create runner with ChatGPT-Codex by using the runner-agents.md from the core

.gitea/workflows/tests.yml

@@ -0,0 +1,20 @@
+name: tests
+
+on:
+  push:
+    branches:
+      - "**"
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Neovim (AppImage)
+        run: |
+          curl -L -o nvim.appimage https://github.com/neovim/neovim/releases/download/v0.11.4/nvim.appimage
+          chmod +x nvim.appimage
+          sudo mv nvim.appimage /usr/local/bin/nvim
+      - name: Run tests
+        run: bash run_test.sh

AGENTS.md

@@ -0,0 +1,247 @@
+# runner-agents.md — test-samurai Runner-API
+
+Ziel: Diese Datei beschreibt die öffentliche Runner-API, die ein neuer Runner
+implementieren muss, damit alle Commands vollständig unterstützt werden.
+
+## Modulform (Pflicht)
+
+- Der Runner ist ein Lua-Modul, das eine Table mit Funktionen zurückgibt.
+- Beispiel:
+  - `local runner = {}`
+  - `return runner`
+- Der Modulpfad in `runner_modules` muss exakt zum Dateipfad unter `lua/` passen.
+  - Beispiel: `lua/test-samurai-go-runner/init.lua` -> `require("test-samurai-go-runner")`.
+  - `lua/init.lua` wäre `require("init")` und ist kollisionsanfällig; vermeiden.
+
+## Pflichtfunktionen (volle Command- und Keymap-Unterstützung)
+
+- `is_test_file(bufnr) -> boolean`
+  - Wird für die Runner-Auswahl genutzt.
+- `find_nearest(bufnr, row, col) -> spec|nil, err?`
+  - Für `TSamNearest`.
+  - `spec.file` muss gesetzt sein.
+  - Bei Fehler/kein Treffer: `nil, "reason"` zurückgeben.
+- `build_command(spec) -> command_spec`
+  - Für `TSamNearest`.
+- `build_file_command(bufnr) -> command_spec`
+  - Für `TSamFile`.
+- `build_all_command(bufnr) -> command_spec`
+  - Für `TSamAll`.
+- `build_failed_command(last_command, failures, scope_kind) -> command_spec`
+  - Für `TSamFailedOnly`.
+- `parse_results(output) -> results`
+  - Fallback-Parser für vollständigen Output.
+- `output_parser() -> { on_line, on_complete }`
+  - Muss Streaming unterstützen (Listing wird live befüllt).
+- `parse_test_output(output) -> table`
+  - Detail-Ausgabe pro Test (Detail-Float und `<cr>` im Listing).
+- `collect_failed_locations(failures, command, scope_kind) -> items`
+  - Quickfix-Unterstützung (Keymap `<leader>qn`).
+
+## Output-Parsing (Listing + Summary)
+
+Beide Varianten müssen vorhanden sein:
+
+- `parse_results(output) -> results`
+  - `results` muss enthalten:
+    - `passes` (Array von Namen)
+    - `failures` (Array von Namen)
+    - `skips` (Array von Namen)
+  - Optional:
+    - `display = { passes = {}, failures = {}, skips = {} }`
+    - `failures_all` (für Streaming-Parser, um alle bisherigen Failures zu liefern)
+  - Wenn `display` fehlt, werden `passes/failures/skips` direkt im Listing angezeigt.
+- oder `output_parser() -> { on_line, on_complete }`
+  - `on_line(line, state)` kann `results` liefern (siehe oben).
+  - `on_complete(output, state)` kann `results` liefern (siehe oben).
+  - `on_line` muss Ergebnisse liefern, damit das Listing-Float immer gestreamt wird.
+
+Hinweis: `parse_results` darf intern `output_parser().on_complete` nutzen, aber beide Funktionen müssen existieren.
+
+## Listing-Gruppierung fuer Parent/Subtests
+
+- Wenn Testnamen das Format `Parent/Subtest` oder verschachtelt `Parent/Sub/Subtest` haben
+  und der Parent ebenfalls in den Ergebnissen vorhanden ist, gruppiert das Listing:
+  - Parent kommt vor seinen direkten Kindern.
+  - Mehrstufige Subtests werden hierarchisch gruppiert (Parent -> Kind -> Enkel).
+  - Die Reihenfolge der Kinder folgt der Eingangsreihenfolge des Runners.
+  - Ohne Parent-Eintrag bleibt die normale Reihenfolge erhalten.
+
+## Detail-Output (TSamShowOutput / <cr> im Listing)
+
+- `parse_test_output(output) -> table`
+  - Rückgabeform:
+    - `{ [test_name] = { "line1", "line2", ... } }`
+  - `test_name` muss mit `results.*` korrespondieren (gleiches Namensschema).
