Testing Pyramid

EGAD! How Do We Start Writing (Better) Tests?

Some have never automated tests and can’t check themselves before they wreck themselves. Others have 1000s of tests that are flaky, duplicative, and slow. Wa-do-we-do? Well, I gave a talk about this problem at PyOhio 2018. The language used for example code was Python, but the principles apply to any language. Please watch it below!

The Testing Pyramid

The “Testing Pyramid” is an industry-standard guideline for functional test case development. Love it or hate it, the Pyramid has endured since the mid-2000’s because it continues to be practical. So, what is it, and how can it help us write better tests?

Layers

The Testing Pyramid has three classic layers:

  • Unit tests are at the bottom. Unit tests directly interact with product code, meaning they are “white box.” Typically, they exercise functions, methods, and classes. Unit tests should be short, sweet, and focused on one thing/variation. They should not have any external dependencies – mocks/monkey-patching should be used instead.
  • Integration tests are in the middle. Integration tests cover the point where two different things meet. They should be “black box” in that they interact with live instances of the product under test, not code. Service call tests (REST, SOAP, etc.) are examples of integration tests.
  • End-to-end tests are at the top. End-to-end tests cover a path through a system. They could arguably be defined as a multi-step integration test, and they should also be “black box.” Typically, they interact with the product like a real user. Web UI tests are examples of integration tests because they need the full stack beneath them.

All layers are functional tests because they verify that the product works correctly.

Proportions

The Testing Pyramid is triangular for a reason: there should be more tests at the bottom and fewer tests at the top. Why?

  1. Distance from code. Ideally, tests should catch bugs as close to the root cause as possible. Unit tests are the first line of defense. Simple issues like formatting errors, calculation blunders, and null pointers are easy to identify with unit tests but much harder to identify with integration and end-to-end tests.
  2. Execution time. Unit tests are very quick, but end-to-end tests are very slow. Consider the Rule of 1’s for Web apps: a unit test takes ~1 millisecond, a service test takes ~1 second, and a Web UI test takes ~1 minute. If test suites have hundreds to thousands of tests at the upper layers of the Testing Pyramid, then they could take hours to run. An hours-long turnaround time is unacceptable for continuous integration.
  3. Development cost. Tests near the top of the Testing Pyramid are more challenging to write than ones near the bottom because they cover more stuff. They’re longer. They need more tools and packages (like Selenium WebDriver). They have more dependencies.
  4. Reliability. Black box tests are susceptible to race conditions and environmental failures, making them inherently more fragile. Recovery mechanisms take extra engineering.

The total cost of ownership increases when climbing the Testing Pyramid. When deciding the level at which to automate a test (and if to automate it at all), taking a risk-based strategy to push tests down the Pyramid is better than writing all tests at the top. Each proportionate layer mitigates risk at its optimal return-on-investment.

Practice

The Testing Pyramid should be a guideline, not a hard rule. Don’t require hard proportions for test counts at each layer. Why not? Arbitrary metrics cause bad practices: a team might skip valuable end-to-end tests or write needless unit tests just to hit numbers. W. Edwards Deming would shudder!

Instead, use loose proportions to foster better retrospectives. Are we covering too many input combos through the Web UI when they could be checked via service tests? Are there unit test coverage gaps? Do we have a pyramid, a diamond, a funnel, a cupcake, or some other wonky shape? Each layer’s test count should be roughly an order of magnitude smaller than the layer beneath it. Large Web apps often have 10K unit tests, 1K service tests, and a few hundred Web UI tests.

Resources

Check out these other great articles on the Testing Pyramid:

Quality Metrics 101: Test Quality

New to the series? Start from the beginning!

Test quality metrics make sure that testing efforts are worthwhile. Though “testing” and “quality” may be synonymous as organizational titles, testing is only one method of enforcing quality. In software, it just happens to be the most effective one. Testing is expensive, though, because it slows down time-to-market. Some people even devalue testing work because it doesn’t add new features to a product. Below are aspects of test quality to consider measuring to prove and even increase the value of testing efforts.

roofing

Coverage

Quality Aspect How much functionality is covered by tests?
Desired State High – More coverage means less risk. Note that 100% complete coverage is impossible.
Metrics Coverage may be measured for both manual and automated tests. However, automated test coverage is usually more important because automated tests are meant to be defensive without gaps.

