v0.10.2

Patch Changes

• #103 2e6c3bf - Clarify declared states, conditional skips and expected failures in CLI help, API comments, recipes and CI reports.
  Correct config inheritance and failure-scope descriptions. Add first-use guidance, concepts and troubleshooting,
  preserve existing documentation URLs, and publish the diff schema alongside the inventory schema.

• #104 cf08b52 - Add a checked complete library example, a CLI/library/MCP walkthrough, compact agent guidance and feature-level
  compatibility notes. Prioritize essential Markdown references and distinguish illustrative excerpts from runnable
  code by their copy controls.

• #105 9325281 - Add focused documentation guides for each supported test framework and separate shared scanner limits from runner-specific exceptions.

• #107 ef35e6e - Add checked examples for Cypress component and E2E selection, WebdriverIO with Jasmine, Playwright projects and Bun case expansion to the framework guides.

• #106 30dc03f - Keep skipped-test queries consistent with the first scan, add focused Vitest, Jest and Mocha examples, and link framework guides to relevant next steps.

• #108 7891cc0 - Read context modifiers and subtests in named node:test callbacks, retain valid parent test IDs when callbacks appear before their callers, recognize Windows patterns naming linked node:test folders, and reject TOML inline tables with missing commas.

  Stop claiming that Jest runs tests when a required local preset is missing. Config-based discovery returns no files, while explicit globs retain the declarations with a file-not-run diagnostic.

  Validate node:test concurrency values, report unknown values, and inherit concurrency through parent tests without resetting it for null or undefined options.

v0.10.1

Patch Changes

• #99 df8968d - Keep Playwright declaring modifiers when functions are stored in variables or selected inline. Read logical choices with the same rules as the other runners, and classify each nested branch once to avoid exponential traversal.

  Match Vite's alias merge precedence and preserve object aliases when mixed with lists. Mark imports as unfollowed when an unreadable alias entry or find could match them.

  Ignore annotation and step line numbers in inventory diffs, so moving unchanged tests to another line does not report metadata changes.

• #100 fa10d39 - Read locally declared describe callbacks passed by name, preserving suite placement, inherited state, lexical values and expanded table arguments. Keep suite hashes, dependencies and nested parent links complete, including when a child is written before its calling parent. Apply the existing helper traversal limits and runner callback checks to these callbacks.

v0.10.0

Minor Changes

• #96 86e1765 - Compare two complete saved scans with test-inventory diff old.json new.json or diffInventories(before, after).
  Reports additions, removals, renamed and moved tests, changed code and dependencies, state, tags and other metadata.
  Matches ids first, then body hashes with file, suite or dependency context, leaving ambiguous matches unmatched.
  Includes exact body clone groups, JSON Schema, and escaped Markdown output through --format markdown or
  formatDiffMarkdown. Hash comparisons and rename matching require identical tool versions; incomplete scans warn.

Patch Changes

• #96 86e1765 - Keep unreadable config aliases as dependency gaps rather than silently treating their imports as installed packages.
  Read string concatenation and templates in config paths. Follow config aliases in setup paths, Jasmine helper globs,
  Cypress's default and configured support files, Bun preload files, and Node script preloads in dependencies and
  dependencyHash, so --related keeps tests affected by changes to those files.

• #96 86e1765 - Apply Mocha's focus filtering, including Cypress and WebdriverIO on Mocha, so an outer describe.only does not focus
  tests that a more specific focus excludes. A root focused test in standalone Mocha excludes child suites across
  files. Contain refused calls inside node:test and Bun describe callbacks: Node keeps earlier tests and tests outside
  the describe; Bun loses that describe only. Vitest reads allowOnly per project, does not fail unconditionally skipped
  focused tests, and refuses modifiers after a called each or for table.

v0.9.1

Patch Changes

• #92 0001eea - Bun: a test whose function is passed by name, followed by a timeout or options, as in
  test('reprices a basket', repricesBasket, 5_000), is read as a test. The scan took the timeout for the function and
  reported a test-without-body error, listing every test in the file as not loaded, though Bun runs them.

• #92 0001eea - A __proto__ key in a bunfig.toml or .mocharc.yml is read as a plain key, as Bun and Mocha read it. A
  bunfig.toml with a [__proto__] table changed Object.prototype for the whole process, so every later scan in it,
  such as those of a long-running MCP server, read settings the project never set. A .mocharc.yml one gave its map the
  settings inside it. TOML keys named constructor or toString are read too, where they were refused.

