasets. Then we ensemble the learners to capture comprehensive features of code snippets. The experiments show that MuCoS has better results than the existing state-of-the-art methods. Our source code and data are anonymously available at https://github.com/Xzh0u/MuCoS.

Software developers all over the world use Stack Overflow (SO) to interact and exchange code snippets. Research also uses SO to harvest code snippets for use with recommendation systems. However, previous work has shown that code on SO may have quality issues, such as security or license problems. We analyse Python code on SO to determine its coding style compliance. From 1,962,535 code snippets tagged with 'python', we extracted 407,097 snippets of at least 6 statements of Python code. Surprisingly, 93.87% of the extracted snippets contain style violations, with an average of 0.7 violations per statement and a huge number of snippets with a considerably higher ratio. Researchers and developers should, therefore, be aware that code snippets on SO may not representative of good coding style. Furthermore, while user reputation seems to be unrelated to coding style compliance, for posts with vote scores in the range between -10 and 20, we found a strong correlation (r = -0.87, p <; 10^-7) between the vote score a post received and the average number of violations per statement for snippets in such posts.

As the popularity of modern social coding paradigm such as Stack Overflow grows, its potential security risks increase as well (e.g., insecure codes could be easily embedded and distributed). To address this largely overlooked issue, in this paper, we bring an important new insight to exploit social coding properties in addition to code content for automatic detection of insecure code snippets in Stack Overflow. To determine if the given code snippets are insecure, we not only analyze the code content, but also utilize various kinds of relations among users, badges, questions, answers, code snippets and keywords in Stack Overflow. To model the rich semantic relationships, we first introduce a structured heterogeneous information network (HIN) for representation and then use meta-path based approach to incorporate higher-level semantics to build up relatedness over code snippets. Later, we propose a novel network embedding model named snippet2vec for representation learning in HIN where both the HIN structures and semantics are maximally preserved. After that, a multi-view fusion classifier is constructed for insecure code snippet detection. To the best of our knowledge, this is the first work utilizing both code content and social coding properties to address the code security issues in modern software coding platforms. Comprehensive experiments on the data collections from Stack Overflow are conducted to validate the effectiveness of the developed system ICSD which integrates our proposed method in insecure code snippet detection by comparisons with alternative approaches.

To implement a program functionality, developers can reuse previously written code snippets by searching through a large-scale codebase. Over the years, many code search tools have been proposed to help developers. The existing approaches often treat source code as textual documents and utilize information retrieval models to retrieve relevant code snippets that match a given query. These approaches mainly rely on the textual similarity between source code and natural language query. They lack a deep understanding of the semantics of queries and source code. In this paper, we propose a novel deep neural network named CODEnn (Code-Description Embedding Neural Network). Instead of matching text similarity, CODEnn jointly embeds code snippets and natural language descriptions into a high-dimensional vector space, in such a way that code snippet and its corresponding description have similar vectors. Using the unified vector representation, code snippets related to a natural language query can be retrieved according to their vectors. Semantically related words can also be recognized and irrelevant/noisy keywords in queries can be handled. As a proof-of-concept application, we implement a code search tool named DeepCS using the proposed CODEnn model. We empirically evaluate DeepCS on a large scale codebase collected from GitHub. The experimental results show that our approach can effectively retrieve relevant code snippets and outperforms previous techniques.

Developers are usually unaware of the quality and lineage of information available on popular Web resources, leading to potential maintenance problems and license violations when reusing code snippets from these resources. In this paper, we study the duplication of code snippets between two popular sources of software development information: the Stack Overflow QA a significant number (31%) of answers that contained a duplicate code block were chosen as the accepted answer. Qualitative analysis reveals that developers commonly use Stack Overflow to ask clarifying questions about code they reused from tutorials, and copy code snippets from tutorials to provide answers to questions.