Code Coverage – Code coverage tools check what paths of code are actually exercised by automated tests. While they cannot tell if tests are good or bad, they are great for exposing gaps in coverage. Unit test code coverage is easy because most frameworks have plugins, but above-unit code coverage requires instrumented builds. Look for tools that track more than just lines of code. Target 90%+ coverage. Add new tests to cover any major gaps.

Feature Coverage – Feature coverage is a manual way to score features on test coverage based on planning and review. For this metric to be successful, a team must consistently specify features well; otherwise, this metric will give useless data. Gherkin scenarios a great way to do this – for example, each scenario can be marked as untested, manual, or automated. Feature coverage is unscientific, but it can give a better picture of functionalities actually covered (instead of just the raw lines of code covered).

Automation DebtTechnical debt increases when tests are not automated and thus lack coverage. Teams are often unable to automate all tests originally planned, and test automation is frequently jettisoned from the Definition of Done. Or, a project may not start automating tests until a large chunk of the project is already complete. The best way to track automation debt is to create a backlog for incomplete automation work. Backlog tasks can be sized, prioritized, and planned according to whatever development process is used (Scrum, Kanban, etc.). Appropriate process metrics can then be used to understand the magnitude of the work and, thus, the lack of automated test coverage.

Warning: Test case count, test length, and test code line count are terrible metrics for coverage because they encourage largeness rather than uniqueness. The goal of testing is to have the greatest coverage with the lowest risk for the least work. Anybody can blindly write tests or variations that add no meaningful value.

gears-image1

Reliability

Quality Aspect Do automated tests consistently reach completion? And how trustworthy are the results?
Desired State High – Reliability means less time for failure triage or (horrors) reruns.
Metrics Failure Reasons – Track the failure reason for each test case run. Ideally, tests should fail only when they discover product bugs. However, tests may also fail when:

  • an acceptable product change caused an automation error because tests were not updated, indicating poor communication or careless updates
  • an environmental change or interruption caused an automation error, indicating deployment or sysadmin problems
  • the automation code itself has a bug

Remember, “successful” test runs either pass with appropriate coverage or fail due to product bugs. “Unsuccessful” test runs fail or crash for reasons other than product bugs. Aim to minimize unsuccessful test runs. Never hack a test just to get it passing – always work to fix the problems behind test failures.

is-the-speedometer-reliable-in-telling-me-my-vehicles-real-time-speed

Speed

Quality Aspect How much time do test runs take?
Desired State Fast – Tests should complete in the shortest time possible.
Metrics Test Case Execution Time – Test case execution times indicate the efficiency of the automation code. Track the start-to-end execution time for every individual test case run. Then, analyze the data using common sense. For example, outliers may be inefficient tests that need tuning or should be removed altogether. It may be wise to separate test runs by result type or coverage area. Historical data can also be used as a baseline to determine performance impacts when making cross-cutting automation changes.

Test Suite Execution Time – Test suites are sets of test cases, but their execution times are not merely the sum of their tests’ times. A test suite run may include environmental setup, deployment, parallel execution, reporting, and other things. The purpose of tracking test suite execution time is to determine the start-to-end time of the suite in total, because that indicates the speed of feedback and, in CI, delivery. Tracking test suite execution time will also reveal the effect of adding more test cases to the suite, which then factors into the risk-based decisions of including or excluding tests.

Test Pyramid Balance – The Test Pyramid separates tests between unit (bottom), integration (middle), and end-to-end (top) layers. Ideally, there should be more tests at the bottom than at the top. Why? Higher-level tests are more expensive – they take more time to develop, they are more time consuming to triage, and they have slower execution times. Consider the “Rule of 1’s”: a unit test takes ~1ms, an integration test takes ~1s, and an end-to-end test takes ~1m. When scaled to thousands of tests with continuous integration, end-to-end tests simply take too much time. Tracking the proportion of tests at each layer will give a rough picture of the balance. There’s no perfect ratio between layers, but make sure that the tests form a pyramid and not a cupcake, hourglass, or ice cream cone. Rebalance test efforts as appropriate.

piggybankmoney

Return on Investment

Quality Aspect Do the tests add greater value than their cost?
Desired State High – Tests need to be worth the effort. Don’t test for the sake of testing!
Metrics Measuring return on investment in terms of hard dollars is objectively impossible. The true cost of bugs can never be fully known: if a bug is caught early, the potential cost to fix it later can merely be estimated. The intangible value of protecting brand reputation may be more important than the tangible value of money saved by finding specific bugs. Better quality practices might prevent developers from causing bugs that would have otherwise happened – and there’s no good way to measure that.