• #92 0001eea - Cypress: specPattern and excludeSpecPattern are read as globby, which Cypress uses, reads them. A specPattern that
  starts with ! leaves files out, where the scan took it to match every other file and listed nearly the whole project;
  an excludeSpecPattern drops its !, where the scan left out everything else; and a folder in either stands for every
  file below it, where it matched nothing.

• #92 0001eea - Jasmine: a describe callback with a TypeScript this parameter, a parameter with a default value or a rest parameter,
  as function (this: object), is no longer reported as taking arguments. Jasmine checks the function's length, which
  counts none of these, and runs the specs; the scan listed them as not loaded.

• #92 0001eea - Jasmine: an xdescribe inside an fdescribe stays skipped, as in Jasmine, where the scan listed its specs as active;
  a focus inside the xdescribe still runs. WebdriverIO on Jasmine now follows Jasmine's focus rules: a fit inside an
  xdescribe runs, and in an fdescribe with a fit below it only the fit has isOnly, where the scan applied
  Mocha's rules.

• #92 0001eea - A <rootDir> entry in Jest's moduleDirectories, as ['node_modules', '<rootDir>/src'], is followed as a folder to
  look in, as Jest does. The scan looked for a folder named <rootDir>/src and left the files out of dependencies
  without a warning.

• #92 0001eea - Jest's moduleNameMapper replaces the whole import name, as Jest does, also when its pattern matches only part of it,
  as '\\.css$' does. The scan replaced only the matched part, so a mapped style or file mock was missing from
  dependencies and its import got a dependency-not-followed warning. Vite's aliases still replace only the part they
  match.

• #92 0001eea - Test files reached through a symbolic link are read the way each runner reads them. Jasmine and Bun follow such links
  and run the file once, as Vitest and Mocha do, so the file is no longer left out; Bun names it by its own path, as its
  reports do. node:test runs a linked file, twice when its own path matches too, but a wildcard in its pattern doesn't
  enter a linked folder. The symlink-ignored warning now names the runner in use, where it named Playwright for every
  runner but Jest.

• #92 0001eea - Mocha: an absolute glob in spec, as a .mocharc.js builds it with path.join(__dirname, 'test', '**', '*.spec.js'),
  runs the files it matches. The scan matched it against paths relative to the config's folder, found nothing, and
  reported no-files-matched.

• #92 0001eea - Mocha, Cypress and WebdriverIO on Mocha: this.skip() in an afterEach hook no longer marks the tests as skipped.
  The test has passed by the time the hook runs, and Mocha reports it as a pass.

• #92 0001eea - Mocha: a .only in a project whose config sets forbid-only (or forbidOnly) gets a forbidden-only error, as Mocha
  stops the run before any test runs, or a warning when the setting depends on a condition, as Playwright's forbidOnly
  does.

• #92 0001eea - A .mocharc.yml is read as js-yaml, Mocha's reader, reads it. Files that Mocha reads and the scan refused, which lost
  every setting and so the files to test, are now read: a byte order mark, a comment or a %YAML line before ---, a
  ... at the end, quoted list items with commas, as ['test/**/*.{ts,js}'], YAML's escapes in double quotes, as
  \x41, and a tab or an escaped quote followed by # in quotes. A quote inside a value, as in grep: buy 'limit # note,
  no longer hides the comment after it, -a.js on its own line is a value, not a list item, and what Mocha refuses is
  reported rather than read: a key set twice, a tab in the indent, a: b in a value, a list inside a list, and values
  that start with @, ` or %. -.5 is a string, as js-yaml reads it. A long line is read in linear time; one of
  160 KB took 5 seconds.

• #92 0001eea - node:test and Bun: a skipped describe skips every test inside it, over the test's own todo, as node:test creates none of
  them, and in Bun the outermost skipped or todo describe decides, over the describes inside it and the test's own state.
  The scan took the state nearest the test, so test.todo() in describe.skip() was todo, and in Bun test.skip() in
  describe.todo() was skipped.

• #92 0001eea - node:test: a test whose function is passed by name, as test('fills a market order', fillsMarketOrder) or
  test(fillsMarketOrder), gets that function's bodyHash, and without a title the function's name, as Node.js names it.
  The scan took the function for the test's options, with an unresolved-option warning and no hash, or for a title it
  couldn't work out.