+
+## Quickfix-Unterstützung (Failures)
+
+- `collect_failed_locations(failures, command, scope_kind) -> items`
+  - `items`: Array von Quickfix-Items
+    - `{ filename = "...", lnum = <number>, col = <number>, text = "..." }`
+
+## Erwartete Datenformen
+
+- `command_spec`:
+  - `{ cmd = { "binary", "arg1", ... }, cwd = "..." }`
+  - `cmd` darf nicht leer sein.
+  - `cwd` ist optional; wenn nicht gesetzt, nutzt der Core das aktuelle CWD.
+- `spec` (von `find_nearest`):
+  - Muss mindestens `file` enthalten, z. B.:
+    - `{ file = "...", cwd = "...", test_name = "...", full_name = "...", kind = "..." }`
+- `results` (für Listing-Float + Keymaps):
+  - `{ passes = { "Name1", ... }, failures = { "Name2", ... }, skips = { "Name3", ... } }`
+  - Optional: `display = { passes = { "DisplayName1", ... }, failures = { "DisplayName2", ... }, skips = { "DisplayName3", ... } }`
+  - `failures` steuert `[ FAIL ]`-Zeilen im Listing und wird von `<leader>nf`/`<leader>pf` genutzt.
+- `items` (für Quickfix):
+  - `{ { filename = "...", lnum = 1, col = 1, text = "..." }, ... }`
+  - Wird von `<leader>qn` verwendet.
+
+## Keymaps (Datenlieferung)
+
+- `<leader>nf` / `<leader>pf`
+  - benötigt `[ FAIL ]`-Einträge im Listing.
+  - Runner muss `results.failures` (und optional `display.failures`) liefern.
+- `<leader>qn`
+  - springt in die Quickfix-Liste.
+  - Runner muss `collect_failed_locations` implementieren und gültige `items` liefern.
+- `<cr>` im Listing
+  - öffnet Detail-Float.
+  - Runner muss `parse_test_output` liefern und Testnamen konsistent zu `results.*` halten.
+
+## Optional empfohlene Metadaten
+
+- `name` (String)
+  - Wird in Fehlermeldungen und Logs angezeigt.
+- `framework` (String)
+  - Wird zur Framework-Auswahl (z. B. JS) genutzt.
+
+## Prompt-Beispiel
+
+"Erstelle mir anhand der `runner-agents.md` einen neuen Runner für Rust."
+
+## Minimaler Runner-Skeleton (Template)
+
+```lua
+local runner = {
+  name = "my-runner",
+  framework = "my-framework",
+}
+
+function runner.is_test_file(bufnr)
+  return false
+end
+
+function runner.find_nearest(bufnr, row, col)
+  return nil, "no test call found"
+end
+
+function runner.build_command(spec)
+  return { cmd = { "echo", "not-implemented" }, cwd = spec.cwd }
+end
+
+function runner.build_file_command(bufnr)
+  return { cmd = { "echo", "not-implemented" } }
+end
+
+function runner.build_all_command(bufnr)
+  return { cmd = { "echo", "not-implemented" } }
+end
+
+function runner.build_failed_command(last_command, failures, scope_kind)
+  return { cmd = { "echo", "not-implemented" }, cwd = last_command and last_command.cwd or nil }
+end
+
+function runner.parse_results(output)
+  return { passes = {}, failures = {}, skips = {}, display = { passes = {}, failures = {}, skips = {} } }
+end
+
+function runner.output_parser()
+  return {
+    on_line = function(_line, _state)
+      return nil
+    end,
+    on_complete = function(_output, _state)
+      return runner.parse_results(_output)
+    end,
+  }
+end
+
+function runner.parse_test_output(output)
+  return {}
+end
+
+function runner.collect_failed_locations(failures, command, scope_kind)
+  return {}
+end
+
+return runner
+```
+
+## Checkliste für neue Runner
+
+- is_test_file implementiert
+- find_nearest implementiert (setzt `spec.file`)
+- build_command implementiert
+- build_file_command implementiert
+- build_all_command implementiert
+- build_failed_command implementiert
+- parse_results implementiert
+- output_parser implementiert (Streaming)
+- parse_test_output implementiert
+- collect_failed_locations implementiert
+- command_spec `{ cmd, cwd }` korrekt zurückgegeben
+
+## Projekt- und Prozessanforderungen
+
+- Rolle: **TDD-first Entwickler**.
+  - Jede neue Funktion, jedes neue Kommando und jede Verhaltensänderung muss durch Tests abgesichert sein.
+  - Nach jeder Code-Änderung Tests via `bash run_test.sh` ausführen und bei Fehlern korrigieren, bis alles grün ist.
+- **Nicht raten**:
+  - Bei unklaren oder mehrdeutigen Anforderungen Arbeit stoppen und Klarstellung verlangen.
+  - TODO/NOTE im Code ist zulässig, stilles Raten nicht.
+- **Keine stillen Änderungen**:
+  - Bestehende Features dürfen nicht unbemerkt geändert oder ersetzt werden.