Instead, return on investment is better measured by a collection of metrics that validate both code line protection and defect discovery. Use a weighted scorecard to get a more holistic view of ROI. Scorecards can be used with estimates for planning tests, as well as plugged in with actual values to measure the degree of success. Note that some aspects of ROI may be too difficult to measure accurately – in those cases, a LOW-MID-HIGH grading scale may be best. Others may seem like micromanagement.

  • Priority – Assign each test a priority for its coverage importance. Core functionalities should have the highest priority, while fringe functionalities should have the lowest priority. Focus on high-priority tests. Another way to look at importance is risk, or the chances that bugs will escape if explicit testing for a feature is not done.
  • Test Execution Frequency – Track how many times tests are actually run. Higher frequency is better. Tests that are rarely run should either be included in more regular runs or removed/archived. This could easily be tracked by a test management tool or database.
  • Coverage Uniqueness – Duplicate test coverage wastes resources. Unfortunately, this one is difficult to measure. Tools for code coverage or static analysis might help. Manual review, however, is typically a better approach.
  • Development Cost and Maintenance Cost – Track how much effort it takes to make and keep tests, including man-hours and resources. Lower costs are better, of course. Planning tools may help with this.
  • Bug Discovery – Track bugs discovered in terms of severity and when and how they were caught. Ideally, the number of bugs caught by customers after a release (meaning, not caught by tests during development) should be minimal, and their severity should be low. Bug tracking tools should easily provide this data. Be warned, though, that the raw bug count is a poor metric. Consider this question: Is a high bug count good or bad? Trick question – during a release, it indicates good test quality but poor product quality; after a release, it indicates all-around poor quality. What matters is that a minimal number of bugs happen at all, and that most of those bugs are caught and fixed before a release. Plus, keep in mind that bugs happen by accident. Finally, focusing exclusively on bug count to determine test value ignores the positive side of testing – that passing tests give confidence that features work correctly.

To Infinity and Beyond: A Guide to Parallel Testing

Are your automated tests running in parallel? If not, then they probably should be. Together with continuous integration, parallel testing the best way to fail fast during software development and ultimately enforce higher software quality. Switching tests from serial to parallel execution, however, is not a simple task. Tests themselves must be designed to run concurrently without colliding, and extra tools and systems are needed to handle the extra stress. This article is a high-level guide to good parallel testing practices.

What is Parallel Testing?

Parallel testing means running multiple automated tests simultaneously to shorten the overall start-to-end runtime of a test suite. For example, if 10 tests take a total of 10 minutes to run, then 2 parallel processes could execute 5 tests each and cut the total runtime down to 5 minutes. Even better, 10 processes could execute 1 test each to shrink runtime to 1 minute. Parallel testing is usually managed by either a test framework or a continuous integration tool. It also requires more compute resources than serial testing.

Why Go Parallel?

Running automated tests in parallel does require more effort (and potentially cost) than running tests serially. So, why go through the trouble?

The answer is simple: time. It is well documented that software bugs cost more when they are discovered later. That’s why current development practices like Agile and BDD strive to avoid problems from the start through small iterations and healthy collaboration (“shift left“), while CI/CD defensively catches regressions as soon as they happen (“fail fast“). Reducing the time to discover a problem after it has been introduced means higher quality and higher productivity.

Ideally, a developer should be told if a code change is good or bad immediately after committing it. The change should automatically trigger a new build that runs all tests. Unfortunately, tests are not instantaneous – they could take minutes, hours, or even days to complete. A test automation strategy based on the Testing Pyramid will certainly shorten start-to-end execution time but likely still require parallelization. Consider the layers of the Testing Pyramid and their tests’ average runtimes, the Testing Pyramid Rule of 1’s:

The Testing Pyramid with Times

Each layer is listed above with the rough runtime of a typical test. Though actual runtimes will vary, the Rule of 1’s focuses on orders of magnitude. Unit tests typically run in milliseconds because they often exercise product code in memory. Integration tests exercise live products but are limited in scope and often cover low-level areas (like REST service calls). End-to-end tests, however, cover full paths through a live system, which requires extra setup and waiting (like Selenium WebDriver interaction).