• #92 0001eea - node:test: the scan reads the node --test commands of every script the test script runs through npm run, pnpm,
  yarn, bun run, npm-run-all, run-s or run-p, as in "test": "npm run test:unit && npm run test:integration".
  It read only the first script that runs node --test, and left out the files of the others without a warning.

• #92 0001eea - From Playwright 1.65, test.skip('reason'), test.fixme('reason') and test.fail('reason') with only a string give
  that string as the record's reason, in a test body and at file or describe level, and no longer a skip-without-body
  warning, as Playwright 1.65 takes the string as the description. An empty string still applies nothing. Older versions
  keep the old reading.

• #92 0001eea - A dependency-not-followed warning for an import a config maps to a file that doesn't exist gives that file's path
  relative to the root, as every path in the output is. It gave the absolute path, as path.resolve() or <rootDir>
  built it, so the output differed from machine to machine.

• #92 0001eea - A test or describe function chosen in place, as in (process.env.CI ? it.skip : it)('books a fill', ...) or
  (process.env.CI && it.skip || it)(...), is read as one kept in a variable: only what both choices share applies. The
  scan dropped these tests without a warning, for every runner but Playwright, whose case is still open. Jasmine
  variables that hold xit, fit, xdescribe or fdescribe, as const later = xit, keep the x or f, as the docs
  said; their tests were dropped too.

• #92 [0001eea](https://github.com/Huldoser/test-inventory...

v0.9.0

Minor Changes

• #88 2d370c9 - Bun 1.2.23 and newer: --framework bun, also found from a package.json script that runs bun test or a
  bunfig.toml. Its version comes from .bun-version, packageManager or engines.bun, else @types/bun or
  bun-types. The scan reads Bun's globals and imports of bun:test, @jest/globals and vitest, which Bun rewrites;
  chains of .skip, .todo, .failing, .only, .concurrent, .serial and .each in any order, refusing the
  combinations Bun refuses; the conditions .if, .skipIf, .todoIf, .failingIf, .concurrentIf and .serialIf;
  options second or third, a number as the timeout or { timeout, retry, repeats }; and .each titles as Bun formats
  them. The files are Bun's *.test.*, *_test.*, *.spec.* and *_spec.* under bunfig.toml's [test] root, less
  pathIgnorePatterns, with its retry and concurrentTestGlob applied. A small TOML reader reads bunfig.toml.

  The oldest version a framework supports may now name a patch, as Bun's 1.2.23 does.

• #87 ae2bd90 - dependencyHash follows imports through the config a file runs with: Vite's resolve.alias and Vitest's
  test.alias, as an object or a list of { find, replacement } with a name or a regex, and Jest's moduleNameMapper,
  modulePaths and moduleDirectories, with <rootDir>. They were read from tsconfig paths only, so tests that
  imported through such aliases had hasUnfollowedDependencies and a hash that missed those files. A mapping to no file
  is a dependency-not-followed warning that names it.

• #83 9ef9c59 - Cypress 13 and newer: --framework cypress, also found from a cypress package or a cypress.config.* file. Tests are
  read as Mocha's, with Cypress's test config before the function: retries, of which cypress run takes runMode,
  browser as a skip condition, @cypress/grep's tags, and from Cypress 16 an error for env, which Cypress refuses
  there. this.timeout() is not read, as Cypress has only command timeouts; this.retries() in a describe and
  .retries() after it() or describe() are errors, as Cypress throws on them while it loads the spec. The config's
  testing types are the projects, e2e and component, each with its specPattern, excludeSpecPattern and retries.

  A project whose package.json lists cypress next to another runner, as Jest or Vitest for unit tests, now needs
  --framework, as the scan asks for one whenever it finds several runners.

• #82 04e7328 - Jasmine 4 and newer: --framework jasmine, also found from a jasmine or karma-jasmine package or a
  spec/support/jasmine.* config. The scan reads describe, it and their x and f forms with Jasmine's focus, a
  spec without a function as pending, pending(reason) in a spec or its beforeEach or afterEach, from Jasmine 7
  notApplicable(reason), the hooks, and a spec's timeout. The config's spec_dir and spec_files, with !
  exclusions, decide the files; without a config, as in Karma and Angular projects, the files given by glob run. A call
  Jasmine refuses stops the whole run at the top level of a file, and is an error that leaves the file's tests as they
  are inside a describe, which Jasmine catches.