+  - Notwendige Anpassungen zur Koexistenz mehrerer Features müssen klar erkennbar sein.
+- Antworten immer auf Deutsch.
+- Eine englischsprachige `README.md` ist zu erstellen und wird bei Änderungen automatisch aktualisiert.
+- TDD-Vorgaben (aus `AGENTS.md`) uebernehmen:
+  - Neue Funktionen/Commands/Verhaltensaenderungen muessen getestet werden.
+  - Tests nach jeder Code-Aenderung ausfuehren.
+  - Im Runner erstellter Quellcode ist ebenfalls zu testen.
+  - Eine eigene `run_test.sh` wird im Runner-Repo angelegt.
+- Eine Gitea-Action ist zu erstellen, die bei jedem Push die Tests ausfuehrt.
+  - Neovim wird per AppImage installiert (kein `apt`).
+  - Runner laeuft auf `gitea-act-runner` mit Raspberry Pi 5 (ARM).
+  - Beispiel (anpassbarer Workflow):
+
+```yaml
+name: tests
+
+on:
+  push:
+    branches:
+      - "**"
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Neovim (AppImage)
+        run: |
+          curl -L -o nvim.appimage https://github.com/neovim/neovim/releases/download/v0.11.4/nvim.appimage
+          chmod +x nvim.appimage
+          sudo mv nvim.appimage /usr/local/bin/nvim
+      - name: Run tests
+        run: bash run_test.sh
+```
+
+## Zusaetzliche Runner-Guidelines (framework-agnostisch)
+
+- **Testnamen-Konvention:** Runner sollen eine konsistente, dokumentierte Full-Name-Bildung verwenden (z. B. `Parent/Subtest`), inklusive Mehrfach-Nesting. Diese Konvention muss in `results.*`, `parse_test_output` und `collect_failed_locations` uebereinstimmen.
+- **TSamNearest-Prioritaet:** Falls moeglich, gelten folgende Regeln: Test-Block > Describe/Context-Block > File-Command. Das Verhalten muss getestet werden (Cursor im Test, zwischen Tests, ausserhalb von Describe/Context).
+- **Reporter-Payload-Schema:** Wenn ein Custom-Reporter verwendet wird, soll dessen JSON-Payload dokumentiert und stabil sein (z. B. `{ name, status, file, location, output }`), damit Parser/Quickfix/Detail-Output konsistent bleiben.
+- **Failed-Only-Logik:** Failed-Only muss auf den letzten Fehlermeldungen basieren und nur die fehlerhaften Tests erneut ausfuehren. Die Pattern-Strategie (z. B. Titel-only vs. Full-Name) muss getestet werden.
+- **CI-Installations-Snippet:** Die Neovim-Installation in CI soll als „authoritative snippet“ behandelt werden und in Runner-Repos 1:1 uebernommen werden.

README.md

@@ -0,0 +1,61 @@
+# test-samurai-mocha-runner
+
+Mocha.js runner for the test-samurai Neovim plugin.
+
+## Features
+
+- Detects Mocha test files by checking `package.json` dependencies.
+- Supports nearest, file, all, and failed-only commands.
+- Streams results via Mocha's `json-stream` reporter.
+- Provides quickfix locations and per-test output.
+- Uses `--grep` with escaped patterns to match titles safely, even when running through `npm test`.
+- Executes tests via `npx mocha` for direct Mocha invocation.
+- `TSamAll` runs `test/**/*.{test,spec}.{t,j}s` to discover tests reliably.
+
+## Full Name Convention
+
+The runner builds stable full names using the active suite stack joined by `/`,
+followed by the test title:
+
+```
+Suite/Subsuite/Test
+```
+
+This convention is used consistently in `results.*`, `parse_test_output`, and
+`collect_failed_locations`. Failed-only runs translate `/` back to spaces for
+Mocha's `--grep` matching. Avoid `/` in your titles if you rely on that mapping.