Developers often rely on online resources, such as Stack Overflow (SO), to seek assistance for programming tasks. To facilitate effective search and resource discovery, manual tagging of questions and posts with the appropriate programming language is essential. However, accurate tagging is not consistently achieved, leading to the need for the automated classification of code snippets into the correct programming language as a tag. In this study, we introduce a novel approach to automated classification of code snippets from Stack Overflow (SO) posts into programming languages using generative pre-trained transformers (GPT). Our method, which does not require additional training on labeled data or dependency on pre-existing labels, classifies 224,107 code snippets into 19 programming languages. We employ the text-davinci-003 model of ChatGPT-3.5 and postprocess its responses to accurately identify the programming language. Our empirical evaluation demonstrates that our GPT-based model (SCC-GPT) significantly outperforms existing methods, achieving a median F1-score improvement that ranges from +6% to +31%. These findings underscore the effectiveness of SCC-GPT in enhancing code snippet classification, offering a cost-effective and efficient solution for developers who rely on SO for programming assistance.

Code search methods, especially those that allow programmers to raise queries in a natural language, plays an important role in software development. It helps to improve programmers' productivity by returning sample code snippets from the Internet and/or source-code repositories for their natural-language queries. Meanwhile, there are many code search methods in the literature that support natural-language queries. Difficulties exist in recognizing the strengths and weaknesses of each method and choosing the right one for different usage scenarios, because (1) the implementations of those methods and the datasets for evaluating them are usually not publicly available, and (2) some methods leverage different training datasets or auxiliary data sources and thus their effectiveness cannot be fairly measured and may be negatively affected in practical uses. To build a common ground for measuring code search methods, this paper builds CosBench, a dataset that consists of 1000 projects, 52 code-independent natural-language queries with ground truths, and a set of scripts for calculating four metrics on code research results. We have evaluated four IR (Information Retrieval)-based and two DL (Deep Learning)-based code search methods on CosBench. The empirical evaluation results clearly show the usefulness of the CosBench dataset and various strengths of each code search method. We found that DL-based methods are more suitable for queries on reusing code, and IR-based ones for queries on resolving bugs and learning API uses.

To accelerate software development, developers frequently search and reuse existing code snippets from a large-scale codebase, e.g., GitHub. Over the years, researchers proposed many information retrieval (IR)-based models for code search, but they fail to connect the semantic gap between query and code. An early successful deep learning (DL)-based model DeepCS solved this issue by learning the relationship between pairs of code methods and corresponding natural language descriptions. Two major advantages of DeepCS are the capability of understanding irrelevant/noisy keywords and capturing sequential relationships between words in query and code. In this article, we proposed an IR-based model CodeMatcher that inherits the advantages of DeepCS (i.e., the capability of understanding the sequential semantics in important query words), while it can leverage the indexing technique in the IR-based model to accelerate the search response time substantially. CodeMatcher first collects metadata for query words to identify irrelevant/noisy ones, then iteratively performs fuzzy search with important query words on the codebase that is indexed by the Elasticsearch tool and finally reranks a set of returned candidate code according to how the tokens in the candidate code snippet sequentially matched the important words in a query. We verified its effectiveness on a large-scale codebase with ~41K repositories. Experimental results showed that CodeMatcher achieves an MRR (a widely used accuracy measure for code search) of 0.60, outperforming DeepCS, CodeHow, and UNIF by 82%, 62%, and 46%, respectively. Our proposed model is over 1.2K times faster than DeepCS. Moreover, CodeMatcher outperforms two existing online search engines (GitHub and Google search) by 46% and 33%, respectively, in terms of MRR. We also observed that: fusing the advantages of IR-based and DL-based models is promising; improving the quality of method naming helps code search, since method name plays an important role in connecting query and code.

Searching and reusing online code is a common activity in software development. Meanwhile, like many general-purposed searches, code search also faces the session search problem: in a code search session, the user needs to iteratively search for code snippets, exploring new code snippets that meet his/her needs and/or making some results highly ranked. This paper presents Cosoch, a reinforcement learning approach to session search of code documents (code snippets with textual explanations). Cosoch is aimed at generating a session that reveals user intentions, and correspondingly searching and reranking the resulting documents. More specifically, Cosoch casts a code search session into a Markov decision process, in which rewards measuring the relevances between the queries and the resulting code documents guide the whole session search. We have built a dataset, say CosoBe, from StackOverflow, containing 103 code search sessions with 378 pieces of user feedback. We have also evaluated Cosoch on CosoBe. The evaluation results show that Cosoch achieves an average NDCG@3 score of 0.7379, outperforming StackOverflow by 21.3%.