• #81 85fb8a9 - Mocha 10 and newer: --framework mocha, also found from a mocha package, a .mocharc.* file or a mocha key in
  package.json. The scan reads describe, context, it, specify, their x forms, .only and .skip, a test
  without a function as pending, ui: tdd's suite and test, the hooks, this.timeout(), this.retries() and
  this.skip() where this is Mocha's, and .timeout() and .retries() after it() and describe(). The config's
  spec, extension, recursive, ignore and file decide the files, and its timeout and retries apply to every
  test; a YAML config is read too. A describe without a function, it() or describe() without a title, and an
  unknown chain such as it.concurrent are errors, as Mocha then runs no test.

  A project whose package.json lists mocha next to another runner now needs --framework, as the scan asks for one
  whenever it finds several runners.

• #86 1085fb7 - node:test on Node.js 22 and newer: --framework node, also found from a package.json script that runs node --test
  or a node.config.json. Its version is the Node.js version the project pins in .node-version, .nvmrc, Volta's
  node or engines.node. The scan reads test, it, describe and suite from node:test, imported or required,
  with .skip, .todo, .only and from 24.14 .expectFailure, and the options skip, todo, only, timeout,
  concurrency, expectFailure and from 24.19 and 26.2 tags; t.skip() and t.todo() in a test's body; and the
  hooks. The files are those the script gives node --test, or Node.js's default patterns, with TypeScript files from
  22.18.

  A test declared in a test's body, with t.test() or test(), is a subtest that node:test runs while it runs the test:
  it is listed with the test's title in suitePath and the test's id in the new key parentTestId, which is null for
  every other test.

• #80 cae53c1 - A new invalid-fixture-argument error: a rest property among a Playwright function's fixture arguments, as
  async ({ page, ...fixtures }) => {} in a test, hook, test.skip() condition or test.extend() fixture written in
  the test file. Playwright refuses it and runs no tests at all, so the file's tests are now notLoaded instead of
  active. A nested rest, as ({ page: { mouse, ...pointer } }), and a test.step() callback are allowed, as
  Playwright allows them.

• #85 90df9aa - scan() takes a signal, an AbortSignal: the scan checks it before it starts and between files, and rejects with
  the signal's reason once it aborts, so a server can stop a scan its client no longer waits for.

• #85 90df9aa - A Vitest .only that allowOnly makes fail is a forbidden-only diagnostic, as for Playwright's forbidOnly: an
  error when the config sets test.allowOnly to false, and a warning when it is unset, as Vitest then turns it off in
  CI, or depends on a condition. Vitest marks such a test failed instead of running it.

• #84 c68bac7 - WebdriverIO 9 and newer: --framework webdriverio, also found from a @wdio/cli package or a wdio.conf.* file. The
  config's config export, ES or CommonJS, decides the files: specs, with the nested lists that group files into one
  worker, less exclude, matched as WebdriverIO matches them. Its framework decides how the tests are read, Mocha by
  default or Jasmine, with mochaOpts.timeout, retries and ui, or jasmineOpts.defaultTimeoutInterval, on every
  test; Cucumber is reported, as its feature files aren't read. A number after a test's function is WebdriverIO's retry
  count, and with Jasmine a second one, after the timeout. Mocha's and Jasmine's version gates follow the versions
  WebdriverIO depends on. With Mocha, a call the runner refuses stops the file loading there, and the tests declared
  above it still run. The config's rootDir and each capability's own specs and exclude are read too.

  A project whose package.json lists @wdio/cli next to another runner now needs --framework, as the scan asks for
  one whenever it finds several runners.

Patch Changes

• #85 90df9aa - expectCount counts Playwright's expect.configure(...), called in place as expect.configure({ timeout: 5_000 })(locator)
  or kept in a variable as const slowExpect = expect.configure(...); it missed both.

• #83 9ef9c59 - The error for an unknown framework option or --framework value names every runner the scan reads; it still named
  only Playwright, Vitest and Jest.

• #89 fd4b13a - --help names all nine runners --framework takes, and says which files node:test and Bun run without globs.

• #80 [cae53c1](https://github.com/Huld...

v0.8.0

Minor Changes

• #75 f6690bb - --framework (library: framework) is now optional. Without it, the scan takes the runner the project uses: the one
  whose package a package.json in the root or below it lists, or whose config file is there, else one a package.json
  above the root lists. A project with several runners, or none, gets a usage error (a TypeError in the library) that
  names what it found.

