0
0
mirror of https://github.com/nodejs/node.git synced 2024-11-29 23:16:30 +01:00
nodejs/doc/api/esm.md
Daniel Nalborczyk b5e818197f doc: pass path in URL constructor
PR-URL: https://github.com/nodejs/node/pull/28161
Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
Reviewed-By: Tiancheng "Timothy" Gu <timothygu99@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Yongsheng Zhang <zyszys98@gmail.com>
2019-06-14 17:19:21 +08:00

26 KiB
Raw Blame History

ECMAScript Modules

Stability: 1 - Experimental

Introduction

ECMAScript modules are the official standard format to package JavaScript code for reuse. Modules are defined using a variety of import and export statements.

Node.js fully supports ECMAScript modules as they are currently specified and provides limited interoperability between them and the existing module format, CommonJS.

Node.js contains support for ES Modules based upon the Node.js EP for ES Modules and the ECMAScript-modules implementation.

Expect major changes in the implementation including interoperability support, specifier resolution, and default behavior.

Enabling

The --experimental-modules flag can be used to enable support for ECMAScript modules (ES modules).

Once enabled, Node.js will treat the following as ES modules when passed to node as the initial input, or when referenced by import statements within ES module code:

  • Files ending in .mjs.

  • Files ending in .js, or extensionless files, when the nearest parent package.json file contains a top-level field "type" with a value of "module".

  • Strings passed in as an argument to --eval or --print, or piped to node via STDIN, with the flag --input-type=module.

Node.js will treat as CommonJS all other forms of input, such as .js files where the nearest parent package.json file contains no top-level "type" field, or string input without the flag --input-type. This behavior is to preserve backward compatibility. However, now that Node.js supports both CommonJS and ES modules, it is best to be explicit whenever possible. Node.js will treat the following as CommonJS when passed to node as the initial input, or when referenced by import statements within ES module code:

  • Files ending in .cjs.

  • Files ending in .js, or extensionless files, when the nearest parent package.json file contains a top-level field "type" with a value of "commonjs".

  • Strings passed in as an argument to --eval or --print, or piped to node via STDIN, with the flag --input-type=commonjs.

package.json "type" field

Files ending with .js or .mjs, or lacking any extension, will be loaded as ES modules when the nearest parent package.json file contains a top-level field "type" with a value of "module".

The nearest parent package.json is defined as the first package.json found when searching in the current folder, that folders parent, and so on up until the root of the volume is reached.

// package.json
{
  "type": "module"
}
# In same folder as above package.json
node --experimental-modules my-app.js # Runs as ES module

If the nearest parent package.json lacks a "type" field, or contains "type": "commonjs", extensionless and .js files are treated as CommonJS. If the volume root is reached and no package.json is found, Node.js defers to the default, a package.json with no "type" field.

import statements of .js and extensionless files are treated as ES modules if the nearest parent package.json contains "type": "module".

// my-app.js, part of the same example as above
import './startup.js'; // Loaded as ES module because of package.json

Package Scope and File Extensions

A folder containing a package.json file, and all subfolders below that folder down until the next folder containing another package.json, is considered a package scope. The "type" field defines how .js and extensionless files should be treated within a particular package.json files package scope. Every package in a projects node_modules folder contains its own package.json file, so each projects dependencies have their own package scopes. A package.json lacking a "type" field is treated as if it contained "type": "commonjs".

The package scope applies not only to initial entry points (node --experimental-modules my-app.js) but also to files referenced by import statements and import() expressions.

// my-app.js, in an ES module package scope because there is a package.json
// file in the same folder with "type": "module".

import './startup/init.js';
// Loaded as ES module since ./startup contains no package.json file,
// and therefore inherits the ES module package scope from one level up.

import 'commonjs-package';
// Loaded as CommonJS since ./node_modules/commonjs-package/package.json
// lacks a "type" field or contains "type": "commonjs".

import './node_modules/commonjs-package/index.js';
// Loaded as CommonJS since ./node_modules/commonjs-package/package.json
// lacks a "type" field or contains "type": "commonjs".

Files ending with .mjs are always loaded as ES modules regardless of package scope.