Past studies have proposed solutions that analyze Stack Overflow content to help users find desired information or aid various downstream software engineering tasks. A common step performed by those solutions is to extract suitable representations of posts; typically, in the form of meaningful vectors. These vectors are then used for different tasks, for example, tag recommendation, relatedness prediction, post classification, and API recommendation. Intuitively, the quality of the vector representations of posts determines the effectiveness of the solutions in performing the respective tasks. In this work, to aid existing studies that analyze Stack Overflow posts, we propose a specialized deep learning architecture Post2Vec which extracts distributed representations of Stack Overflow posts. Post2Vec is aware of different types of content present in Stack Overflow posts, i.e., title, description, and code snippets, and integrates them seamlessly to learn post representations. Tags provided by Stack Overflow users that serve as a common vocabulary that captures the semantics of posts are used to guide Post2Vec in its task. To evaluate the quality of Post2Vec's deep learning architecture, we first investigate its end-to-end effectiveness in tag recommendation task. The results are compared to those of state-of-the-art tag recommendation approaches that also employ deep neural networks. We observe that Post2Vec achieves 15-25 percent improvement in terms of F1-score@5 at a lower computational cost. Moreover, to evaluate the value of representations learned by Post2Vec, we use them for three other tasks, i.e., relatedness prediction, post classification, and API recommendation. We demonstrate that the representations can be used to boost the effectiveness of state-of-the-art solutions for the three tasks by substantial margins (by 10, 7, and 10 percent in terms of F1-score, F1-score, and correctness, respectively). We release our replication package at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/maxxbw/Post2Vec</uri> .

As code search permeates most activities in software development,code-to-code search has emerged to support using code as a query and retrieving similar code in the search results. Applications include duplicate code detection for refactoring, patch identification for program repair, and language translation. Existing code-to-code search tools rely on static similarity approaches such as the comparison of tokens and abstract syntax trees (AST) to approximate dynamic behavior, leading to low precision. Most tools do not support cross-language code-to-code search, and those that do, rely on machine learning models that require labeled training data. We present Code-to-Code Search Across Languages (COSAL), a cross-language technique that uses both static and dynamic analyses to identify similar code and does not require a machine learning model. Code snippets are ranked using non-dominated sorting based on code token similarity, structural similarity, and behavioral similarity. We empirically evaluate COSAL on two datasets of 43,146Java and Python files and 55,499 Java files and find that 1) code search based on non-dominated ranking of static and dynamic similarity measures is more effective compared to single or weighted measures; and 2) COSAL has better precision and recall compared to state-of-the-art within-language and cross-language code-to-code search tools. We explore the potential for using COSAL on large open-source repositories and discuss scalability to more languages and similarity metrics, providing a gateway for practical,multi-language code-to-code search.

Modern software systems are increasingly including machine learning (ML) as an integral component. However, we do not yet understand the difficulties faced by software developers when learning about ML libraries and using them within their systems. To fill that gap this thesis reports on a detailed (manual) examination of 3,243 highly-rated Q&A posts related to ten ML libraries, namely Tensorflow, Keras, scikitlearn, Weka, Caffe, Theano, MLlib, Torch, Mahout, and H2O, on Stack Overflow, a popular online technical Q&A forum. Our findings reveal the urgent need for software engineering (SE) research in this area. The second part of the thesis particularly focuses on understanding the Deep Neural Network (DNN) bug characteristics. We study 2,716 high-quality posts from Stack Overflow and 500 bug fix commits from Github about five popular deep learning libraries Caffe, Keras, Tensorflow, Theano, and Torch to understand the types of bugs, their root causes and impacts, bug-prone stage of deep learning pipeline as well as whether there are some common antipatterns found in this buggy software. While exploring the bug characteristics, our findings imply that repairing software that uses DNNs is one such unmistakable SE need where automated tools could be beneficial; however, we do not fully understand challenges to repairing and patterns that are utilized when manually repairing DNNs. So, the third part of this thesis presents a comprehensive study of bug fix patterns to address these questions. We have studied 415 repairs from Stack Overflow and 555 repairs from Github for five popular deep learning libraries Caffe, Keras, Tensorflow, Theano, and Torch to understand challenges in repairs and bug repair patterns. Our key findings reveal that DNN bug fix patterns are distinctive compared to traditional bug fix patterns and the most common bug fix patterns are fixing data dimension and neural network connectivity. Finally, we propose an automatic technique to detect ML Application Programming Interface (API) misuses. We started with an empirical study to understand ML API misuses. Our study shows that ML API misuse is prevalent and distinct compared to non-ML API misuses. Inspired by these findings, we contributed Amimla (Api Misuse In Machine Learning Apis) an approach and a tool for ML API misuse detection. Amimla relies on several technical innovations. First, we proposed an abstract representation of ML pipelines to use in misuse detection. Second, we proposed an abstract representation of neural networks for deep learning related APIs. Third, we have developed a representation strategy for constraints on ML APIs. Finally, we have developed a misuse detection strategy for both single and multi-APIs. Our experimental evaluation shows that Amimla achieves a high average accuracy of ∼80% on two benchmarks of misuses from Stack Overflow and Github.