• #75 f6690bb - Jest 29 and 30: --framework jest (library: framework: 'jest') lists the tests of Jest files with the same keys as
  Playwright and Vitest tests. It reads Jest's globals, fit, xit, xtest, fdescribe and xdescribe among them,
  imports and require() of @jest/globals, the chains Jest has, such as test.concurrent.only.each, .failing as
  expectedToFail, the timeout passed third, jest.retryTimes() as retries for every test of the file (from 30.5,
  with entireDescribe, of its describe) and waitBeforeRetry as retryDelay, and @group pragmas as tags, as
  jest-runner-groups reads them. --expand-cases titles each case as jest-each 29 and 30 do. Calls Jest refuses, such
  as test.only.concurrent, test('title') without a function or an async describe, mark the file's tests notLoaded.

  Without globs, the test files come from the Jest config, read as Jest reads it: jest.config.*, else package.json,
  with roots, testMatch, testRegex, the ignore patterns, moduleFileExtensions, projects named by
  displayName, testTimeout and injectGlobals, also through Jest 30.3's defineConfig() and mergeConfig();
  without any config, Jest's defaults apply. framework in the output
  can now be jest, an open set, so schemaVersion stays 1.

  The cases-not-expanded message no longer names Vitest, as Jest tables can get it too.

Patch Changes

• #75 f6690bb - require() of a local file is followed as an import is: tables, titles and other values it exports, with
  module.exports or exports.name, are read, and helper functions it exports that declare tests are followed. A
  helper function a shared file imports and exports again, as an index.js of helpers does, is followed to where it is
  written, instead of getting an unresolved-test-import error. A helpers file may export a test function it takes from
  the runner's globals, as export const pendingTest = xit, or from the framework's module, with its modifiers.

• #75 f6690bb - A call the runner refuses, such as test.only.skip(...), that is only reached under a condition no longer marks every
  test of its file notLoaded, as the file loads unless the condition holds; the call is still reported. A test function
  chosen with &&, || or ??, as in const onCi = process.env.CI && test.skip || test, is read like one chosen with
  ?:. In a config, a list with a spread the scan can't work out keeps the items it can.

• #75 f6690bb - A config that require()s a local file, as a CommonJS config does with a shared base, now reads what that file
  exports, as an import of it already did, and a JSON file, required or imported, is read as its value.

• #75 f6690bb - A table or loop gets its caseCount when its length is known even if one of its values isn't, as in
  [NaN, Infinity].forEach(...) or a table with a call in a row. A test or describe titled by a function or class
  written in place, as describe(function OrderBook() {}, ...), takes its name, as the runners do. Code the parser
  fails to hand over, such as a JSX name written as a Unicode escape, gets a parse-error instead of an
  internal-error.

• #75 f6690bb - dependencyHash is faster on large projects: the files a test file and its describes reach are put together once for
  all their tests, and a workspace's packages are found once, not once per folder that imports one. A scan of the Nx
  repository with 17,571 tests went from 19.4 to 11.9 seconds. Hashes are unchanged.

v0.7.0

Minor Changes

• #72 cf784f0 - --expand-cases (library: expandCases: true) lists each case of a .each or .for table and each pass of a loop
  whose values the scan can work out as its own record, with the title the runner gives it. Vitest's titles follow the
  rules of the project's Vitest version, 3.0 to 5: %s, %d, %i, %f, %j, %o, %O, %c, %#, %$ and $name,
  with values quoted and cut as that version does; they match the names Vitest 3.0, 3.1, 3.2, 4.0, 4.1 and 5.0 list.
  Each case's values are bound in its callback, so titles, tags and conditions built from them are worked out per case,
  and an if, ?:, &&, || or ?? that reads nothing but them registers only the branch it takes. Loops in test
  and hook bodies and loops over a var are not expanded, and a file starts expansions up to 1,000 loop passes and
  describe cases.
  Suites and tests have a new key, caseIndex: the case's index from 0, as %# counts, or null without the option. A
  table that can't be expanded stays one record with a new cases-not-expanded warning. Without the option the output
  is as before, apart from caseIndex: null on every suite and test.

• #72 cf784f0 - Values imported from other files of the project are now worked out: a title, tag, annotation, reason, run setting or
  .each table built from a constant, enum or default export of another module, through re-exports, export *,
  import * as and JSON files, found the same way as helper functions (relative paths, tsconfig paths, #imports,
  workspace packages). Values from installed packages are still not read, and keep their diagnostic.

