paper

Crossing the Gap from Imperative to Functional Programming through Refactoring

  • Authors:
Read the paper

📜 Abstract

Java 8 introduces two functional features: lambda expressions and functional operations like map or filter that apply a lambda expression over the elements of a Collection. Refactoring existing code to use these new features enables explicit but unobtrusive parallelism and makes the code more succinct. However, refactoring is tedious: it requires changing many lines of code. It is also error-prone: the programmer must reason about the control-, data-flow, and side-effects. Fortunately, refactorings can be automated. We designed and implemented LambdaFicator, a tool which automates two refactorings. The first refactoring converts anonymous inner classes to lambda expressions. The second refactoring converts for loops that iterate over Collections to functional operations that use lambda expressions. Using 9 open-source projects, we have applied these two refactorings 1263 and 1709 times, respectively. The results show that LambdaFicator is useful: (i) it is widely applicable, (ii) it reduces the code bloat, (iii) it increases programmer productivity, and (iv) it is accurate.

✨ Summary

Summary

The paper introduces LambdaFicator, an automated refactoring tool for migrating imperative Java code toward the functional features introduced in Java 8. It supports two transformations: converting anonymous inner classes into lambda expressions, and converting enhanced for loops over collections into chains of stream operations such as map, filter, reduce, forEach, anyMatch, and noneMatch. The transformations are designed to preserve behavior while addressing Java-specific scoping rules, type inference, variable shadowing, control flow, side effects, and the distinction between lazy and eager stream operations. (dig.cs.illinois.edu)

The tool uses explicit safety preconditions and a variable-availability analysis to infer operation chains. Its evaluation covered nine open-source Java projects totaling approximately 1.1 million source lines. It converted 55% of examined anonymous inner classes and 46.02% of enhanced for loops. Anonymous-inner-class conversions reduced source code by 2,213 lines and achieved reported precision and recall of 100% in the sampled evaluation. For-loop conversions inferred 1,709 operation chains, with an average chain length of 2.72; the reported precision was 0.90 and recall was 0.92. Testing and manual inspection found no semantic regressions in the evaluated transformations. (dig.cs.illinois.edu)

Subsequent influence

The paper established an early practical approach to automated migration from imperative Java loops to Java 8 functional constructs. Its ideas were followed by work porting the enhanced-for-loop lambda refactoring from NetBeans to Eclipse, explicitly using the paper as a reference for bringing equivalent automated refactoring to another IDE. (researchgate.net) Later research on semantics-oriented Java stream refactoring, including Kayak, cited this work while pursuing safer and more broadly applicable transformations of external iteration into stream pipelines. (cprover.org) Subsequent studies and tools for automatic Java Stream API refactoring also cite the paper as prior work on analyzing constraints and preserving semantics during imperative-to-functional transformation. (researchgate.net)