AI-generated code has become an integral part of the mainstream developer workflow today. However, in community-driven platforms like Stack Overflow (SO), where trust, authorship, and credibility are important, it can lead to serious complications. While recent studies have focused on detecting AI-generated code, they have mostly worked with long code samples from repositories and assignments. In contrast, code snippets on SO are often small and context-specific, and thus may prove more challenging for detection. Moreover, another aspect overlooked in prior studies concerns recognizing adversarially prompted AI code deliberately crafted to resemble human-written code. To address these limitations, we have first introduced a large-scale dataset comprising 3500 pairs of SO and ChatGPT answers, along with a curated set of 4500 adversarially prompted AI responses. Next, we evaluate existing code language models over this newly curated dataset. Our evaluation shows that existing models perform well on standard AI answers but fail to detect adversarial ones. Finally, to improve detection, we propose an ensemble approach combining stylometric features of code along with the code embeddings. Our approach shows consistent improvements across multiple models and improves resistance to adversarial prompted code. Our overall findings open promising directions for future research into understanding the nuances of AI code detection with adversarial prompting and code stylometry.

Developers reuse programming-related knowledge (e.g., code snippets) on Q&A sites (e.g., Stack Overflow) that functionally matches the programming problems they encounter in their development. Despite extensive research on Q&A sites, being a high-level and important type of development-related knowledge, architectural solutions (e.g., architecture tactics) and their reuse are rarely explored. To fill this gap, we conducted a mixed-methods study that includes a mining study and a survey study. For the mining study, we mined 984 commits and issues (i.e., 821 commits and 163 issues) from 893 Open-Source Software (OSS) projects on GitHub that explicitly referenced architectural solutions from Stack Overflow (SO) and Software Engineering Stack Exchange (SWESE). For the survey study, we identified practitioners involved in the reuse of these architectural solutions and surveyed 227 of them to further understand how practitioners reuse architectural solutions from Q&A sites in their OSS development. Our main findings are that: (1) OSS practitioners reuse architectural solutions from Q&A sites to solve a large variety (15 categories) of architectural problems, wherein <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Component design issue</i>, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Architectural anti-pattern</i>, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Security issue</i> are dominant; (2) Seven categories of architectural solutions from Q&A sites have been reused to solve those problems, among which <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Architectural refactoring</i>, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Use of frameworks</i>, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Architectural tactic</i> are the three most reused architectural solutions; (3) OSS developers often rely on ad hoc ways (e.g., informal, improvised, or unstructured approaches) to reuse architectural solutions from SO, drawing on personal experience and intuition rather than standardized or systematic practices; (4) Reusing architectural solutions from SO comes with a variety of challenges, e.g., OSS practitioners complain that they need to spend significant time to adapt such architectural solutions to address design concerns raised in their OSS development, and it is challenging to reuse architectural solutions that are not tailored to the design context of their OSS projects. Our findings pave the way for future research directions, including the design and development of approaches and tools (such as IDE plugin tools) to facilitate the reuse of architectural solutions from Q&A sites, and could also be used to offer guidelines to practitioners when they contribute architectural solutions to Q&A sites. Our dataset is publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://doi.org/10.5281/zenodo.10936098</uri>.