Tuesday, August 31, 2010

No Time for Testing!

Not too long ago I met someone who was developing a new software product for release later this year. I casually asked about their test plans and test results to date. The response I got was underwhelming, and, unfortunately, a response that I’ve heard from other development teams. Basically, my paraphrase of their response was: “We’re too busy developing to test right now.”

Immediately, red flags popped into my head. In my opinion, software testing is an integral part of the software development process and must not be left until the end of the development period. The purpose of testing is to not only find defects in the software product but also to ensure that the product requirements are understood well and that the product is designed to be testable. And, what better way to ensure that product requirements are understood well and testable than by developing tests in the early part of a development cycle or iteration (if developing using an agile methodology).

This means that test planning occurs at the start of a project or iteration. I recommend that developers follow a test-driven development (TDD) methodology that mandates that an automated test is written for each feature/function before any code is written. When the test passes, the code is complete. TDD is especially helpful for development organizations using an agile methodology. For each development iteration, TDD requires the developer to understand the requirements of the feature/function they are implementing and ensures that the feature/function is designed and coded to be testable. TDD also ensures that the features/functions of the iteration work at the end of an iteration.

However, test-driven development does not represent the only testing that should occur and does not mean that a product has no defects. A product development team must consider all the different types of testing that can and should be done, such as unit testing (using TDD), functional testing, system testing, integration testing, and regression testing. Full test design and implementation should be done in parallel with software design and implementation. This ensures that product requirements are written clearly and are testable. Full testing throughout the product development cycle also ensures that defects are identified as early in the development cycle as possible.

Eileen Boerger
Agilis Solutions


No comments:

Post a Comment