  A test in a for...of, for...in, forEach or map loop over values the scan can work out now has caseCount, the
  number of tests the runner registers, and no test-in-loop warning; the warning now marks a loop the scan can't
  count. Values also work out arithmetic on numbers, as in 30 * 1000, ?:, &&, ||, ??, comparisons,
  Object.keys(), Object.values(), Object.entries(), JSON.stringify(), String(), and a string's toUpperCase(),
  toLowerCase() and trim() and an array's join(). A skip condition that works out to a constant, such as
  side === 'short' in a helper called with 'buy', now sets the state it gives instead of being kept as a condition.

  Code after an if whose branch always returns or throws, as in if (!strategy.enabled) return; in a helper, is now
  under the opposite condition, so its tests have isInCondition and a loop around them no count. A tagged-template
  table is counted as Vitest splits it: outer pipes are columns, an incomplete last row is left out, and a template
  without values is one case.

  A test whose title was its source text and is now worked out gets a new id, once.

Patch Changes

• #71 bf8383f - A name destructured from a test object, as in const { describe, beforeEach } = test, is now read as that member:
  beforeEach(...) is a hook, describe(...) a describe and skip(...) a skipped test, for Playwright and Vitest,
  also nested, as in const { describe: { serial } } = test. Every such name was read as test itself, so a hook got
  test-without-body and every test of the file notLoaded, and the tests inside a describe were left out with
  nested-test warnings.

• #71 bf8383f - When the root and the folders above it don't have the framework, its version is now read from the shallowest
  package.json below the root that declares it and holds scanned files, as code/package.json in a monorepo, with
  that package's lockfile or installed copy first. The version was null there, so the newest rules applied: Vitest
  5's, which refuse a file that calls vi.mock inside a block, made every test of such a file notLoaded in a project
  on Vitest 3.

v0.6.1

Patch Changes

• #69 41ce74d - An import that matches a tsconfig paths pattern but none of its files, as an installed package does with
  "paths": { "*": ["./src/types/*"] }, now goes on to the usual lookup, as TypeScript does. Such imports were reported
  as unreadable, so every test that used a package or a built-in module such as node:fs got
  hasUnfollowedDependencies. An @/ or ~/ alias whose files are missing is still reported.

• #69 41ce74d - A file that only passes names on, as a barrel index.ts with export * from './fees' or
  export { OrderTicket } from './order-ticket' does, is now in the dependencies of the tests that import through it,
  so --related keeps them when the file changes. Which module a re-export points at is now part of dependencyHash, as
  an import's already was, so the hash of a test that imports through a re-export changes once.

• #69 41ce74d - An imported file that isn't text, such as import logo from './logo.png', is now taken by its bytes in
  dependencyHash. It was read as text, where every byte that isn't UTF-8 reads the same, so two different images could
  give the same hash.

• #69 41ce74d - A <script> block in a Vue or Svelte component whose type isn't JavaScript, such as
  <script type="application/ld+json"> in <svelte:head>, is no longer read as the component's code. It made the
  component unreadable, so the modules its real script imports were missing from dependencies.

• #68 fa9b97a - dependencyHash now follows what a Vue or Svelte single-file component imports, also the components and helpers only
  its template uses, so a change to a child component changes the hash of the tests that mount its parent. A component
  whose script doesn't parse is still taken by its text, and its tests get hasUnfollowedDependencies.

• #69 41ce74d - A file with brackets nested more than 500 deep, such as a generated fixture, now gets a parse-error instead of
  crashing the scan. oxc-parser overflows the native stack on such a file, about 5,000 deep, and the process dies with a
  segmentation fault that no error handling can catch, so no inventory was written.

• #69 41ce74d - expectCount now counts expect.element(), the assertion of Vitest browser mode, so a browser test that checks the
  page only with await expect.element(locator).toBeVisible() no longer looks like a test without assertions.

• #69 41ce74d - An import written with a .jsx extension now finds the .tsx file it refers to, as .js already found .ts, so
  import { OrderHeader } from './OrderHeader.jsx' is followed instead of marking the test with
  hasUnfollowedDependencies.

• #69 41ce74d - From Playwright 1.57, the config's tag is now read: every test gets it first in its tags, and a project's grep or
  grepInvert matches it, so a project with grepInvert: /@nightly/ in a config with tag: '@nightly' no longer runs
  the tests. When the scan can't work out the tag, as for process.env.CI_ENVIRONMENT_NAME, it gets a
  config-value-unknown warning, and the tests of a project with a grep get projects: null.