Files ending with .cjs are always loaded as CommonJS regardless of package scope.

import './legacy-file.cjs';
// Loaded as CommonJS since .cjs is always loaded as CommonJS.

import 'commonjs-package/src/index.mjs';
// Loaded as ES module since .mjs is always loaded as ES module.

The .mjs and .cjs extensions may be used to mix types within the same package scope:

  • Within a "type": "module" package scope, Node.js can be instructed to interpret a particular file as CommonJS by naming it with a .cjs extension (since both .js and .mjs files are treated as ES modules within a "module" package scope).

  • Within a "type": "commonjs" package scope, Node.js can be instructed to interpret a particular file as an ES module by naming it with an .mjs extension (since both .js and .cjs files are treated as CommonJS within a "commonjs" package scope).

--input-type flag

Strings passed in as an argument to --eval or --print (or -e or -p), or piped to node via STDIN, will be treated as ES modules when the --input-type=module flag is set.

node --experimental-modules --input-type=module --eval \
  "import { sep } from 'path'; console.log(sep);"

echo "import { sep } from 'path'; console.log(sep);" | \
  node --experimental-modules --input-type=module

For completeness there is also --input-type=commonjs, for explicitly running string input as CommonJS. This is the default behavior if --input-type is unspecified.

Package Entry Points

The package.json "main" field defines the entry point for a package, whether the package is included into CommonJS via require or into an ES module via import.

// ./node_modules/es-module-package/package.json
{
  "type": "module",
  "main": "./src/index.js"
}
// ./my-app.mjs

import { something } from 'es-module-package';
// Loads from ./node_modules/es-module-package/src/index.js

An attempt to require the above es-module-package would attempt to load ./node_modules/es-module-package/src/index.js as CommonJS, which would throw an error as Node.js would not be able to parse the export statement in CommonJS.

As with import statements, for ES module usage the value of "main" must be a full path including extension: "./index.mjs", not "./index".

If the package.json "type" field is omitted, a .js file in "main" will be interpreted as CommonJS.

The "main" field can point to exactly one file, regardless of whether the package is referenced via require (in a CommonJS context) or import (in an ES module context). Package authors who want to publish a package to be used in both contexts can do so by setting "main" to point to the CommonJS entry point and informing the packages users of the path to the ES module entry point. Such a package would be accessible like require('pkg') and import 'pkg/module.mjs'. Alternatively the package "main" could point to the ES module entry point and legacy users could be informed of the CommonJS entry point path, e.g. require('pkg/commonjs').

import Specifiers

Terminology

The specifier of an import statement is the string after the from keyword, e.g. 'path' in import { sep } from 'path'. Specifiers are also used in export from statements, and as the argument to an import() expression.

There are four types of specifiers:

  • Bare specifiers like 'some-package'. They refer to an entry point of a package by the package name.

  • Deep import specifiers like 'some-package/lib/shuffle.mjs'. They refer to a path within a package prefixed by the package name.

  • Relative specifiers like './startup.js' or '../config.mjs'. They refer to a path relative to the location of the importing file.

  • Absolute specifiers like 'file:///opt/nodejs/config.js'. They refer directly and explicitly to a full path.

Bare specifiers, and the bare specifier portion of deep import specifiers, are strings; but everything else in a specifier is a URL.

Only file:// URLs are supported. A specifier like 'https://example.com/app.js' may be supported by browsers but it is not supported in Node.js.

Specifiers may not begin with / or //. These are reserved for potential future use. The root of the current volume may be referenced via file:///.

import.meta

  • {Object}

The import.meta metaproperty is an Object that contains the following property:

  • url {string} The absolute file: URL of the module.

Differences Between ES Modules and CommonJS

Mandatory file extensions

A file extension must be provided when using the import keyword. Directory indexes (e.g. './startup/index.js') must also be fully specified.

This behavior matches how import behaves in browser environments, assuming a typically configured server.

No NODE_PATH

NODE_PATH is not part of resolving import specifiers. Please use symlinks if this behavior is desired.

No require, exports, module.exports, __filename, __dirname

These CommonJS variables are not available in ES modules.

require can be imported into an ES module using module.createRequire().

An equivalent for __filename and __dirname is import.meta.url.

No require.extensions

require.extensions is not used by import. The expectation is that loader hooks can provide this workflow in the future.

No require.cache

require.cache is not used by import. It has a separate cache.

