Advanced Computing in the Age of AI | Thursday, May 23, 2024

AI Tool Helps Coders Accelerate Testing 


There is relief for Java developers bogged down with necessary but tedious testing of individual units of source code: an automated platform that automatically generates regression unit tests for Java applications.

Diffblue, an Oxford University spin out, recently released a community edition of its enterprise Cover platform used by early customers such as Amazon Web Services (NASDAQ: AMZN) and Goldman Sachs (NYSE: GS).

The AI-based system applies software verification and synthesis technologies to reflect the current behavior of an application. The unsupervised learning framework can then be used to find regressions in new code commits, producing a unit test of individual source code as much as 100 times faster than humans.

The testing tool addresses soaring enterprise requirements for deploying new code, often thousands of times per day. At the same time, market trackers predict that enterprise demand for application developers will grow five-fold through 2021.

To help solve that problem, startups like Diffblue are attempting to free Java developers from tedious source code testing so they can focus on pushing bug-free code to production faster. Unlike other tooling, Diffblue claims its Cover tool automatically maintains its test suite as code evolves, updating only those unit tests that must be re-written after a change.

While lacking the bell and whistles of its paid version, the community edition of Cover includes a plug-in to the IntellJ development environment and automated unit tests. Early users have noted restrictions in the community edition, most notably that is works with Java 11 but not Java 14. Another is that coders must use a later version of Spring Boot, the open source Java-based framework for application development.

Diffblue settled on the ubiquitous Java programming language for the first version of its unit tester. Support for C, JavaScript and Python are expected soon. (Usage surveys that have consistently ranked Python at or near the top of data scientists’ programming language preferences. One reason is the growing number of tools and libraries used to explore big data sets.)

An “individual” option is also planned, priced well below the enterprise version.

Founded in 2016, Diffblue is backed by Goldman Sachs and Oxford Sciences Innovation, Oxford University’s early-stage venture capital fund.

Diffblue reported that about two weeks after releasing its community edition, developers used it to write about 21,000 Java unit tests. That works out to an estimated time savings of 580 hours, the company estimates.

Download the free community edition of the Cover unit test generator here.

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).