Testing Unit and Systems
Before any developers new code can be integrated into the mainline production code base, it must be tested, vetted and peer reviewed.
Software Testing has been written about extensively, so I won’t spend time here discussing testing directly, as there are as many ways to test software as there are technologies.
Test Driven Development
Test Driven Development (TDD) a principal for which I agree is necessary, but I don’t always adhere to in practice. This is usually out of laziness, haste or carelessness (I know sad but I’m just being honest here). However, it is worth paying attention to…
Anyways, the basic rules are:
- Write your test before writing the code
- The test will fail initially
- It is the programmers job to write code such that the test will pass.
That is hard to argue with. However, taking the upfront time to build the tests before just hammering out the code, is much harder to do than say.
TDD is hard, not technically, but emotionally
The hard part about writing tests upfront is that it forces us to really define what we are building. You have to drill down a level or two.
When defining tests, ask the “Five whys?”
Once you have done the hard brain work and exposed the applications, true requirements implementing the test becomes striaght forward.
TODO Reference to TDD
Testing movements can be taken too far, for example spending a lot of time writing tests that are effectively worthless just for the sake of writing a test.
However, testing is critical and we must find a good balance. The following are some general points that a good set of tests provide:
- Tests must cover all features and bug code changes
- Tests must be 100% repeatable
- Tests should be automated 100%, if possible
- Tests should combine white and black box testing
- Tests should be easy to run selectively
- Tests should produce nice reports
Faithfully writing tests along with new code contributes over time to a increasingly stable code base.
Testing Builds Confidence
As software projects evolve over time, developers and users gain benefits beyond just stable software including:
Confidence to the point of taking the software for granted. They then begin to rely on the software without even realizing it.
Operational problems become rare because most issue are caught and fixed in a lab or a controlled system on the client site.
When problems do occur they are typically resolved very quickly. Stable products tend to be well documented, as well as having meaningful logs and great support tools.
Software Development Becomes Faster
The test environment creates a factory for producing well machined software.
High quality code can be developed and released quickly, with little risk of new failures.
A good test rig will allow the developer to easily run selective tests from the developers development process (REPL). The result is a much higher rate of software delivery without an increase in error rates.
Automated vs. Manual Tests
Tests are far more effective when they are clearly written down. Even better, written down AND automated.
It is inevitable that some tests, at least in certain industries, may have to be conducted manually. For example, making sure a robot hand does not break a glass bottle. Ideally however, you will try automating tests as close to 100% as possible.
Complete Test Runs ALL the Time
With automated tests it is easy to run a complete suite of tests for every software change and addition of new code.
Nothing will give a developer more confidence than comprehensive regression tests that are easy to run and easy to add new tests and modify existing ones.
Automated tests will most likely provide the ability to run a subset of test or even pick a single specific one to run.
The key is a developer is free to run and re-run tests to her hearts content until she is convinced the code is ready for everybody else.
Whitebox vs. Blackbox Testing
White Box testing means the test function has internal access to the the actual code being tested. In otherwords, the code can directly call any internal function or examine internal data structures.
Unit Testing is a standard form of Whitebox Test.
Other whitebox tests may include performance and stress tests.
Black Box Testing
Blackbox tests have no access or knoweledge of any internal functions or data structures, other than the applications “Application Programming Interface”.
System Tests are a standard form of Blackbox Testing
Other blackbox tests may also include performance and stress tests.
Acceptance Tests ~ Pass without Fail
Acceptance test is the term we will use for our final system tests. The tests that validate the delivery agreement between development and client.
Before any release passes to a customer, the entire suite of Acceptance Tests must pass without fail.
Acceptance tests are System Tests in that they will often involve co-ordination among two or more sub-systems or micro-services, as we like to call them now. For example, a database and inputs from a sensor are two sub-systems that are frequently found in embedded and IoT style software projects.