Abstract:To help programmers find proper API methods and learn API usages, researchers have proposed various code search engines. Given an API of interest, a code search engine can retrieve its code samples from online software repositories. Through such tools, Internet code becomes a major resource for learning API usages. Besides Internet code, local library code also contains API usages, and researchers have found that library code contains many API usages that are more concise than those in client code. As samples from a code search engine are typically client code, it is interesting to explore the API usages inside library code, but the samples inside library code contain internal method invocations. If an empirical study does not remove them from API usages, it can significantly overestimate the API usages from library code. Due to this challenge, no prior study has ever analyzed API usages inside libraries, and many research questions are still open. For example, how many API usages are there inside libraries? The answers are useful to motivate future research on APIs and code search engines. The internal usages in library code will introduce compilation errors when they are directly called from the client side. To support the exploration of the above questions, in this paper, we propose CodeEx that extracts Internet code samples from a popular code search engine and local code samples by removing internal usages from library code. With the support of CodeEx, we conduct the first empirical study on API usages of five libraries, and summarize our results into six findings as the answers to five research questions. Our results are useful for researchers to motivate their future research. For example, our results show that although code samples from library code are only half of those from the code search engine, they cover 4.0 times more API classes, 4.7 times more API methods, and 3.0 times more call sequences. Meanwhile, in a controlled experiment, we compare their effectiveness in assisting programming. We find that more API usages do not lead to more complete tasks, and it highlights the importance of code recommendation approaches.

APIExample: An effective web search based usage example recommendation system for java APIs.

API method recommendation without worrying about the task-API knowledge gap.

An Exploratory Study of API Usage Examples on the Web

APIReal: an API Recognition and Linking Approach for Online Developer Forums

Generating API-usage Example for Project Developers

Apiuaas: A Reference Architecture For Facilitating Api Usage From A Data Analytics Perspective

BIKER: a Tool for Bi-information Source Based API Method Recommendation

RESTful API Analysis, Recommendation, and Client Code Retrieval

Web APIs: Features, Issues, and Expectations – A Large-Scale Empirical Study of Web APIs from Two Publicly Accessible Registries Using Stack Overflow and a User Survey

Mining API Usage Examples from Test Code

API Recommendation for Novice Programmers: Build a Bridge of Query-Task Knowledge Gap

Suggesting API Usage to Novice Programmers with the Example Guru

An Efficient API Search Algorithm

Automatic Parameter Recommendation for Practical API Usage

Detecting and Ranking API Usage Pattern in Large Source Code Repository: A LFM Based Approach.

A Fusion of Java Domain Knowledge Base and Siamese Network for Java API Recommendation

MAPO: mining and recommending api usage patterns

An Empirical Study on API Usages from Code Search Engine and Local Library

Application programming interface recommendation according to the knowledge indexed by app feature mined from app stores

Documenting API Input/Output Examples

API recommendation for the development of Android App features based on the knowledge mined from App stores