+
+## Reporter Payload
+
+The runner expects Mocha's built-in `json-stream` reporter, which emits one JSON
+object per line. It consumes the following fields when present:
+
+```
+{
+  "event": "suite" | "suite end" | "pass" | "fail" | "pending",
+  "title": "Test or suite title",
+  "fullTitle": "Mocha full title",
+  "file": "/path/to/test.js",
+  "err": { "message": "string", "stack": "string" }
+}
+```
+
+## Usage
+
+Add the module to your test-samurai configuration:
+
+```lua
+require("test-samurai").setup({
+  runner_modules = {
+    "test-samurai-mocha-runner",
+  },
+})
+```
+
+## Development
+
+Run tests:
+
+```bash
+bash run_test.sh
+```

lua/test-samurai-mocha-runner/init.lua

@@ -0,0 +1,503 @@
+local runner = {
+  name = "test-samurai-mocha-runner",
+  framework = "javascript",
+}
+
+local function read_file(path)
+  local ok, lines = pcall(vim.fn.readfile, path)
+  if not ok then
+    return nil
+  end
+  return table.concat(lines, "\n")
+end
+
+local function find_project_root(start_path)
+  if not start_path or start_path == "" then
+    return nil
+  end
+  local dir = vim.fn.fnamemodify(start_path, ":h")
+  local prev = nil
+  while dir and dir ~= prev do
+    local candidate = dir .. "/package.json"
+    if vim.fn.filereadable(candidate) == 1 then
+      return dir
+    end
+    prev = dir
+    dir = vim.fn.fnamemodify(dir, ":h")
+  end
+  return nil
+end
+
+local function load_package_json(root)
+  if not root then
+    return nil
+  end
+  local content = read_file(root .. "/package.json")
+  if not content then
+    return nil
+  end
+  local ok, data = pcall(vim.json.decode, content)
+  if not ok then
+    return nil
+  end
+  return data
+end
+
+local function has_mocha_dependency(pkg)
+  if not pkg then
+    return false
+  end
+  local deps = pkg.dependencies or {}
+  local dev_deps = pkg.devDependencies or {}
+  return deps.mocha ~= nil or dev_deps.mocha ~= nil
+end
+
+local function count_char(line, char)
+  local _, count = line:gsub(char, "")
+  return count
+end
+
+local function extract_title(line, names)
+  for _, name in ipairs(names) do
+    local pattern = name .. "%s*%(%s*(['\"])(.-)%1"
+    local _, title = line:match(pattern)
+    if title and title ~= "" then
+      return name, title
+    end
+  end
+  return nil, nil
+end
+
+local function build_full_name(suites, title)
+  if #suites == 0 then
+    return title
+  end
+  return table.concat(suites, "/") .. "/" .. title
+end
+
+local function build_mocha_title(suites, title)
+  if #suites == 0 then
+    return title
+  end
+  return table.concat(suites, " ") .. " " .. title
+end
+
+local function parse_buffer_scope(lines, row)
+  local depth = 0
+  local stack = {}
+  local suite_names = { "describe", "context" }
+  local test_names = { "it", "test" }
+
+  for i = 1, row do
+    local line = lines[i] or ""
+    local kind, title = extract_title(line, suite_names)
+    if title then
+      local opens = count_char(line, "{")
+      local closes = count_char(line, "}")
+      local new_depth = depth + opens - closes
+      table.insert(stack, { kind = "suite", title = title, depth = new_depth, line = i })
+      depth = new_depth
+    else
+      kind, title = extract_title(line, test_names)
+      local opens = count_char(line, "{")
+      local closes = count_char(line, "}")
+      local new_depth = depth + opens - closes
+      if title then
+        table.insert(stack, { kind = "test", title = title, depth = new_depth, line = i })
+      end
+      depth = new_depth
+    end
+
+    while #stack > 0 and stack[#stack].depth > depth do
+      table.remove(stack)
+    end
+  end
+
+  return stack
+end
+
+local function suite_titles_from_stack(stack)
+  local suites = {}
+  for _, entry in ipairs(stack) do
+    if entry.kind == "suite" then
+      table.insert(suites, entry.title)
+    end
+  end
+  return suites
+end
+
+local function find_last_of_kind(stack, kind)
+  for i = #stack, 1, -1 do
+    if stack[i].kind == kind then
+      return stack[i]
+    end
+  end
+  return nil
+end
+
+local function build_grep_pattern(title)
+  local pattern = title or ""
+  pattern = pattern:gsub("%s+", ".*")
+  return pattern
+end
+
+local function ensure_state(state)
+  state.results = state.results or { passes = {}, failures = {}, skips = {} }
+  state.seen = state.seen or { passes = {}, failures = {}, skips = {} }
+  state.outputs = state.outputs or {}
+  state.locations = state.locations or {}
+  state.suite_stack = state.suite_stack or {}
+  state.mocha_titles = state.mocha_titles or {}
+  return state
+end
+
+local function extract_location(stack)
+  if not stack then
