Why I Switched to Table Driven Testing approach in Go

Table driven tests, also known as parameterized tests, have became very popular over the past few years, due to their ability to eliminate repetition. Table driven tests make it quite a bit easier to re-use the same values for different sets of tests by just moving the table outside of the scope of the test function. Different tests may benefit from the same input, and each test may have completely different configration, concurrency etc... The table structure Well, think of it as a systematic way of testing your code by providing a set of inputs and expected outputs in a structured table format. Instead of writing individual test cases for each input-output pair, we organize them neatly in a table, making it easier to manage and maintain our tests. Each row in the table represents a separate test case, and the columns represent different aspects or parameters of the test, such as inputs, expected outputs, and any other relevant conditions. Here's a simplified example of what a table-driven test might look like for a function that calculates the square of a number: package main import ( "testing" ) // We first have the function we may want to test // Consider the following function that calculates the addition // of two numbers func Add(a, b int) int { return a + b } /* We want to write tests that ensure our add function returns the correct result, no matter what numbers we provide. For the sake of keeping this blog to the point, I'm just going to include two test cases */ func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ "10+3": { inputA: 10, inputA: 3, expected: 13, }, "2+6": { inputA: 2, inputA: 6, expected: 8, }, } for _, tt := range tests { t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } Here, once we have declared the test cases via the tests map, the test function iterates over all the map entries, i.e. the test cases. The Add function is called in each iteration using the input arguments of the current test case. If the actual result doesn’t match the result expected by the test case, the test fails with an error message. Why Use Table-Driven Tests? Now, you might be wondering, why bother with table-driven tests? Here are a few reasons: Clarity: By organizing test cases in a table, it's easier to see all the inputs and expected outputs at a glance. This makes it simpler to understand what the test is doing. Scalability: As your codebase grows, you'll likely have more test cases to manage. Table-driven tests make it easier to add new test cases without cluttering your test code. Maintainability: Since test cases are organized in a structured format, it's easier to update or modify them when needed. What constitutes a table driven test? Typically every Table-driven tests case will have a description, input, and expected output values. Let's explore some techniques: Maps and slices Usually, test cases contain a name field to uniquely identify and describe the test case, and these are stored in slice. tests := []struct { name string inputA int inputB int expected int } Although, slices are better approach when the test cases are simple, It's is preferred to use maps for storing when a lot of complex test cases start occupying too much space. Other benefit of using maps instead of the slice is that, it is an undefined iteration order. This means that test cases might be executed in a different order each time, which could expose issues in your test setup. tests := map[string]struct { inputA int inputB int expected int } It is also an issue because, if your tests expect certain conditions to be met in a specific order, using maps for table-driven tests might not be the best choice because the iteration order is not guaranteed. It's essential to consider this when designing your test setup to ensure reliable and consistent testing results. Solution to this that the order of test cases are maintained explicitly. This can be done by using an ordered slice of structs instead of a map. func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ {"Test case 1", 1, 2, 3}, {"Test case 2", -1, 1, 0}, {"Test case 3", 0, 0, 0}, } for _, tt := range tests { tt := tt // capture range variable t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } Here, we use a slice of structs to maintain the order of test cases. Inside the loop, we create a local copy of tt using := to ensure that each iteration has its own copy of the test case. This prevents any unintended sharing of state between iterations. By using an ordered slice instead of a map, we ensure that the test cases are executed in the defined order, addressing the concern of non-deterministic behavior introduced by map iteration. This approach provides more reliability and consistency in test execution. Error handling & Logging When a comparison between an actual value and an expected value reveals a mismatch, there are two ways to handle it. Either we log an error message using Errorf or we terminate the test using Fatalf. In the first scenario, we log the error but allow the test to continue with subsequent checks. The latter option stops the entire test immediately upon encountering a failure. It's important to use these options thoughtfully. Generally, it's recommended to log the failure and continue with the test using Errorf. Only if proceeding with the test wouldn't make sense due to failed preconditions, should Fatalf be used to halt the test immediately. For instance, if the lengths of two slices, actual and test.expected, don't match, we immediately fail the test with Fatalf: if len(actual) != len(tt.expected) { t.Fatalf("lengths don't match: expected %v, got %v", len(tt.expected), len(actual)) } Assuming the lengths are the same, we compare the values within the slices one by one. If a mismatch occurs, we log an error using Errorf but continue with the loop: for i, value := range tt.expected { if actual[i] != value { t.Errorf("values at index %d don't match: expected %v, got %v", i, value, actual[i]) } } Structured test cases To manage complex test cases with many parameters or dependencies, we can use helper structs. These structs store input parameters and expected results separately: func TestAdd(t *testing.T) { type Given struct { inputA int inputB int } type Expected struct { result int } tests := map[string]struct{ given Given expected Expected }{ } } By defining helper structs within the test function, we can keep the test cases clear and manageable. This approach is particularly