Concept: Developer Testing
Developers regression test their code on a continuous basis to ensure that it works as expected.
Relationships
Main Description

Developer testing is the act of regression testing source code by developers. This is sometimes called "unit regression testing" but many developer tests go beyond unit testing to address integration testing as well.

Testing Philosophies

Here are some important philosophies with regard to developer testing:

  1. The goal is to find defects. Successful tests find bugs, but correcting the bugs falls into other areas.
  2. Test early and often. The cost of change rises exponentially the longer it takes to find and then remove a defect. The implication is that you want to test as early as possible (the earliest you could possibly test is first, see Guideline: Test-first Design).
  3. Testing builds confidence. Many people fear making a change to their code because they are afraid that they will break it, but with a full test suite in place if you do break something you know you will detect it and then fix it.
  4. One test is worth a thousand opinions. You can say that your application works, but until you show the test results you might not be believed.
  5. Test to the risk. The riskier something is, the more it needs to be reviewed and tested. In other words you should invest significant effort testing in the algorithm for estimating radiation doses but nowhere near as much effort testing the "change font size" function of the same application.
  6. You can validate all artifacts. You can test all your artifacts, not just your source code, although the focus of this guidance is testing code.

Qualities of a Good Developer Test

These are the qualities of a good developer test:
  • It runs fast. It has short setup, run time, and clean-up.
  • It runs in isolation. You should be able to reorder your tests.
  • It is understandable. Good tests have consistent and informative names and use data that makes them easy to read and to understand.
  • It uses real data. For example, use copies of production data when appropriate, but remember that you'll typically have to create some specific "artificial" test data as well.
  • It is minimally cohesive. The test represents one step toward your overall goal. The test should address one and one only issue.

Approaches for Test Setup

To successfully run a test, the system must be in a known state. To do this you will need objects or components in memory, rows in the database, etc. that you will test against. The easiest approach is to hardcode the required data and the setup code within the test itself. The primary advantage is that all the information that you need about the test is in one place and that the test is potentially self-sufficient.

Another approach is to define an external data set which is loaded into memory or into the database at the beginning of the test run. There are several advantages to this approach:

  • It decouples the test data from the test.
  • More than one test can use the same data set.
  • It is easy to modify and/or multiply the test data.

There are some disadvantages to this approach:

  • Increased complexity for maintaining the external data
  • Potential coupling between test cases. When they share a common test data bed it becomes very easy to write tests that depend on other tests running first, thereby coupling them together.

Coding for Testability

Add code instrumentation for testing and debugging. Pay special attention to the implementation of the observation/control points, such as critical functions or objects, as these aspects might need special support that has to be implemented in the component under test.

Reviewing Tests

If a test will be long-lived, ask a person with less inside knowledge of the component to run it and check if there is enough support information. Review it with other people within the development team and other interested parties as needed.