+    return nil
+  end
+  local filename, lnum, col = stack:match("%(([^%s%(%):]+):(%d+):(%d+)%)")
+  if not filename then
+    filename, lnum, col = stack:match("at%s+[^%s]+%s+([^%s%(%):]+):(%d+):(%d+)")
+  end
+  if not filename then
+    filename, lnum, col = stack:match("([^%s%(%):]+):(%d+):(%d+)")
+  end
+  if not filename then
+    return nil
+  end
+  return {
+    filename = filename,
+    lnum = tonumber(lnum),
+    col = tonumber(col),
+  }
+end
+
+local function record_result(state, event, title, payload)
+  local full_name = build_full_name(state.suite_stack, title)
+  local mocha_title = build_mocha_title(state.suite_stack, title)
+  state.mocha_titles[full_name] = mocha_title
+  local added = false
+
+  if event == "pass" then
+    if not state.seen.passes[full_name] then
+      state.seen.passes[full_name] = true
+      table.insert(state.results.passes, full_name)
+      added = true
+    end
+  elseif event == "fail" then
+    if not state.seen.failures[full_name] then
+      state.seen.failures[full_name] = true
+      table.insert(state.results.failures, full_name)
+      added = true
+    end
+  elseif event == "pending" then
+    if not state.seen.skips[full_name] then
+      state.seen.skips[full_name] = true
+      table.insert(state.results.skips, full_name)
+      added = true
+    end
+  end
+
+  if event == "fail" and payload then
+    local lines = {}
+    local message = nil
+    local stack = nil
+    if type(payload.err) == "table" then
+      message = payload.err.message
+      stack = payload.err.stack
+    elseif type(payload.err) == "string" then
+      message = payload.err
+    end
+    if type(payload.stack) == "string" and payload.stack ~= "" then
+      stack = payload.stack
+    end
+
+    if message and message ~= "" then
+      table.insert(lines, message)
+    end
+    if stack and stack ~= "" then
+      for _, line in ipairs(vim.split(stack, "\n")) do
+        table.insert(lines, line)
+      end
+    end
+    if #lines > 0 then
+      state.outputs[full_name] = lines
+    end
+
+    local location = extract_location(stack or "")
+    if location then
+      location.text = message or "test failure"
+      state.locations[full_name] = location
+    end
+  end
+
+  if added then
+    return full_name
+  end
+  return nil
+end
+
+local function parse_json_stream_line(line, state)
+  local ok, payload = pcall(vim.json.decode, line)
+  if not ok or type(payload) ~= "table" then
+    return nil
+  end
+
+  local event = payload.event
+  local data = payload
+  if event == nil and payload[1] then
+    event = payload[1]
+    data = payload[2] or {}
+  end
+  if event == "suite" then
+    if data.title and data.title ~= "" then
+      table.insert(state.suite_stack, data.title)
+    end
+  elseif event == "suite end" then
+    if data.title and data.title ~= "" and #state.suite_stack > 0 then
+      table.remove(state.suite_stack)
+    end
+  elseif event == "pass" or event == "fail" or event == "pending" then
+    local name = record_result(state, event, data.title or "", data)
+    if name then
+      return { event = event, name = name }
+    end
+  end
+
+  return data
+end
+
+local function parse_json_stream_output(output)
+  local state = ensure_state({})
+  for line in output:gmatch("[^\n]+") do
+    parse_json_stream_line(line, state)
+  end
+  return state
+end
+
+local function project_root_for_buf(bufnr)
+  local file = vim.api.nvim_buf_get_name(bufnr)
+  if file == "" then
+    return nil
+  end
+  return find_project_root(file)
+end
+
+function runner.is_test_file(bufnr)
+  local file = vim.api.nvim_buf_get_name(bufnr)
+  if file == "" then
+    return false
+  end
+  local root = find_project_root(file)
+  if not root then
+    return false
+  end
+  local pkg = load_package_json(root)
+  return has_mocha_dependency(pkg)
+end
+
+function runner.find_nearest(bufnr, row, _col)
+  local file = vim.api.nvim_buf_get_name(bufnr)
+  if file == "" then
+    return nil, "buffer has no name"
+  end
+  local root = find_project_root(file)
+  if not root then
+    return nil, "no package.json found"
+  end
+  local lines = vim.api.nvim_buf_get_lines(bufnr, 0, -1, false)
+  local stack = parse_buffer_scope(lines, row)
+  local suites = suite_titles_from_stack(stack)
+
+  local test_block = find_last_of_kind(stack, "test")
+  if test_block then
+    local full_name = build_full_name(suites, test_block.title)
+    return {
+      file = file,
+      cwd = root,
+      test_name = test_block.title,
+      full_name = full_name,
+      mocha_full_title = build_mocha_title(suites, test_block.title),