• #69 41ce74d - From Playwright 1.64, a project with default: false is no longer listed in a test's projects, as Playwright runs it
  only with --project, unless a project that runs lists it in dependencies or teardown. A default the scan can't
  work out gets a config-value-unknown warning, and the tests it may run get projects: null.

• #69 41ce74d - Playwright's defineConfig(base, config) is now merged as Playwright merges it: a project of config merges into the
  project of base with the same name instead of replacing all of base's projects, and expect, use and build
  merge key by key. A setup project from a base config is no longer lost, and an expect.toHaveScreenshot.pathTemplate
  from the base still marks the tests that compare screenshots with hasUnfollowedDependencies.

• #69 41ce74d - A Vitest retry object in a test's or describe's options before Vitest 4.1 now gets an unresolved-option warning
  and retries: null. Vitest 4.0 and 3.2 never run such a test; the scan used to ignore the object and give the
  config's retry count instead.

• #69 41ce74d - A Vitest test's own skip, only, todo or fails option now replaces the one its tags set, as Vitest does, so
  test('title', { tags: ['flaky'], skip: false }, fn) is active instead of skipped, and a conditional skip replaces
  the tag's unconditional one.

• #69 41ce74d - An import that starts with @/ or ~/, as a Vite resolve.alias allows, now gives the test
  hasUnfollowedDependencies and a dependency-not-followed warning when no tsconfig paths define it. It was taken for
  an installed package, so the aliased file was missing from dependencies without a sign, and --related left the test
  out when that file changed.

• #69 41ce74d - Vitest 5 tests follow the option rules of Vitest 5.0.5: the last of .skip, skipIf() and runIf() on a test
  decides its skip, so test.skip.skipIf(false) is active, and a skipIf() or runIf() replaces the skip of a tag or of
  the options a describe passes down. An option set to undefined, as in test.skip('title', { skip: undefined }, fn),
  no longer undoes the chain or a tag.

• #69 41ce74d - In Vitest 5, the projects of a config that another config lists in projects are named inside that config's name, as
  Vitest names them: the unit project of apps/web/vitest.config.ts, named web, is web (unit).

• #69 41ce74d - A Vitest project without a test.name now has the name Vitest gives it, in projects and in configs: the name in
  the package.json of its folder, else the folder's name, so projects: ['packages/*'] names its projects
  @trading/pricing or ledger instead of "". An inline project without a name is named after its place among the
  inline projects, as 0, and in Vitest 5 no longer takes the name of the config it extends.

v0.6.0

Minor Changes

• #61 926352f - Each test now has a dependencyHash of everything it runs in the project, beside the unchanged bodyHash: its own call
  with its options, the hooks, test.use() and options of its file and of every describe around it, the fixtures it,
  those hooks, conditions and test.use() values ask for, with the fixtures those ask for, auto ones and overrides of
  Playwright's own, through every extend() and mergeTests(), the file's other top-level code, the config and the
  configs it builds on, setup files, global setup and teardown, the Playwright setup projects the test runs after, and the
  snapshot files and screenshots it compares against, and the functions, classes, constants and modules all of these use,
  followed by scope and through imports, re-exports, barrels, require(), import() and Vitest's vi.mock() with a
  fixed path, with the code each module runs when it loads. A helper that declares the test is in it with the arguments of
  the call that declares it, so two calls give two hashes. Formatting and comments don't change it, nor does moving the
  test. Installed packages are left out, and types too, except for type tests.

  dependencies lists the files that went into the hash. When the test reaches something the scan can't follow, a local
  file that is missing or doesn't parse, an import with a path worked out at run time or a pattern, a value only a
  declaration file describes, or a setup file the config names in a way the scan can't work out,
  hasUnfollowedDependencies is true and the new dependency-not-followed warning says where, once per place.

• #65 31a9bee - --related <file> and the related filter of filterInventory() keep the tests a change to a file can affect: those
  whose dependencies include it, and those that reach code the scan can't follow, which a change may affect unseen.
  --related - reads the files from standard input, one per line, so git diff --name-only origin/main... | npx test-inventory --framework playwright --related - lists the tests a pull request can affect. Paths are taken from the
  current folder, as git writes them, also when --root is a folder in the repository.

v0.5.0

Minor Changes

• #51 acec7e7 - Each file record has a new config key: the config the file uses, as the runner started at the root would pick it.
  That is the outermost config at or above the file that runs it, and for Vitest projects the project that does,
  read from Playwright's testDir, testMatch, testIgnore and .gitignore files, and from Vitest's include,
  exclude, includeSource, typecheck.include and benchmark.include with each version's defaults. Before, the
  nearest config was taken, which is wrong for a root config whose include covers packages that have their own
  vite.config.ts. A file with tests that no config runs gets the new file-not-run warning.