useful for complex test cases. Additionally, if a function becomes too complex, it's a sign that it may be doing too much. In such cases, it's beneficial to split the function into smaller, more manageable pieces with clearly defined responsibilities. This simplifies both the function itself and its associated test cases. Prallel execution of test cases While all of above tests run very quickly, there is no reason that we shouldn't make use of t.Parallel. By adding this to each test, we can run all the tests that specify t.Parallel at the same time in parallel, provided they don't share state or dependencies. Conclusion Table driven tests are a popular model for testing in Go and a clear improvement to traditional testing of one function per case. As it offers a convenient way to test functions with various inputs. Since, the approach of "table" are easy and better to scale, most language testing frameworks have adopted some concept of dataprovider. They are great when you've a simply set of primitives as inputs, and you get out something simple in return. For anything more complex. use of a load of t.Runs inside a TestMyFunc, becomes necessary. Personally, a mix is fine, since it allows to test application in more than one scenario as well as are easy to maintain and document. FAQ's Are having dependencies not a total blocker to a table-driven test ? You can mock dependencies as additional fields in test structs, enabling comprehensive testing of functions with external dependencies in table-driven tests. How can I manage complex test cases efficiently? Utilize helper struct more, helper structs separate input parameters and expected results, enhancing test case clarity and manageability by structuring data into distinct categories for better organization and readability. Can we use Table Driven test in Golang only? While table driven tests are most popular with Go, TypeScript is also a perfect language for them. Being a strongly-typed superset of JavaScript, TypeScript allows us to explicitly define our input and expected types, an important feature to have when we are writing our tests. Table driven tests , also known as parameterized tests, have became very popular over the past few years, due to their ability to eliminate repetition. Table driven tests make it quite a bit easier to re-use the same values for different sets of tests by just moving the table outside of the scope of the test function. Different tests may benefit from the same input, and each test may have completely different configration, concurrency etc... Table driven tests The table structure The table structure Well, think of it as a systematic way of testing your code by providing a set of inputs and expected outputs in a structured table format. Instead of writing individual test cases for each input-output pair, we organize them neatly in a table, making it easier to manage and maintain our tests. Each row in the table represents a separate test case, and the columns represent different aspects or parameters of the test, such as inputs, expected outputs, and any other relevant conditions. Here's a simplified example of what a table-driven test might look like for a function that calculates the square of a number: package main import ( "testing" ) // We first have the function we may want to test // Consider the following function that calculates the addition // of two numbers func Add(a, b int) int { return a + b } /* We want to write tests that ensure our add function returns the correct result, no matter what numbers we provide. For the sake of keeping this blog to the point, I'm just going to include two test cases */ func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ "10+3": { inputA: 10, inputA: 3, expected: 13, }, "2+6": { inputA: 2, inputA: 6, expected: 8, }, } for _, tt := range tests { t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } package main import ( "testing" ) // We first have the function we may want to test // Consider the following function that calculates the addition // of two numbers func Add(a, b int) int { return a + b } /* We want to write tests that ensure our add function returns the correct result, no matter what numbers we provide. For the sake of keeping this blog to the point, I'm just going to include two test cases */ func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ "10+3": { inputA: 10, inputA: 3, expected: 13, }, "2+6": { inputA: 2, inputA: 6, expected: 8, }, } for _, tt := range tests { t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } Here, once we have declared the test cases via the tests map, the test function iterates over all the map entries, i.e. the test cases. The Add function is called in each iteration using the input arguments of the current test case. If the actual result doesn’t match the result expected by the test case, the test fails with an error message. tests Add Why Use Table-Driven Tests? Why Use Table-Driven Tests? Now, you might be wondering, why bother with table-driven tests? Here are a few reasons: Clarity: By organizing test cases in a table, it's easier to see all the inputs and expected outputs at a glance. This makes it simpler to understand what the test is doing. Scalability: As your codebase grows, you'll likely have more test cases to manage. Table-driven tests make it easier to add new test cases without cluttering your test code. Maintainability: Since test cases are organized in a structured format, it's easier to update or modify them when needed. Clarity: By organizing test cases in a table, it's easier to see all the inputs and expected outputs at a glance. This makes it simpler to understand what the test is doing. Clarity: Scalability: As your codebase grows, you'll likely have more test cases to manage. Table-driven tests make it easier to add new test cases without cluttering your test code. Scalability: Maintainability: Since test cases are organized in a structured format, it's easier to update or modify them when needed. Maintainability: What constitutes a table driven test? Typically every Table-driven tests case will have a description, input, and expected output values. Let's explore some techniques: Maps and slices Maps and slices Usually, test cases contain a name field to uniquely identify and describe the test case, and these are stored in slice. tests := []struct { name string inputA int inputB int expected int } tests := []struct { name string inputA int inputB int expected int } Although, slices are better