Solving the Search for Source Code

Abstract

Programmers frequently search for source code to reuse using keyword searches. The search effectiveness in facilitating reuse, however, depends on the programmer's ability to specify a query that captures how the desired code may have been implemented. Further, the results often include many irrelevant matches that must be filtered manually. More semantic search approaches could address these limitations, yet existing approaches are either not flexible enough to find approximate matches or require the programmer to define complex specifications as queries. We propose a novel approach to semantic code search that addresses several of these limitations and is designed for queries that can be described using a concrete input/output example. In this approach, programmers write lightweight specifications as inputs and expected output examples. Unlike existing approaches to semantic search, we use an SMT solver to identify programs or program fragments in a repository, which have been automatically transformed into constraints using symbolic analysis, that match the programmer-provided specification. We instantiated and evaluated this approach in subsets of three languages, the Java String library, Yahoo! Pipes mashup language, and SQL select statements, exploring its generality, utility, and trade-offs. The results indicate that this approach is effective at finding relevant code, can be used on its own or to filter results from keyword searches to increase search precision, and is adaptable to find approximate matches and then guide modifications to match the user specifications when exact matches do not already exist. These gains in precision and flexibility come at the cost of performance, for which underlying factors and mitigation strategies are identified.

Understanding the impact of support for iteration on code search

“…Surveys asking programmers why they search for code on the Internet have been conducted by Sim et al [107], Sadowski et al [97], Stolee et al [116], and Hucka and Graham [53]. In Sim et al's survey, all 69 participants reported that they searched on the Internet either for code to use as a reference or for code that they can use as-is.…”

“…To address this, our approach was simply to do a literature search to determine how many results people, in general, look at. We discovered that, in general, they do not look past the top ten results from a query [21], [56], [117]. As such, we decided to set , the size of , to be 10.…”

“…In a counterbalanced design, each participant was assigned two of the four code search engines and used them to complete eight different code search tasks, alternating the search engine used with each task. The search tasks covered a range from those that are more open ended to those that are more well defined, in order to represent a variety of real-world internet-scale code search scenarios [7], [48], [50], [76], [107], [116]. After each search task, each participant was asked to rate their experience in using the assigned search engine for that task.…”

“…solver that takes well-defined constraints on the input/output of the code, and the search engine then uses an SMT solver to find code satisfying those constraints [116]. There is also research focusing on locating more specific kinds of functionality, such as code that transforms a variable of one type into another.…”

Goal-Based Analytic Composition for On-and Off-line Execution at Scale

“…Stolee et al [25] examined the use of semantic models of source code as an indexing strategy to help find a block of code that will pass a set of test cases -one form of goal-based search. Such semantic searches have additionally been used in web service composition [26], [27].…”