Code search with input/output queries: Generalizing, ranking, and assessment

Stolee, Kathryn T.; Elbaum, Sebastian; Dwyer, Matthew B.

doi:10.1016/j.jss.2015.04.081

Cited by 38 publications

(7 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Satsy [32] utilizes symbolic execution engines to support the encoding work on multi-path programs, and then ranks code candidates based on the matching degree between the query and encoded program paths. Quebio [33] encodes code snippets into path constraints via symbolic analysis and returns the fulfilled programs.…”

Section: B Constraint-based Code Searchmentioning

confidence: 99%

Multimodal Representation for Neural Code Search

Gu,

Chen,

Monperrus

2021

Preprint

View full text Add to dashboard Cite

Semantic code search is about finding semantically relevant code snippets for a given natural language query. In the state-of-the-art approaches, the semantic similarity between code and query is quantified as the distance of their representation in the shared vector space. In this paper, to improve the vector space, we introduce tree-serialization methods on a simplified form of AST and build the multimodal representation for the code data. We conduct extensive experiments using a single corpus that is large-scale and multi-language: CodeSearchNet. Our results show that both our tree-serialized representations and multimodal learning model improve the performance of code search. Last, we define intuitive quantification metrics oriented to the completeness of semantic and syntactic information of the code data, to help understand the experimental findings.

show abstract

Section: B Constraint-based Code Searchmentioning

confidence: 99%

Multimodal Representation for Neural Code Search

Gu,

Chen,

Monperrus

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…According to the fact that black-box behavior is representable through program input-output, several works incorporate program input-output as their query. Users can provide either test cases (Lemos, et al, 2007), input-output data types (Thummalapenta & Xie, 2007;Reiss, 2009), or input-output query model (Stolee, et al, 2016) to refine their retrieval result. On the one hand, test cases are incorporated for retrieving only source codes which output is similar to given test-case output while its respective input is given.…”

Section: Related Workmentioning

confidence: 99%

LANGUAGE-AGNOSTIC SOURCE CODE RETRIEVAL USING KEYWORD & IDENTIFIER LEXICAL PATTERN

Karnalim¹

2018

IJSECS

View full text Add to dashboard Cite

This paper proposes a language-agnostic approach to retrieve source codes. Language-agnostic means that our approach could incorporate various programming languages automatically since it does not rely directly on programming language structure. Instead, it incorporates Keyword & Identifier lexical pattern. Keyword & Identifier lexical pattern is selected as our main concern based on twofold. First, Keyword & Identifier is the most declarative token type for representing author intention on source code. Second, Keyword & Identifier lexical patterns are typically similar in most programming language.

show abstract

“…Other approaches use symbolic execution [218], formal logic and theorem prover to identify relevant components [184]. Other semantic code search engines execute test cases on candidate code [121,195].…”

Section: Code Recommendationmentioning

confidence: 99%

Exploring regularities in software with statistical models and their applications

Nguyen¹

View full text Add to dashboard Cite

contribution to research and x my mistakes. I would also like to thank my committee members for their guide and eort of sharing ideas. My work would not become successful without collaboration of my colleagues, especially Dr. Hoan Nguyen, Dr. Tung Nguyen and Hung Nguyen. They did not only support my work, contribute their thoughts, but also share their knowledge to develop my skills and my understanding about SE. Last but foremost, I would like to thanks my parents and other family's members. My parents were the rst to encourage me to follow my PhD degree at Iowa State University. Only with them, I could nd the best motivation to follow my research career, to overcome many barriers that I met. My father, who denied his PhD work to take care of his family, always want me to nish the degree that he could not. My father passed away and could not wait until these days to see I completing his desire. However, the way of his thinking, his excitement in nishing his jobs, his love to contribute to the development of society is the model that I want to follow. I feel proud that I can partly learn his characteristics and use them to deal with a long, hard PhD student career. My mother, who spend all of her love to my family and me, xvi ABSTRACT Software systems are becoming popular. They are used with dierent platforms for dierent applications. Software systems are developed with support from programming languages, which help developers work conveniently. Programming languages can have dierent paradigms with dierent form, syntactic structures, keywords, representation ways. In many cases, however, programming languages are similar in dierent important aspects: 1. They are used to support description of specic tasks, 2. Source codes are written in languages and includes a limit set of distinctive tokens, many tokens are repeated like keywords, function calls, and 3. They follow specic syntactic rules to make machine understanding. Those points also reect the similarity between programming language and natural language. Due to its critical role in many applications, natural language processing (NLP) has been studied much and given many promising results like automatic cross-language translation, speech-to-text, information searching, etc. It is interesting to observe if there are similar characteristics between natural language and programming language and whether techniques in NLP can be reused for programming language processing? Recent works in software engineering (SE) shows that their similarities between NLP and programming language processing and techniques in NLP can be reused for PLP. This dissertation introduces my works with contributions in study of characteristics of programming languages, the models which employed them and the main applications that show the usefulness of the proposed models. Study in both three aspects has draw interests from software engineering community and received awards due to their innovation and applicability I hope that this dissertation will bring a systematic view of how advantage t...

show abstract

Code search with input/output queries: Generalizing, ranking, and assessment

Cited by 38 publications

References 35 publications

Multimodal Representation for Neural Code Search

Multimodal Representation for Neural Code Search

LANGUAGE-AGNOSTIC SOURCE CODE RETRIEVAL USING KEYWORD & IDENTIFIER LEXICAL PATTERN

Exploring regularities in software with statistical models and their applications

Contact Info

Product

Resources

About