+      kind = "test",
+    }
+  end
+
+  local suite_block = find_last_of_kind(stack, "suite")
+  if suite_block then
+    local full_name = table.concat(suites, "/")
+    return {
+      file = file,
+      cwd = root,
+      test_name = suite_block.title,
+      full_name = full_name,
+      mocha_full_title = table.concat(suites, " "),
+      kind = "suite",
+    }
+  end
+
+  return {
+    file = file,
+    cwd = root,
+    kind = "file",
+  }
+end
+
+function runner.build_command(spec)
+  if not spec.cwd or spec.cwd == "" then
+    return { cmd = { "echo", "no package.json found" } }
+  end
+  if spec.kind == "file" then
+    return runner.build_file_command(spec.file)
+  end
+
+  local grep = spec.mocha_full_title or spec.full_name or spec.test_name
+  return {
+    cmd = {
+      "npx",
+      "mocha",
+      "--reporter",
+      "json-stream",
+      "--grep",
+      build_grep_pattern(grep),
+      spec.file,
+    },
+    cwd = spec.cwd,
+  }
+end
+
+function runner.build_file_command(bufnr_or_file)
+  local file = bufnr_or_file
+  if type(bufnr_or_file) == "number" then
+    file = vim.api.nvim_buf_get_name(bufnr_or_file)
+  end
+  local cwd = file and find_project_root(file) or nil
+  if not cwd then
+    return { cmd = { "echo", "no package.json found" } }
+  end
+  return {
+    cmd = {
+      "npx",
+      "mocha",
+      "--reporter",
+      "json-stream",
+      file,
+    },
+    cwd = cwd,
+  }
+end
+
+function runner.build_all_command(bufnr)
+  local cwd = project_root_for_buf(bufnr)
+  if not cwd then
+    return { cmd = { "echo", "no package.json found" } }
+  end
+  return {
+    cmd = {
+      "npx",
+      "mocha",
+      "--reporter",
+      "json-stream",
+      "test/**/*.{test,spec}.{t,j}s",
+    },
+    cwd = cwd,
+  }
+end
+
+function runner.build_failed_command(last_command, failures, _scope_kind)
+  local cwd = last_command and last_command.cwd or nil
+  if not failures or #failures == 0 then
+    return {
+      cmd = {
+        "npx",
+        "mocha",
+        "--reporter",
+        "json-stream",
+      },
+      cwd = cwd,
+    }
+  end
+
+  local patterns = {}
+  for _, failure in ipairs(failures) do
+    local mocha_title = runner._last_mocha_titles and runner._last_mocha_titles[failure]
+    if not mocha_title then
+      mocha_title = failure:gsub("/", " ")
+    end
+    table.insert(patterns, build_grep_pattern(mocha_title))
+  end
+
+  return {
+    cmd = {
+      "npx",
+      "mocha",
+      "--reporter",
+      "json-stream",
+      "--grep",
+      table.concat(patterns, "|"),
+    },
+    cwd = cwd,
+  }
+end
+
+function runner.parse_results(output)
+  local state = parse_json_stream_output(output)
+  return state.results
+end
+
+function runner.output_parser()
+  return {
+    on_line = function(line, state)
+      state = ensure_state(state or {})
+      local info = parse_json_stream_line(line, state)
+      runner._last_locations = state.locations
+      runner._last_mocha_titles = state.mocha_titles
+      if not info or not info.event or not info.name then
+        return nil
+      end
+      local results = { passes = {}, failures = {}, skips = {} }
+      if info.event == "pass" then
+        results.passes = { info.name }
+      elseif info.event == "fail" then
+        results.failures = { info.name }
+        results.failures_all = vim.deepcopy(state.results.failures)
+      elseif info.event == "pending" then
+        results.skips = { info.name }
+      end
+      return results
+    end,
+    on_complete = function(output, state)
+      state = ensure_state(state or {})
+      local parsed = parse_json_stream_output(output)
+      runner._last_locations = parsed.locations
+      runner._last_mocha_titles = parsed.mocha_titles
+      return nil
+    end,
+  }
+end
+
+function runner.parse_test_output(output)
+  local state = parse_json_stream_output(output)
+  return state.outputs
+end
+
+function runner.collect_failed_locations(failures, _command, _scope_kind)
+  local items = {}
+  if not failures or not runner._last_locations then
+    return items
+  end
+
+  for _, name in ipairs(failures) do
+    local location = runner._last_locations[name]
+    if location then
+      table.insert(items, {
+        filename = location.filename,
+        lnum = location.lnum,
+        col = location.col,
+        text = location.text or name,
+      })
+    end
+  end
+  return items
+end
+
+return runner

run_test.sh

@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+nvim --headless -u NONE -c "lua dofile('tests/run.lua')" -c "qa"

tests/run.lua

@@ -0,0 +1,252 @@
+local function ok(condition, message)
