Test setup

Neotest-golang is tested using neotest-plenary by running lua unit tests and integration tests.

Tests can be executed either from within Neovim (using neotest-plenary) or in the terminal. To run tests from the terminal, use these commands (requires Taskfile):

task test - Run all tests
task test-file -- spec/unit/convert_spec.lua - Run a single test file

Tests timing out

Nvim-nio will hit a hard-coded 2000 millisecond timeout if you are running the entire test suite. I have mitigated this in the bootstraping script and also opened an issue about that in nvim-neotest/nvim-nio#30.

Test execution flow

When you run tests, the following sequence occurs:

Bootstrap script runs first (spec/bootstrap.lua): - Resets Neovim's runtime path to a clean state - Downloads and installs required plugins (e.g. plenary.nvim, neotest, nvim-nio, nvim-treesitter) - Configures the test environment with proper packpath - Installs the Go treesitter parser - Initializes neotest with the golang adapter - Ensures PlenaryBustedDirectory command is available
PlenaryBustedDirectory executes (via the -c flag): - Discovers all *_spec.lua files in the spec/ directory - For each test file, creates a fresh Neovim instance using the minimal init - Some integration tests use spec/helpers/integration.lua to run actual Go tests
Minimal init runs per test (spec/minimal_init.lua): - Resets runtime path to clean state for each test - Sets basic Neovim options (no swapfile, correct packpath) - Provides isolated environment for individual test execution

Neovim vs Busted execution

The below outlines why BustedPlenary was chosen instead of Busted.

This setup uses Neovim's -c flag to execute commands within Neovim's context, rather than Busted's -l flag which loads Lua files externally.

Strengths of plenary-busted approach:

Tests run within actual Neovim instances, providing authentic plugin behavior
Full access to Neovim APIs (vim.*, treesitter, etc.) during testing
Each test gets a clean Neovim environment via the minimal init
Integration tests can interact with real neotest functionality
No need to mock Neovim-specific behavior

Weaknesses compared to pure Busted:

Slower execution due to Neovim startup overhead per test
More complex setup and bootstrapping process
Harder to debug test failures (headless Neovim environment)
Potential for Neovim version-specific test behavior
More memory and resource intensive

Writing tests

Unit tests

Unit tests (in ./spec/unit) are meant to test specific lua function capabilities within a small scope, but sometimes with a large amount of permutations in terms of input arguments or output.

Integration tests

Integration tests (in ./spec/integration) performs end-to-end validation by executing actual Go tests via the neotest-golang adapter.

The general workflow of adding a new integration test:

Add a new yourtestname_test.go file in tests/go/internal/yourpkgname
Add a new lua integration test in spec/integration/yourpkgname[_yourtestname]_spec.lua and from it, execute all tests in a dir, a file or specifiy individual test(s):

Lua

local integration = require("spec.helpers.integration")

-- Run all tests in a directory
local result = integration.execute_adapter_direct("/path/to/directory")

-- Run all tests in a file
local result = integration.execute_adapter_direct("/path/to/file_test.go")

-- Run a specific test function
local result = integration.execute_adapter_direct("/path/to/file_test.go::TestFunction")

-- Run a specific subtest
local result = integration.execute_adapter_direct("/path/to/file_test.go::TestFunction::\"SubTest\"")

-- Run a nested subtest
local result = integration.execute_adapter_direct("/path/to/file_test.go::TestFunction::\"SubTest\"::\"TableTest\"")

-- Use blocking/synchronous execution (legacy, but could potentially be useful for debugging)
local result = integration.execute_adapter_direct("/path/to/file_test.go", { use_blocking = true })

Best practices

When writing tests...

Always use gotestsum as runner to prevent flaky tests due to failures with parsing JSON in stdout. Set any other options which might be required by the test.
Follow the Arrange, Act, Assert (AAA) pattern.
Assert on the full wanted test results (want) from the gotten test results (got) wrapped in vim.inspect for easier debugging. Set want.somefield = got.somefield if you don't care about asserting explicity values.
Invoke the Go test by calling execute_adapter_direct using the Neotest.Position.type as argument, mimicing what happens when require("neotest").run().run() executes:
Run all tests in dir: /path/to/folder
Run all tests in file: /path/to/folder/file_test.go
Run test (and/or sub-tests): /path/to/folder/file_test.go::TestSomething or /path/to/folder/file_test.go::TestSomething::"TestSubTest"
Use { use_blocking = true } option for synchronous execution

Debugging Testify Suite Issues

The testify suite feature in neotest-golang requires special handling to map receiver methods to their parent suites. This section documents debugging techniques for testify-related issues.