URL-based paths

ES modules are resolved and cached based upon URL semantics. This means that files containing special characters such as # and ? need to be escaped.

Modules will be loaded multiple times if the import specifier used to resolve them have a different query or fragment.

import './foo.mjs?query=1'; // loads ./foo.mjs with query of "?query=1"
import './foo.mjs?query=2'; // loads ./foo.mjs with query of "?query=2"

For now, only modules using the file: protocol can be loaded.

Interoperability with CommonJS

require

require always treats the files it references as CommonJS. This applies whether require is used the traditional way within a CommonJS environment, or in an ES module environment using module.createRequire().

To include an ES module into CommonJS, use import().

import statements

An import statement can reference either ES module or CommonJS JavaScript. Other file types such as JSON and Native modules are not supported. For those, use module.createRequire().

import statements are permitted only in ES modules. For similar functionality in CommonJS, see import().

The specifier of an import statement (the string after the from keyword) can either be an URL-style relative path like './file.mjs' or a package name like 'fs'.

Like in CommonJS, files within packages can be accessed by appending a path to the package name.

import { sin, cos } from 'geometry/trigonometry-functions.mjs';

Currently only the “default export” is supported for CommonJS files or packages:

import packageMain from 'commonjs-package'; // Works

import { method } from 'commonjs-package'; // Errors

There are ongoing efforts to make the latter code possible.

import() expressions

Dynamic import() is supported in both CommonJS and ES modules. It can be used to include ES module files from CommonJS code.

(async () => {
  await import('./my-app.mjs');
})();

CommonJS, JSON, and Native Modules

CommonJS, JSON, and Native modules can be used with module.createRequire().

// cjs.js
module.exports = 'cjs';

// esm.mjs
import { createRequire } from 'module';
import { fileURLToPath as fromURL } from 'url';

const require = createRequire(fromURL(import.meta.url));

const cjs = require('./cjs');
cjs === 'cjs'; // true

Builtin modules

Builtin modules will provide named exports of their public API, as well as a default export which can be used for, among other things, modifying the named exports. Named exports of builtin modules are updated when the corresponding exports property is accessed, redefined, or deleted.

import EventEmitter from 'events';
const e = new EventEmitter();
import { readFile } from 'fs';
readFile('./foo.txt', (err, source) => {
  if (err) {
    console.error(err);
  } else {
    console.log(source);
  }
});
import fs, { readFileSync } from 'fs';

fs.readFileSync = () => Buffer.from('Hello, ESM');

fs.readFileSync === readFileSync;

JSON Modules

JSON modules follow the WHATWG JSON modules specification.

The imported JSON only exposes a default. There is no support for named exports. A cache entry is created in the CommonJS cache, to avoid duplication. The same object will be returned in CommonJS if the JSON module has already been imported from the same path.

Assuming an index.mjs with

import packageConfig from './package.json';

Experimental Wasm Modules

Importing Web Assembly modules is supported under the --experimental-wasm-modules flag, allowing any .wasm files to be imported as normal modules while also supporting their module imports.

This integration is in line with the ES Module Integration Proposal for Web Assembly.

For example, an index.mjs containing:

import * as M from './module.wasm';
console.log(M);

executed under:

node --experimental-modules --experimental-wasm-modules index.mjs

would provide the exports interface for the instantiation of module.wasm.

Experimental Loader hooks

Note: This API is currently being redesigned and will still change.

To customize the default module resolution, loader hooks can optionally be provided via a --loader ./loader-name.mjs argument to Node.js.

When hooks are used they only apply to ES module loading and not to any CommonJS modules loaded.

Resolve hook

The resolve hook returns the resolved file URL and module format for a given module specifier and parent file URL:

const baseURL = new URL(`${process.cwd()}/`, 'file://');

export async function resolve(specifier,
                              parentModuleURL = baseURL,
                              defaultResolver) {
  return {
    url: new URL(specifier, parentModuleURL).href,
    format: 'module'
  };
}

The parentModuleURL is provided as undefined when performing main Node.js load itself.

The default Node.js ES module resolution function is provided as a third argument to the resolver for easy compatibility workflows.