approach when the test cases are simple, It's is preferred to use maps for storing when a lot of complex test cases start occupying too much space. Other benefit of using maps instead of the slice is that, it is an undefined iteration order. This means that test cases might be executed in a different order each time, which could expose issues in your test setup. tests := map[string]struct { inputA int inputB int expected int } tests := map[string]struct { inputA int inputB int expected int } It is also an issue because, if your tests expect certain conditions to be met in a specific order, using maps for table-driven tests might not be the best choice because the iteration order is not guaranteed. It's essential to consider this when designing your test setup to ensure reliable and consistent testing results. Solution to this that the order of test cases are maintained explicitly. This can be done by using an ordered slice of structs instead of a map. ordered slice of structs func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ {"Test case 1", 1, 2, 3}, {"Test case 2", -1, 1, 0}, {"Test case 3", 0, 0, 0}, } for _, tt := range tests { tt := tt // capture range variable t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } func TestAdd(t *testing.T) { tests := []struct { name string inputA int inputB int expected int }{ {"Test case 1", 1, 2, 3}, {"Test case 2", -1, 1, 0}, {"Test case 3", 0, 0, 0}, } for _, tt := range tests { tt := tt // capture range variable t.Run(tt.name, func(t *testing.T) { result := Add(tt.inputA, tt.inputB) if result != tt.expected { t.Errorf("got %d, want %d", result, tt.expected) } }) } } Here, we use a slice of structs to maintain the order of test cases. Inside the loop, we create a local copy of tt using := to ensure that each iteration has its own copy of the test case. This prevents any unintended sharing of state between iterations. tt := By using an ordered slice instead of a map, we ensure that the test cases are executed in the defined order, addressing the concern of non-deterministic behavior introduced by map iteration. This approach provides more reliability and consistency in test execution. Error handling & Logging When a comparison between an actual value and an expected value reveals a mismatch, there are two ways to handle it. Either we log an error message using Errorf or we terminate the test using Fatalf . Errorf Errorf Fatalf Fatalf In the first scenario, we log the error but allow the test to continue with subsequent checks. The latter option stops the entire test immediately upon encountering a failure. It's important to use these options thoughtfully. Generally, it's recommended to log the failure and continue with the test using Errorf. Only if proceeding with the test wouldn't make sense due to failed preconditions, should Fatalf be used to halt the test immediately. For instance, if the lengths of two slices, actual and test.expected, don't match, we immediately fail the test with Fatalf: if len(actual) != len(tt.expected) { t.Fatalf("lengths don't match: expected %v, got %v", len(tt.expected), len(actual)) } if len(actual) != len(tt.expected) { t.Fatalf("lengths don't match: expected %v, got %v", len(tt.expected), len(actual)) } Assuming the lengths are the same, we compare the values within the slices one by one. If a mismatch occurs, we log an error using Errorf but continue with the loop: for i, value := range tt.expected { if actual[i] != value { t.Errorf("values at index %d don't match: expected %v, got %v", i, value, actual[i]) } } for i, value := range tt.expected { if actual[i] != value { t.Errorf("values at index %d don't match: expected %v, got %v", i, value, actual[i]) } } Structured test cases To manage complex test cases with many parameters or dependencies, we can use helper structs. These structs store input parameters and expected results separately: func TestAdd(t *testing.T) { type Given struct { inputA int inputB int } type Expected struct { result int } tests := map[string]struct{ given Given expected Expected }{ } } func TestAdd(t *testing.T) { type Given struct { inputA int inputB int } type Expected struct { result int } tests := map[string]struct{ given Given expected Expected }{ } } By defining helper structs within the test function, we can keep the test cases clear and manageable. This approach is particularly useful for complex test cases. Additionally, if a function becomes too complex, it's a sign that it may be doing too much. In such cases, it's beneficial to split the function into smaller, more manageable pieces with clearly defined responsibilities. This simplifies both the function itself and its associated test cases. Prallel execution of test cases While all of above tests run very quickly, there is no reason that we shouldn't make use of t.Parallel . By adding this to each test, we can run all the tests that specify t.Parallel at the same time in parallel, provided they don't share state or dependencies. t.Parallel t.Parallel Conclusion Table driven tests are a popular model for testing in Go and a clear improvement to traditional testing of one function per case. As it offers a convenient way to test functions with various inputs. Since, the approach of "table" are easy and better to scale, most language testing frameworks have adopted some concept of dataprovider. They are great when you've a simply set of primitives as inputs, and you get out something simple in return. For anything more complex. use of a load of t.Runs inside a TestMyFunc, becomes necessary. Personally, a mix is fine, since it allows to test application in more than one scenario as well as are easy to maintain and document. t.Runs FAQ's Are having dependencies not a total blocker to a table-driven test ? You can mock dependencies as additional fields in test structs, enabling comprehensive testing of functions with external dependencies in table-driven tests. How can I manage complex test cases efficiently? Utilize helper struct more, helper structs separate input parameters and expected results, enhancing test case clarity and manageability by structuring data into distinct categories for better organization and readability. Can we use Table Driven test in Golang only? While table driven tests are most popular with Go, TypeScript is also a perfect language for them. Being a strongly-typed superset of JavaScript, TypeScript allows us to explicitly define our input and expected types, an important feature to have when we are writing our tests.