+  if not condition then
+    error(message or "assertion failed")
+  end
+end
+
+vim.opt.runtimepath:append(vim.fn.getcwd())
+package.path = package.path
+  .. ";"
+  .. vim.fn.getcwd()
+  .. "/lua/?.lua;"
+  .. vim.fn.getcwd()
+  .. "/lua/?/init.lua"
+
+local function eq(actual, expected, message)
+  if actual ~= expected then
+    error(message or string.format("expected %s, got %s", tostring(expected), tostring(actual)))
+  end
+end
+
+local function write_file(path, content)
+  local dir = vim.fn.fnamemodify(path, ":h")
+  vim.fn.mkdir(dir, "p")
+  vim.fn.writefile(vim.split(content, "\n"), path)
+end
+
+local function make_buffer(path, content)
+  local bufnr = vim.api.nvim_create_buf(false, true)
+  vim.api.nvim_buf_set_name(bufnr, path)
+  vim.api.nvim_buf_set_lines(bufnr, 0, -1, false, vim.split(content, "\n"))
+  return bufnr
+end
+
+local function with_project(structure, fn)
+  local root = vim.fn.tempname()
+  vim.fn.mkdir(root, "p")
+  for rel_path, content in pairs(structure) do
+    write_file(root .. "/" .. rel_path, content)
+  end
+  fn(root)
+end
+
+local function test_is_test_file_with_mocha_dependency()
+  with_project({
+    ["package.json"] = [[{"devDependencies":{"mocha":"10.0.0"}}]],
+  }, function(root)
+    local runner = require("test-samurai-mocha-runner")
+    local bufnr = make_buffer(root .. "/test/sample.spec.js", "describe('x', function() {})")
+    ok(runner.is_test_file(bufnr), "expected mocha project to be detected")
+  end)
+end
+
+local function test_is_test_file_without_mocha_dependency()
+  with_project({
+    ["package.json"] = [[{"devDependencies":{"jest":"29.0.0"}}]],
+  }, function(root)
+    local runner = require("test-samurai-mocha-runner")
+    local bufnr = make_buffer(root .. "/test/sample.spec.js", "describe('x', function() {})")
+    ok(not runner.is_test_file(bufnr), "expected non-mocha project to be ignored")
+  end)
+end
+
+local function test_find_nearest_priorities()
+  with_project({
+    ["package.json"] = [[{"devDependencies":{"mocha":"10.0.0"}}]],
+  }, function(root)
+    local runner = require("test-samurai-mocha-runner")
+    local content = table.concat({
+      "describe(\"Math\", function() {",
+      "  it(\"adds\", function() {",
+      "    expect(1).to.equal(1)",
+      "  })",
+      "",
+      "  it(\"subs\", function() {",
+      "    expect(1).to.equal(1)",
+      "  })",
+      "})",
+      "",
+      "const value = 1",
+    }, "\n")
+    local bufnr = make_buffer(root .. "/test/math.spec.js", content)
+
+    local spec_test = assert(runner.find_nearest(bufnr, 3, 0))
+    eq(spec_test.kind, "test", "expected test kind")
+    eq(spec_test.full_name, "Math/adds", "expected full test name")
+
+    local spec_suite = assert(runner.find_nearest(bufnr, 5, 0))
+    eq(spec_suite.kind, "suite", "expected suite kind")
+    eq(spec_suite.full_name, "Math", "expected suite name")
+
+    local spec_file = assert(runner.find_nearest(bufnr, 11, 0))
+    eq(spec_file.kind, "file", "expected file kind")
+  end)
+end
+
+local function test_missing_package_json_errors()
+  local runner = require("test-samurai-mocha-runner")
+  local content = table.concat({
+    "describe(\"Math\", function() {",
+    "  it(\"adds\", function() {",
+    "    expect(1).to.equal(1)",
+    "  })",
+    "})",
+  }, "\n")
+  local bufnr = make_buffer("/tmp/no-root.test.js", content)
+
+  local spec, err = runner.find_nearest(bufnr, 2, 0)
+  ok(spec == nil, "expected no spec without package.json")
+  eq(err, "no package.json found", "expected package.json error")
+
+  local command = runner.build_file_command(bufnr)
+  eq(command.cmd[1], "echo")
+  eq(command.cmd[2], "no package.json found")
+end
+
+local function test_command_building()
+  with_project({
+    ["package.json"] = [[{"devDependencies":{"mocha":"10.0.0"}}]],
+  }, function(root)
+    local runner = require("test-samurai-mocha-runner")
+    local spec = {
+      file = root .. "/test/math.spec.js",
+      cwd = root,
+      kind = "test",
+      mocha_full_title = "Math adds",
+      full_name = "Math/adds",
+    }
+    local command = runner.build_command(spec)
+    eq(command.cmd[1], "npx")
+    eq(command.cmd[2], "mocha")
+    ok(vim.tbl_contains(command.cmd, "--reporter"), "expected reporter flag")
+    ok(vim.tbl_contains(command.cmd, "json-stream"), "expected json-stream reporter")
+    ok(vim.tbl_contains(command.cmd, "--grep"), "expected grep flag")
+  end)
+end
+
+local function test_build_all_command_includes_glob()
+  with_project({
+    ["package.json"] = [[{"devDependencies":{"mocha":"10.0.0"}}]],
+  }, function(root)