Now, consider how many tests from each layer could be run within given time limits, if the tests are run serially:

Test Layer 1 Minute
Near-Instant
10 Minutes
Coffee Break
1 Hour
There Goes Today
Unit 60,000 600,000 3,600,000
Integration 60 600 3,600
End-to-End 1 10 60

Unit test numbers look pretty good, though keep in mind 1 millisecond is often the best-case runtime for a unit test. Integration and end-to-end runtimes, however, pose a more pressing problem. It is not uncommon for a project to have thousands of above-unit tests, yet not even a hundred end-to-end tests could complete within an hour, nor could a thousand integration tests complete within 10 minutes. Now, consider two more facts: (1) tests often run as different phases in a CI pipeline, to total runtimes are stacked, and (2) multiple commits would trigger multiple builds, which could cause a serious backup. Serial test execution would starve engineering feedback in any continuous integration system of scale. A team would need to drastically shrink test coverage or give up on being truly “continuous” in favor of running tests daily or weekly. Neither alternative is acceptable these days. CI needs parallel testing to be truly continuous.

The Danger of Collisions

The biggest danger for parallel testing is collision – when tests interfere with each other, causing invalid test failures. Collisions may happen in the product under test if product state is manipulated by more than one test at a time, or they may happen in the automation code itself if the code is not thread-safe. Collisions are also inherently intermittent, which makes them all the more difficult to diagnose. As a design principle, automated tests must avoid collisions for correct parallel execution.

Making tests run in parallel is not as simple as flipping a switch or adding a new config file. Automated tests must be specifically designed to run in parallel. A team may need to significantly redevelop their automation code to make parallel execution work right.

train-collision-in-iran-at-least-5-killed-aa48cb33e0af63ac1632c64194621469

A train collision in Iran in November 2016. Don’t let this happen to your tests!

Handling Product-Level Collisions

Product-level collisions essentially reduce to how environments are set up and handled.

Separate Environments

The most basic way to avoid product-level collisions would be to run each test thread or process against its own instance of the product in an exclusive environment. (In the most extreme case, every single test could have its own product instance.) No collisions would happen in the product because each product instance would be touched by only one test instance at a time. Separate environments are possible to implement using various configuration and deployment tools. Docker containers are quick and easy to spin up. VMs with Vagrant, Puppet, Chef, and/or Ansible can also get it done.

However, it may not always be sensible to make separate environments for each test thread/process:

  • Creating a new environment is inefficient – it takes extra time to set up that may cancel out any time saved from parallel execution.
  • Many projects simply don’t have the money or the compute resources to handle a massive scale-out.
  • Some tests may not cause collisions and therefore may not need total isolation.
  • Some product environments are extremely large and complicated and would not be practical to replicate for each test individually.

Shared Environments

Environments with a shared product instance are quite common. One could be a common environment that everyone on a team shares, or one could be freshly created during a CI run and accessed by multiple test threads/processes. Either way, product-level collisions are possible, and tests must be designed to avoid clashing product states. Any test covering a persistent state is vulnerable; usually, this is the vast majority of tests. Consider web app testing as an example. Tests to load a page and do some basic interactions can probably run in parallel without extra protection, but tests that use a login to enter data or change settings could certainly collide. In this case, collisions could be avoided by using different logins for each simultaneous test instance – by using either a pool of logins, a unique login per test case, or a unique login per thread/process. Each product is different and will require different strategies for avoiding collisions.

the_earth_seen_from_apollo_17

We all share certain environments. Take care of them when you do. (Photo: The Blue Marble, taken by the Apollo 17 crew on Dec 7, 1972)

Handling Automation-Level Collisions

Automation-level collisions can happen when automation code is not thread-safe, which could mean more than simply locks and semaphores.

#1: Test Independence

Test cases must be completely independent of each other. One test must not require another test to run before it for the sake of setup. A test case should be able to run by itself without any others. A test suite should be able to run successfully in random order.

#2: Proper Variable Scope

If parallel tests will be run in the same memory address space, then it is imperative to properly scope all variables. Global or static mutable variables (e.g., “non-constants”) must not be allowed because they could be changed unexpectedly. The best pattern for handling scope is dependency injection. Thread-safe singletons would be a second choice. (Typically, global or static variables are used to subvert design patterns, so they may reveal further necessary automation rework when discovered.)

#3: External Resources

Automation may sometimes interact with external resources, such as test config files or test result databases/services. Make sure no external interactions collide. For example, make sure test run updates don’t overwrite each other.

#4: Logging

Logs are very difficult to trace when multiple tests are simultaneously printed to the same file. The best practice is to generate separate log files for each test case, thread, or process to make them readable.