In addition to returning the resolved file URL value, the resolve hook also returns a format property specifying the module format of the resolved module. This can be one of the following:

format Description
'builtin' Load a Node.js builtin module
'commonjs' Load a Node.js CommonJS module
'dynamic' Use a dynamic instantiate hook
'json' Load a JSON file
'module' Load a standard JavaScript module
'wasm' Load a WebAssembly module

For example, a dummy loader to load JavaScript restricted to browser resolution rules with only JS file extension and Node.js builtin modules support could be written:

import path from 'path';
import process from 'process';
import Module from 'module';

const builtins = Module.builtinModules;
const JS_EXTENSIONS = new Set(['.js', '.mjs']);

const baseURL = new URL(`${process.cwd()}/`, 'file://');

export function resolve(specifier, parentModuleURL = baseURL, defaultResolve) {
  if (builtins.includes(specifier)) {
    return {
      url: specifier,
      format: 'builtin'
    };
  }
  if (/^\.{0,2}[/]/.test(specifier) !== true && !specifier.startsWith('file:')) {
    // For node_modules support:
    // return defaultResolve(specifier, parentModuleURL);
    throw new Error(
      `imports must begin with '/', './', or '../'; '${specifier}' does not`);
  }
  const resolved = new URL(specifier, parentModuleURL);
  const ext = path.extname(resolved.pathname);
  if (!JS_EXTENSIONS.has(ext)) {
    throw new Error(
      `Cannot load file with non-JavaScript file extension ${ext}.`);
  }
  return {
    url: resolved.href,
    format: 'module'
  };
}

With this loader, running:

NODE_OPTIONS='--experimental-modules --loader ./custom-loader.mjs' node x.js

would load the module x.js as an ES module with relative resolution support (with node_modules loading skipped in this example).

Dynamic instantiate hook

To create a custom dynamic module that doesn't correspond to one of the existing format interpretations, the dynamicInstantiate hook can be used. This hook is called only for modules that return format: 'dynamic' from the resolve hook.

export async function dynamicInstantiate(url) {
  return {
    exports: ['customExportName'],
    execute: (exports) => {
      // Get and set functions provided for pre-allocated export names
      exports.customExportName.set('value');
    }
  };
}

With the list of module exports provided upfront, the execute function will then be called at the exact point of module evaluation order for that module in the import tree.

Resolution Algorithm

Features

The resolver has the following properties:

  • FileURL-based resolution as is used by ES modules
  • Support for builtin module loading
  • Relative and absolute URL resolution
  • No default extensions
  • No folder mains
  • Bare specifier package resolution lookup through node_modules

Resolver Algorithm

The algorithm to load an ES module specifier is given through the ESM_RESOLVE method below. It returns the resolved URL for a module specifier relative to a parentURL, in addition to the unique module format for that resolved URL given by the ESM_FORMAT routine.

The "module" format is returned for an ECMAScript Module, while the "commonjs" format is used to indicate loading through the legacy CommonJS loader. Additional formats such as "addon" can be extended in future updates.

In the following algorithms, all subroutine errors are propagated as errors of these top-level routines.

isMain is true when resolving the Node.js application entry point.

Resolver algorithm specification

ESM_RESOLVE(specifier, parentURL, isMain)

  1. Let resolvedURL be undefined.
  2. If specifier is a valid URL, then
    1. Set resolvedURL to the result of parsing and reserializing specifier as a URL.
  3. Otherwise, if specifier starts with "/", then
    1. Throw an Invalid Specifier error.
  4. Otherwise, if specifier starts with "./" or "../", then
    1. Set resolvedURL to the URL resolution of specifier relative to parentURL.
  5. Otherwise,
    1. Note: specifier is now a bare specifier.
    2. Set resolvedURL the result of PACKAGE_RESOLVE(specifier, parentURL).
  6. If the file at resolvedURL does not exist, then
    1. Throw a Module Not Found error.
  7. Set resolvedURL to the real path of resolvedURL.
  8. Let format be the result of ESM_FORMAT(resolvedURL, isMain).
  9. Load resolvedURL as module format, format.

