Skip to content

testing.md

Latest commit

 

History

History
131 lines (99 loc) · 5.04 KB

testing.md

File metadata and controls

131 lines (99 loc) · 5.04 KB
Raw

Testing

obsidian-cli.nvim ships a test suite using plenary.nvim's busted-style test harness. The suite runs in Neovim headless mode with no real Obsidian binary required — all CLI interactions are mocked.

What's covered

Spec file Tests Coverage
tests/util_spec.lua 25 Path helpers — is_absolute (Unix + Windows), absolute, relative_to_vault, in_vault, split_lines, expand
tests/config_spec.lua 9 Default values, deep-merge, buffer_options=false, immutability
tests/cli_spec.lua 18 Error detection (6 patterns), JSON parsing, "No X found." handling, vault path caching + reset
tests/setup_spec.lua 8 All 44 command registrations, keymap wiring on/off
tests/daily_spec.lua 14 Today, Yesterday, Tomorrow (with date math), Task CLI fallback, TasksToday JSON + filtering
tests/daily_advanced_spec.lua 5 In-buffer task append, plain text append, Todo alias, cursor jump, background-buffer non-jump
tests/navigation_spec.lua 15 Outline JSON, Links, Orphans, Deadends (all with empty cases), FollowLink (5 cases), Backlinks
tests/unresolved_spec.lua 11 Unresolved (string/comma-separated/missing sources), ResolveLink (create, alias, heading, block-ref, no-link, multi-link line)
tests/tasks_spec.lua 6 TaskToggle: [ ]↔[x], uppercase [X], no checkbox, first-only, correct line
tests/crud_spec.lua 34 New, NewFrom, Find, Recent, Search (5 cases), Rename (7 cases incl. collision + extension), Move, Delete, OpenInApp, Templates, TemplateInsert
tests/plugins_spec.lua 7 Command, CommandList, Install, Enable, Disable (confirm/cancel), Reload, RestrictedMode (query/on/off)
tests/completion_spec.lua 7+5p Source creation, trigger chars, enabled() (4 cases), context detection, auto-pairs interop. 5 tests pending (mock-fragile get_completions data flow)

Total: 122 tests across 12 spec files as of v0.0.5.

Running tests

All tests

make test

Or directly:

nvim --headless --noplugin -u tests/minimal_init.lua \
  -c "PlenaryBustedDirectory tests {minimal_init = 'tests/minimal_init.lua'}"

Single file

make test-file FILE=tests/util_spec.lua

Or:

nvim --headless --noplugin -u tests/minimal_init.lua \
  -c "PlenaryBustedFile tests/util_spec.lua"

Lint + test in one pass

make check

Runs stylua --check, selene, and the test suite. Use this before committing.

Mocking conventions

Tests that exercise CLI interactions (e.g., cli_spec.lua) replace vim.system with a stub that returns canned output:

local real_system = vim.system

local function mock_system(stdout, stderr, code)
  vim.system = function(_, _)
    return {
      wait = function()
        return { code = code or 0, stdout = stdout or "", stderr = stderr or "" }
      end,
    }
  end
end

local function restore_system()
  vim.system = real_system
end

This isolates the pure-Lua logic from the actual obsidian binary, so tests run in CI without needing Obsidian installed.

Don't shell out to the real CLI in tests. The binary isn't available in CI and would couple test stability to a specific Obsidian version. If you need to test a new CLI quirk, add a mock that reproduces the behavior and document the quirk in docs/architecture.md.

Writing new tests

Place new spec files in tests/ with the _spec.lua suffix. Plenary's busted harness auto-discovers them.

Basic structure:

-- tests/my_module_spec.lua
local my_module = require("obsidian-cli.my_module")

describe("my_module.my_function", function()
  it("does the thing", function()
    local result = my_module.my_function("input")
    assert.equals("expected", result)
  end)

  it("handles edge cases", function()
    assert.is_nil(my_module.my_function(nil))
    assert.is_nil(my_module.my_function(""))
  end)
end)

Common assertions (from plenary/luassert):

assert.equals(expected, actual)
assert.is_true(value)
assert.is_false(value)
assert.is_nil(value)
assert.is_not_nil(value)
assert.is_table(value)
assert.is_string(value)
assert.same({a=1, b=2}, result)         -- deep equality
assert.matches("pattern", string)       -- Lua pattern match

Use before_each/after_each for per-test setup/teardown (e.g., restoring mocks).

What NOT to test

  1. The Obsidian CLI itself — we assume it works as documented. Tests for CLI behavior belong upstream.
  2. Neovim APIs — we trust vim.fn.expand, vim.api.nvim_*, etc. Don't test that vim.split splits correctly.
  3. Plugin ecosystem integrations that require external state (Templater, Dataview) — those can only be tested manually against a live vault.
  4. Live CLI invocations — flaky, environment-dependent, doesn't belong in unit tests.

CI

Tests run in GitHub Actions via .github/workflows/ci.yml alongside stylua and selene. Any PR that fails tests won't merge.

The CI installs plenary.nvim fresh on each run (no caching needed — it's small). Neovim is installed via rhysd/action-setup-vim.