#5: Result Aggregation

A test suite is a unified collection of tests, no matter how many threads/processes are used to run its tests in parallel. Make sure test results are aggregated together into one report. Some frameworks will do this automatically, while others will require custom post-processing.

#6: Test Filtering

One strategy to avoid collisions may be to run non-colliding partitions (subsets) of tests in parallel. Test tagging and filtering would make this possible. For example, tests that require a special login could be tagged as such and run together on one thread.

Test Scalability

The previous section on collisions discussed how to handle product environments. It is also important to consider how to handle the test automation environment. These are two different things: the product environment contains the live product under test, while the test environment contains the automation software and resources that run tests against the product. The test environment is where the parallel tests will be executed, and, as such, it must be scalable to handle the parallelization. A common example of a test environment could be a Jenkins master with a few agents for running build pipelines. There are two primary ways to scale the test environment: scale-up and scale-out.

Parallel Scale-Up

Scale-up is when one machine is configured to handle more tests in parallel. For example, scale-up would be when a machine switches from one (serial) thread to two, three, or even more in parallel. Many popular test runners support this type of scale-up by spawning and joining threads in a common memory address space or by forking processes. (For example, the SpecFlow+ Runner lets you choose.)

Scale-up is a simple way to squeeze as much utility out of an existing machine as possible. If tests are designed to handle collisions, and the test runner has out-of-the-box support, then it’s usually pretty easy to add more test threads/processes. However, parallel test scale-up is inherently limited by the machine’s capacity. Each additional test process succumbs to the law of diminishing returns as more memory and processor cycles are used. Eventually, adding more threads will actually slow down test execution because the processor(s) will waste time constantly switching between tests. (Anecdotally, I found the optimal test-thread-to-processor ratio to be 2-to-1 for running C#/SpecFlow/Selenium-WebDriver tests on Amazon EC2 M4 instances.) A machine itself could be upgraded with more threads and processors, but nevertheless, there are limits to a single machine’s maximum capacity. Weird problems like TCP/IP port exhaustion may also arise.

Scale Up

Scale-up adds more threads to one machine.

Parallel Scale-Out

Scale-out is when multiple machines are configured to run tests in parallel. Whereas scale-up had one machine running multiple tests, scale-out has multiple machines each running tests. Scale-out can be achieved in a number of ways. A few examples are:

  • One master test execution machine launches multiple Web UI tests that each use a remote Selenium WebDriver with a service like Selenium Grid, Sauce Labs, or BrowserStack.
  • A Jenkins pipeline launches tests across ten agents in parallel, in which each agent executes a tenth of the tests independently.

Scale-out is a better long-term solution than scale-up because scale-out can handle an unlimited number of machines for parallel testing. The limiting factor with scale-out is not the maximum capacity of the hardware but rather the cost of running more machines. However, scale-out is much harder to implement than scale-up. It requires tests to be evenly divided with some sort of balancer and filter. It also requires some sort of test result aggregation for joint reporting – people won’t want to piece together a bunch of separate reports to get an overall snapshot of quality. Plus, the test environment is more complicated to build and maintain (though tools like CloudBees Jenkins Enterprise or Amazon EC2 can make it easier.)

Scale Out

Scale-out distributes tests across multiple machines.

Upwards and Outwards

Of course, scale-up and scale-out are not mutually exclusive. Scaled-out nodes could individually be scaled-up. Consider a test environment with 10 powerful VMs that could each handle 10 tests in parallel – that means 100 tests could run simultaneously. Using the Rule of 1’s, it would take only about a minute to run 100 Web UI tests, which serially would have taken over an hour and a half! Use both strategies to shorten start-to-end runtime as much as possible.

Conclusion

Parallel testing is a worthwhile endeavor. When done properly, it will not only reduce development time but also improve the development experience. For readers who want to start doing parallel testing, I recommend researching the tools and frameworks you want to use. Many popular test frameworks support parallel execution, and even if the one you choose doesn’t, you can always invoke tests in parallel from the command line. Do well!

The Airing of Grievances: Test Automation Process

Test automation is a big deal for me. It is my chosen specialty within the broad field of software. When I see things done wrong, or when people just don’t get what it’s about, it really grinds my gears. I got a lot of problems with bad test automation processes, and now you’re gonna hear about it!

