Crossing Ecosystems: An Exploration of Analogical Packages Across Diverse Package Ecosystems

Abstract

With an increase in the diversity in technology stacks and third-party library usages, developers may inevitably will face the need to switch programming language and their subsequent libraries specific to that language. In this scenario, developers face the issue of finding similar libraries that can provide comparable features to the libraries that they are already familiar with. One alternative is to have an analogical library package that can span across these different language technology stacks. Hence, we introduce a cross-ecosystem package, which is a package that span across different library ecosystems. To understand this phenomena, we perform an empirical case study of four diverse library ecosystems (i.e., CRAN that serves R, Maven for Java, PyPI for Python, and RubyGems for Ruby). We mined and collected 291,272 packages to uncover 238 cross-ecosystem packages, later exploring the roles and functionalities of these packages. Results show that cross-ecosystem packages play a key role in their respective ecosystems, as evident by the statistical significance in both social (i.e., GitHub watchers, forks, star counts), and dependency among the dependencies (i.e., dependencies and dependents). As part of our qualitative analysis, we present a taxonomy (i.e., utilities, web toolkits, analytic platforms, databases, API service, Algorithm/Optimizer, Interpreter, Testing, Machine Learning and others) of cross-ecosystem packages with analysis to suggest that cross-ecosystem packages characterize different features of each respective programming language (e.g., PyPI with machine learning and Maven for utilities). The study uncovers these cross-ecosystem packages as a viable option for library replacement with implications for both developers and researchers.