  The globals check follows the same choice, so a test in a folder project without a config of its own gets Vitest's
  defaults, and a test the root config runs takes the root's test.globals. Files named by benchmark.include are
  marked isBenchmark, in place of the default *.bench.* and *.benchmark.* names. The config reader now reads
  regular expressions and paths built from __dirname, __filename or import.meta with path.join(),
  path.resolve(), path.dirname(), fileURLToPath() and new URL().

• #50 3e55445 - The output lists the configs the scanned files use in a new top-level configs key, and as config lines in JSON
  Lines. A file uses the nearest playwright.config.*, vitest.config.* or vite.config.* in its folder or a folder
  above, up to the root, as the runner started in that folder would. A config is read as code, never run, and extends
  lists the files it takes settings from through imports, such as a shared base config.

  Vitest's globals now come from test.globals in the config a file uses, read through defineConfig(),
  mergeConfig(), spreads, constants and imported base configs. Before, any globals: true in a Vitest or Vite config
  in the file's folder or a folder above counted, also one Vitest doesn't read for the file, as in a vite.config.ts
  next to a vitest.config.ts. So globals-not-enabled can now appear where it didn't, and go away where a base config
  turns globals on. A test.globals the scan can't work out without running the config, and a config it can't read,
  get the new config-value-unknown warning.

• #55 a714612 - Run settings now come also from the config a file uses: Playwright's timeout, retries and repeatEach, and
  Vitest's testTimeout, retry and repeats, where the file sets none. New timeoutSource, retriesSource and repeatsSource
  keys say where each value comes from: test, suite, file or config, or null when nothing sets it. A test
  takes the settings of the projects that run it, after their grep. When those projects disagree, the config sets the
  value under a condition, as in process.env.CI ? 2 : 0, or the scan can't work out the projects, the value is null
  with its source config. For a value under a condition, timeoutConfigText, retriesConfigText and
  repeatsConfigText hold the config's text for it, so retries: process.env.CI ? 2 : 0 reads as written. Playwright's fullyParallel and Vitest's sequence.concurrent make tests
  isParallel where the file sets no mode, when every project that runs the test sets it.

  A .only in a Playwright file whose config sets forbidOnly gets the new forbidden-only error, since the run then
  fails before any test runs. When forbidOnly depends on a condition, as in !!process.env.CI, it is a warning that
  quotes the condition.

• #59 aeb3eb1 - Vitest tags the config defines in test.tags now apply to the tests that carry them, as Vitest applies them. Their
  skip, todo, fails and only are merged with the test's own and those a describe passes down, and Vitest's order,
  only over skip over todo, gives the state, with the new stateSource value config when a tag sets it. Their
  concurrent, timeout, retry, repeats and meta set those keys where the test sets none: from Vitest 5 over a
  describe's options, and in Vitest 4.1 under them. The lowest priority wins between tags, and a Vitest 5 project's
  own tags replace those of the config it extends.

  A tag the config doesn't define fails the file in Vitest unless strictTags is off, also when there is no config, so
  it now gets the new undefined-tag error, and the file's tests are notLoaded; in a function or under a condition it
  is a warning, as that code may not run. The config-may-change-state warning now shows only when
  the scan can't work out the config's tags.

• #54 c610789 - Each test record has a new projects key: the names of the projects that run the test, of the config its file uses.
  For Playwright, those whose testDir, testMatch and testIgnore include the file, less those whose grep or
  grepInvert leaves the test's title and tags out; for Vitest, those whose files include it, with each browser
  instance as a project, named as Vitest names it, such as unit (chromium). It is [] when no project runs the test,
  and null without a config or when the scan can't tell. Config records list their projects in projects, each with
  its name, testDir and Playwright's dependencies.

• #53 b9471ec - Glob patterns are optional, in scan() and on the command line: without them, the test files are those the
  Playwright or Vitest configs in the root and the folders below it run, each with the config it uses. A config whose
  files the scan can't work out gets a config-value-unknown warning that says to give patterns, and a scan without
  patterns and without a config says so in no-files-matched. An empty list of patterns is still an error.

Patch Changes

• #58 1a34d74 - oxc-parser is updated from 0.154.0 to 0.156.0. The scan output doesn't change.