Saying “They’re Just Test Scripts”

Test automation is not just a bunch of test scripts: it is a full technology stack that requires design, integration, and expertise. Test automation development is a discipline. Saying it is just a bunch of test scripts is derogatory and demeaning. It devalues the effort test automation requires, which can lead to poor work item sizings and an “us vs. them” attitude between developers and QA.

Not Applying the Same Software Development Best Practices

Test automation is software development, and all the same best practices should thus apply. Write clean, well-designed code. Use version control with code reviews. Add comments and doc. Don’t get lazy because “they’re just test scripts” – wrong attitude!

Lip Service

Don’t say automation is important but then never dedicate time or resources to work on it. Don’t leave automation as a task to complete only if there’s time after manual testing is done. Make automation a priority, or else it will never get done! I once worked on an Agile team where automation framework stories were never included into the sprint because there weren’t “enough points to go around.” So, even though this company hired me explicitly to do test automation, I always got shunted into a manual testing scramble every sprint.

Confusing Test Automation with Deployment Automation

Test automation is the automation of test scenarios (for either functional or performance tests). Deployment automation is the automation of product build distribution and installation in a software environment. They are two different concerns. Cucumber is not Ansible.

Forcing 100% Automation

Some people think that automation will totally eliminate the need for any manual testing. That’s simply not true. Automation and manual testing are complementary. Automation should handle deterministic scenarios with a worthwhile return-on-investment to automate, while manual testing should focus on exploratory testing, user experience (UX), and tests that are too complicated to automate properly. Forcing 100% automation will make teams focus on metrics instead of quality and effectiveness.

Downsizing or Eliminating QA

Test automation doesn’t reduce or eliminate the need for testers. On the contrary, test automation requires even more advanced skills than old-school manual testing. There is still a need for testing roles, whether as a dedicated position or as shared collectively by a bunch of developers. The work done by that testing role just becomes more technical.

Saying Product Code and Test Code Must Use the Same Language

For unit tests, this is true, but for above-unit tests, it is simply false. Any general purpose programming language could be used to automate black-box tests. For example, Python tests could run against an Angular web app. A team may choose to use the same language for product and test code for simplicity, but it is not mandatory.

Not Classifying Test Types

Not all tests are the same. You can play buzzword bingo with all the different test type names: unit, integration, end-to-end, functional, performance, system, contract, exploratory, stress, limits, longevity, test-to-break, etc. Different tests need different tools or frameworks. Tests should also be written at the appropriate Testing Pyramid level.

Assuming All Tests Are Equal

Again, not all tests are the same, even within the same test type. Tests vary in development time, runtime, and maintenance time. It’s not accurate to compare individuals or teams merely on test numbers.

Not Prioritizing Tests to Automate

There’s never enough time to automate everything. Pick the most important ones to automate first – typically the core, highest-priority features. Don’t dilly-dally on unimportant tests.

Not Running Tests Regularly

Automated tests need to run at least once daily, if not continuously in Continuous Integration. Otherwise, the return-on-investment is just too low to justify the automation work. I once worked on a QA team that would run an automated test only once or twice during a 2-year release! How wasteful.

HP Quality Center / ALM

This tool f*&@$#! sucks. Don’t use it for automated tests. Pocket the money and just develop a good codebase with decent doc in the code, and rely upon other dashboards (like Jenkins or Kibana) for test reporting.

BDD 101: Unit, Integration, and End-to-End Tests

There are many types of software tests. BDD practices can be incorporated into all aspects of testing, but BDD frameworks are not meant to handle all test types. Behavior scenarios are inherently functional tests – they verify that the product under test works correctly. While instrumentation for performance metrics could be added, BDD frameworks are not intended for performance testing. This post focuses on how BDD automation works into the Testing Pyramid. Please read BDD 101: Manual Testing for manual test considerations. (Check the Automation Panda BDD page for the full table of contents.)

The Testing Pyramid

The Testing Pyramid is a functional test development approach that divides tests into three layers: unit, integration, and end-to-end.

  • Unit tests are white-box tests that verify individual “units” of code, such as functions, methods, and classes. They should be written in the same language as the product under test, and they should be stored in the same repository. They often run as part of the build to indicate immediate success or failure.
  • Integration tests are black-box tests that verify integration points between system components work correctly. The product under test should be active and deployed to a test environment. Service tests are often integration-level tests.
  • End-to-end tests are black-box tests that test execution paths through a system. They could be seen as multi-step integration tests. Web UI tests are often end-to-end-level tests.