+    local runner = require("test-samurai-mocha-runner")
+    local bufnr = make_buffer(root .. "/test/sample.spec.js", "")
+    local command = runner.build_all_command(bufnr)
+    ok(vim.tbl_contains(command.cmd, "test/**/*.{test,spec}.{t,j}s"), "expected test glob")
+  end)
+end
+
+local function test_output_parser_and_locations()
+  local runner = require("test-samurai-mocha-runner")
+  local output_lines = {
+    [=[["suite",{"title":"Math","fullTitle":"Math"}]]=],
+    [=[["pass",{"title":"adds","fullTitle":"Math adds","file":"/tmp/math.spec.js"}]]=],
+    [=[["pass",{"title":"adds","fullTitle":"Math adds","file":"/tmp/math.spec.js"}]]=],
+    [=[["fail",{"title":"subs","fullTitle":"Math subs","file":"/tmp/math.spec.js","err":"oops","stack":"Error: oops\n    at Context.<anonymous> (/tmp/math.spec.js:10:5)"}]]=],
+    [=[["pending",{"title":"skips","fullTitle":"Math skips","file":"/tmp/math.spec.js"}]]=],
+    [=[["suite end",{"title":"Math","fullTitle":"Math"}]]=],
+  }
+  local parser = runner.output_parser()
+  local state = {}
+  local aggregated = { passes = {}, failures = {}, skips = {} }
+  for _, line in ipairs(output_lines) do
+    local results = parser.on_line(line, state)
+    if results then
+      for _, name in ipairs(results.passes or {}) do
+        table.insert(aggregated.passes, name)
+      end
+      for _, name in ipairs(results.failures or {}) do
+        table.insert(aggregated.failures, name)
+      end
+      for _, name in ipairs(results.skips or {}) do
+        table.insert(aggregated.skips, name)
+      end
+    end
+  end
+  eq(aggregated.passes[1], "Math/adds", "expected pass name")
+  eq(#aggregated.passes, 1, "expected no duplicate passes")
+  eq(aggregated.failures[1], "Math/subs", "expected failure name")
+  eq(aggregated.skips[1], "Math/skips", "expected skip name")
+
+  local items = runner.collect_failed_locations(aggregated.failures, {}, "nearest")
+  eq(items[1].filename, "/tmp/math.spec.js", "expected filename from stack")
+  eq(items[1].lnum, 10, "expected line from stack")
+  eq(items[1].col, 5, "expected col from stack")
+end
+
+local function test_failed_only_command()
+  local runner = require("test-samurai-mocha-runner")
+  runner._last_mocha_titles = {
+    ["Math/adds"] = "Math adds",
+    ["Math/subs"] = "Math subs",
+  }
+  local command = runner.build_failed_command({ cwd = "/tmp" }, { "Math/adds", "Math/subs" }, "file")
+  eq(command.cmd[1], "npx")
+  ok(vim.tbl_contains(command.cmd, "--grep"), "expected grep for failed-only")
+  local grep_index
+  for idx, value in ipairs(command.cmd) do
+    if value == "--grep" then
+      grep_index = idx + 1
+      break
+    end
+  end
+  ok(grep_index, "expected grep argument")
+  ok(command.cmd[grep_index]:match("Math%.%*adds"), "expected grep pattern for first failure")
+end
+
+local function test_parse_results_and_test_output()
+  local runner = require("test-samurai-mocha-runner")
+  local output = table.concat({
+    [=[["suite",{"title":"Math","fullTitle":"Math"}]]=],
+    [=[["fail",{"title":"subs","fullTitle":"Math subs","file":"/tmp/math.spec.js","err":"oops","stack":"Error: oops\n    at Context.<anonymous> (/tmp/math.spec.js:10:5)\n    at processImmediate"}]]=],
+    [=[["suite end",{"title":"Math","fullTitle":"Math"}]]=],
+  }, "\n")
+  local results = runner.parse_results(output)
+  eq(results.failures[1], "Math/subs", "expected failure from parse_results")
+
+  local outputs = runner.parse_test_output(output)
+  ok(outputs["Math/subs"], "expected test output for failure")
+  local found = false
+  for _, line in ipairs(outputs["Math/subs"]) do
+    if line:match("oops") then
+      found = true
+      break
+    end
+  end
+  ok(found, "expected error output line")
+end
+
+local function test_output_parser_on_complete_returns_nil()
+  local runner = require("test-samurai-mocha-runner")
+  local parser = runner.output_parser()
+  local state = {}
+  local output = [=[["pass",{"title":"adds","fullTitle":"Math adds","file":"/tmp/math.spec.js"}]]=]
+  local results = parser.on_complete(output, state)
+  ok(results == nil, "expected nil on_complete to avoid duplicate listings")
+end
+
+local tests = {
+  test_is_test_file_with_mocha_dependency,
+  test_is_test_file_without_mocha_dependency,
+  test_find_nearest_priorities,
+  test_missing_package_json_errors,
+  test_command_building,
+  test_build_all_command_includes_glob,
+  test_output_parser_and_locations,
+  test_failed_only_command,
+  test_parse_results_and_test_output,
+  test_output_parser_on_complete_returns_nil,
+}
+
+for _, test_fn in ipairs(tests) do
+  test_fn()
+end