Understanding Testify Architecture

Testify suites use Go receiver methods that are represented as flat test IDs prefixed with the suite name:

Go

// Receiver type
type ExampleTestSuite struct {
    suite.Suite
}

// Test methods (discovered by testify queries)
func (suite *ExampleTestSuite) TestExample() { ... }
func (suite *ExampleTestSuite) TestExample2() { ... }

// Suite runner function (discovered by regular Go queries, but hidden in tree)
func TestExampleTestSuite(t *testing.T) {
    suite.Run(t, new(ExampleTestSuite))
}

The neotest tree structure uses a flat representation with prefixed test IDs:

Text Only

- file_test.go
  ├── TestExampleTestSuite/TestExample (test)
  ├── TestExampleTestSuite/TestExample2 (test)
  └── TestExampleTestSuite/TestSubTest (test)
    └── "subtest" (test)
  └── TestTrivial (regular test)

Note: The suite runner function (TestExampleTestSuite) is not shown in the tree to avoid confusion and improve usability.

Debugging Tree Modification Issues

When testify suites aren't working correctly, the issue is usually in the tree modification process. Add debug output to key functions:

1. Debug Query Discovery

Add this to lua/neotest-golang/query.lua in the detect_tests function:

Lua

-- DEBUG: Test if treesitter finds testify methods directly
print("=== DEBUG TESTIFY QUERY DETECTION ===")
if options.get().testify_enabled == true then
  local testify_matches = testify.query.run_query_on_file(file_path, testify.query.test_method_query)
  print("Testify test methods found:")
  for name, matches in pairs(testify_matches) do
    print("  " .. name .. ": " .. #matches .. " matches")
    for i, match in ipairs(matches) do
      print("    " .. i .. ". " .. match.text)
    end
  end

  local namespace_matches = testify.query.run_query_on_file(file_path, testify.query.namespace_query)
  print("Namespace matches found:")
  for name, matches in pairs(namespace_matches) do
    print("  " .. name .. ": " .. #matches .. " matches")
    for i, match in ipairs(matches) do
      print("    " .. i .. ". " .. match.text)
    end
  end
end
print("=======================================")

2. Debug Tree Structure

Add this to lua/neotest-golang/features/testify/tree_modification.lua in the modify_neotest_tree function:

Lua

-- DEBUG: Check what's in the original tree
print("=== DEBUG TESTIFY TREE MODIFICATION ===")
local positions = {}
for i, pos in tree:iter() do
  table.insert(positions, pos)
end

print("Original tree has " .. #positions .. " positions:")
for i, pos in ipairs(positions) do
  local pos_type = pos.type or "nil"
  local pos_name = pos.name or "nil"
  local pos_id = pos.id or "nil"
  print("  " .. i .. ". " .. pos_type .. " [" .. pos_name .. "] - " .. pos_id)
end

-- DEBUG: Check lookup table
print("Lookup table:")
for file, data in pairs(lookup_table) do
  print("  File: " .. file)
  if data.replacements then
    for receiver, suite in pairs(data.replacements) do
      print("    " .. receiver .. " -> " .. suite)
    end
  end
end
print("=======================================")

3. Debug Method-to-Receiver Mapping

For testify suites with duplicate method names (like multiple TestExample methods), add this debug output:

Lua

-- DEBUG: Show method to receiver mapping
print("=== METHOD TO RECEIVER MAPPING ===")
print("Method positions by name:")
for method, positions in pairs(method_positions) do
  print("  " .. method .. ": " .. #positions .. " instances")
  for i, pos in ipairs(positions) do
    print("    " .. i .. ". " .. pos.receiver)
  end
end
print("=====================================")

Common Issues and Solutions

Issue: Testify methods not discovered

Symptom: Original tree only shows suite functions, no receiver methods
Cause: Testify queries using wrong capture names or invalid syntax
Solution: Ensure testify queries use @test.name and @test.definition (not @test_name)
Check: Enable debug output in query detection to see if methods are found

Issue: Methods not properly prefixed

Symptom: Test IDs missing suite name prefix (e.g., TestMethod instead of SuiteName/TestMethod)
Cause: Tree modification not properly renaming test IDs with suite prefix
Solution: Check method-to-receiver mapping and ID renaming logic
Check: Debug tree structure to verify test IDs have correct format

Issue: Duplicate method names causing confusion

Symptom: Some testify methods missing or assigned to wrong suites
Cause: Multiple receivers with same method names (e.g., two TestExample methods)
Solution: Use position/range information to distinguish duplicate method names
Check: Debug method-to-receiver mapping to see if all instances are found

Issue: Position IDs incorrect

Symptom: Test execution fails or doesn't match expected IDs
Cause: ID replacement logic not updating position IDs correctly
Solution: Ensure ID renaming updates test IDs from ::MethodName to ::SuiteName/MethodName (using / separator to match go test -run format)
Check: Compare expected vs actual position IDs in test assertions

Treesitter Query Compatibility

When upgrading nvim-treesitter (especially from master to main branch):

Capture name format: Main branch requires dots (@test.name) not underscores (@test_name)
Statement list wrappers: Some queries need additional (statement_list ...) wrappers
Query validation: Test queries individually with testify.query.run_query_on_file

Testing Testify Changes

When modifying testify functionality:

Enable testify: Set testify_enabled = true in test options
Use integration tests: Run spec/integration/testifysuites_positions_spec.lua
Check Go command: Verify the generated go test command targets suite functions
Validate tree structure: Ensure namespace hierarchy matches expected test position IDs
Test edge cases: Files with multiple suites, duplicate method names, subtests