PACKAGE_RESOLVE(packageSpecifier, parentURL)

  1. Let packageName be undefined.
  2. Let packageSubpath be undefined.
  3. If packageSpecifier is an empty string, then
    1. Throw an Invalid Specifier error.
  4. If packageSpecifier does not start with "@", then
    1. Set packageName to the substring of packageSpecifier until the first "/" separator or the end of the string.
  5. Otherwise,
    1. If packageSpecifier does not contain a "/" separator, then
      1. Throw an Invalid Specifier error.
    2. Set packageName to the substring of packageSpecifier until the second "/" separator or the end of the string.
  6. Let packageSubpath be the substring of packageSpecifier from the position at the length of packageName plus one, if any.
  7. Assert: packageName is a valid package name or scoped package name.
  8. Assert: packageSubpath is either empty, or a path without a leading separator.
  9. If packageSubpath contains any "." or ".." segments or percent encoded strings for "/" or "\" then,
    1. Throw an Invalid Specifier error.
  10. If packageSubpath is empty and packageName is a Node.js builtin module, then
    1. Return the string "node:" concatenated with packageSpecifier.
  11. While parentURL is not the file system root,
    1. Let packageURL be the URL resolution of "node_modules/" concatenated with packageSpecifier, relative to parentURL.
    2. Set parentURL to the parent folder URL of parentURL.
    3. If the folder at packageURL does not exist, then
      1. Set parentURL to the parent URL path of parentURL.
      2. Continue the next loop iteration.
    4. Let pjson be the result of READ_PACKAGE_JSON(packageURL).
    5. If packageSubpath is empty, then
      1. Return the result of PACKAGE_MAIN_RESOLVE(packageURL, pjson).
    6. Otherwise,
      1. Return the URL resolution of packageSubpath in packageURL.
  12. Throw a Module Not Found error.

PACKAGE_MAIN_RESOLVE(packageURL, pjson)

  1. If pjson is null, then
    1. Throw a Module Not Found error.
  2. If pjson.main is a String, then
    1. Let resolvedMain be the concatenation of packageURL, "/", and pjson.main.
    2. If the file at resolvedMain exists, then
      1. Return resolvedMain.
  3. If pjson.type is equal to "module", then
    1. Throw a Module Not Found error.
  4. Let legacyMainURL be the result applying the legacy LOAD_AS_DIRECTORY CommonJS resolver to packageURL, throwing a Module Not Found error for no resolution.
  5. If legacyMainURL does not end in ".js" then,
    1. Throw an Unsupported File Extension error.
  6. Return legacyMainURL.

ESM_FORMAT(url, isMain)

  1. Assert: url corresponds to an existing file.
  2. Let pjson be the result of READ_PACKAGE_SCOPE(url).
  3. If url ends in ".mjs", then
    1. Return "module".
  4. If url ends in ".cjs", then
    1. Return "commonjs".
  5. If pjson?.type exists and is "module", then
    1. If isMain is true or url ends in ".js", then
      1. Return "module".
    2. Throw an Unsupported File Extension error.
  6. Otherwise,
    1. If isMain is true or url ends in ".js", ".json" or ".node", then
      1. Return "commonjs".
    2. Throw an Unsupported File Extension error.

READ_PACKAGE_SCOPE(url)

  1. Let scopeURL be url.
  2. While scopeURL is not the file system root,
    1. Let pjson be the result of READ_PACKAGE_JSON(scopeURL).
    2. If pjson is not null, then
      1. Return pjson.
    3. Set scopeURL to the parent URL of scopeURL.
  3. Return null.

READ_PACKAGE_JSON(packageURL)

  1. Let pjsonURL be the resolution of "package.json" within packageURL.
  2. If the file at pjsonURL does not exist, then
    1. Return null.
  3. If the file at packageURL does not parse as valid JSON, then
    1. Throw an Invalid Package Configuration error.
  4. Return the parsed JSON source of the file at pjsonURL.

Customizing ESM specifier resolution algorithm

The current specifier resolution does not support all default behavior of the CommonJS loader. One of the behavior differences is automatic resolution of file extensions and the ability to import directories that have an index file.

The --es-module-specifier-resolution=[mode] flag can be used to customize the extension resolution algorithm. The default mode is explicit, which requires the full path to a module be provided to the loader. To enable the automatic extension resolution and importing from directories that include an index file use the node mode.

$ node --experimental-modules index.mjs
success!
$ node --experimental-modules index #Failure!
Error: Cannot find module
$ node --experimental-modules --es-module-specifier-resolution=node index
success!