Below is a visual representation of the Testing Pyramid:

The Testing Pyramid

The Testing Pyramid

From bottom to top, the tests increase in complexity: unit tests are the simplest and run very fast, while end-to-end require lots of setup, logic, and execution time. Ideally, there should be more tests at the bottom and fewer tests at the top. Test coverage is easier to implement and isolate at lower levels, so fewer high-investment, more-fragile tests need to be written at the top. Pushing tests down the pyramid can also mean wider coverage with less execution time. Different layers of testing mitigate risk at their optimal returns-on-investment.

Behavior-Driven Unit Testing

BDD test frameworks are not meant for writing unit tests. Unit tests are meant to be low-level, program-y tests for individual functions and methods. Writing Gherkin for unit tests is doable, but it is overkill. It is much better to use established unit test frameworks like JUnit, NUnit, and pytest.

Nevertheless, behavior-driven practices still apply to unit tests. Each unit test should focus on one main thing: a single call, an individual variation, a specific input combo; a behavior. Furthermore, in the software process, feature-level behavior specs draw a clear dividing line between unit and above-unit tests. The developer of a feature is often responsible for its unit tests, while a separate engineer is responsible for integration and end-to-end tests for accountability. Behavior specs carry a gentleman’s agreement that unit tests will be completed separately.

Integration and End-to-End Testing

BDD test frameworks shine at the integration and end-to-end testing levels. Behavior specs expressively and concisely capture test case intent. Steps can be written at either integration or end-to-end levels. Service tests can be written as behavior specs like in Karate. End-to-end tests are essentially multi-step integrations tests. Note how a seemingly basic web interaction is truly a large end-to-end test:

Given a user is logged into the social media site
When the user writes a new post
Then the user's home feed displays the new post
And the all friends' home feeds display the new post

Making a simple social media post involves web UI interaction, backend service calls, and database updates all in real time. That’s a full pathway through the system. The automated step definitions may choose to cover these layers implicitly or explicitly, but they are nevertheless covered.

Lengthy End-to-End Tests

Terms often mean different things to different people. When many people say “end-to-end tests,” what they really mean are lengthy procedure-driven tests: tests that cover multiple behaviors in sequence. That makes BDD purists shudder because it goes against the cardinal rule of BDD: one scenario, one behavior. BDD frameworks can certainly handle lengthy end-to-end tests, but careful considerations should be taken for if and how it should be done.

There are five main ways to handle lengthy end-to-end scenarios in BDD:

  1. Don’t bother. If BDD is done right, then every individual behavior would already be comprehensively covered by scenarios. Each scenario should cover all equivalence classes of inputs and outputs. Thus, lengthy end-to-end scenarios would primarily be duplicate test coverage. Rather than waste the development effort, skip lengthy end-to-end scenario automation as a small test risk, and compensate with manual and exploratory testing.
  2. Combine existing scenarios into new ones. Each When-Then pair represents an individual behavior. Steps from existing scenarios could be smashed together with very little refactoring. This violates good Gherkin rules and could result in very lengthy scenarios, but it would be the most pragmatic way to reuse steps for large end-to-end scenarios. Most BDD frameworks don’t enforce step type order, and if they do, steps could be re-typed to work. (This approach is the most pragmatic but least pure.)
  3. Embed assertions in Given and When steps. This strategy avoids duplicate When-Then pairs and ensures validations are still performed. Each step along the way is validated for correctness with explicit Gherkin text. However, it may require a number of new steps.
  4. Treat the sequence of behaviors as a unique, separate behavior. This is the best way to think about lengthy end-to-end scenarios because it reinforces behavior-driven thinking. A lengthy scenario adds value only if it can be justified as a uniquely separate behavior. The scenario should then be written to highlight this uniqueness. Otherwise, it’s not a scenario worth having. These scenarios will often be very declarative and high-level.
  5. Ditch the BDD framework and write them purely in the automation programming. Gherkin is meant for collaboration about behaviors, while lengthy end-to-end tests are meant exclusively for intense QA work. Biz roles will write behavior specs but will never write end-to-end tests. Forcing behavior specification on lengthy end-to-end scenarios can inhibit their development. A better practice could be coexistence: acceptance tests could be written with Gherkin, while lengthy end-to-end tests could be written in raw programming. Automation for both test sets could still nevertheless share the same automation code base – they could share the same support modules and even step definition methods.

Pick the approach that best meets the team’s needs.