lib: refactor ES module loader for readability

PR-URL: https://github.com/nodejs/node/pull/16579
Reviewed-By: Michaël Zasso <targos@protonmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>

doc/api/errors.md

@@ -1207,6 +1207,14 @@ for strict compliance with the API specification (which in some cases may accept
 `func(undefined)` and `func()` are treated identically, and the
 [`ERR_INVALID_ARG_TYPE`][] error code may be used instead.
 
+<a id="ERR_MISSING_DYNAMIC_INSTANTIATE_HOOK"></a>
+### ERR_MISSING_DYNAMIC_INSTANTIATE_HOOK
+
+> Stability: 1 - Experimental
+
+Used when an [ES6 module][] loader hook specifies `format: 'dynamic` but does
+not provide a `dynamicInstantiate` hook.
+
 <a id="ERR_MISSING_MODULE"></a>
 ### ERR_MISSING_MODULE
 

lib/internal/errors.js

@@ -376,6 +376,9 @@ E('ERR_IPC_ONE_PIPE', 'Child process can have only one IPC pipe');
 E('ERR_IPC_SYNC_FORK', 'IPC cannot be used with synchronous forks');
 E('ERR_METHOD_NOT_IMPLEMENTED', 'The %s method is not implemented');
 E('ERR_MISSING_ARGS', missingArgs);
+E('ERR_MISSING_DYNAMIC_INSTANTIATE_HOOK',
+  'The ES Module loader may not return a format of \'dynamic\' when no ' +
+  'dynamicInstantiate function was provided');
 E('ERR_MISSING_MODULE', 'Cannot find module %s');
 E('ERR_MODULE_RESOLUTION_LEGACY', '%s not found by import in %s.' +
   ' Legacy behavior in require() would have found it at %s');

lib/internal/loader/Loader.js

@@ -10,7 +10,8 @@ const ModuleRequest = require('internal/loader/ModuleRequest');
 const errors = require('internal/errors');
 const debug = require('util').debuglog('esm');
 
-function getBase() {
+// Returns a file URL for the current working directory.
+function getURLStringForCwd() {
   try {
     return getURLFromFilePath(`${process.cwd()}/`).href;
   } catch (e) {
@@ -23,22 +24,44 @@ function getBase() {
   }
 }
 
+/* A Loader instance is used as the main entry point for loading ES modules.
+ * Currently, this is a singleton -- there is only one used for loading
+ * the main module and everything in its dependency graph. */
 class Loader {
-  constructor(base = getBase()) {
+  constructor(base = getURLStringForCwd()) {
-    this.moduleMap = new ModuleMap();
     if (typeof base !== 'string') {
       throw new errors.TypeError('ERR_INVALID_ARG_TYPE', 'base', 'string');
     }
+
+    this.moduleMap = new ModuleMap();
     this.base = base;
-    this.resolver = ModuleRequest.resolve.bind(null);
+    // The resolver has the signature
+    //   (specifier : string, parentURL : string, defaultResolve)
+    //       -> Promise<{ url : string,
+    //                    format: anything in Loader.validFormats }>
+    // where defaultResolve is ModuleRequest.resolve (having the same
+    // signature itself).
+    // If `.format` on the returned value is 'dynamic', .dynamicInstantiate
+    // will be used as described below.
+    this.resolver = ModuleRequest.resolve;
+    // This hook is only called when resolve(...).format is 'dynamic' and has
+    // the signature
+    //   (url : string) -> Promise<{ exports: { ... }, execute: function }>
+    // Where `exports` is an object whose property names define the exported
+    // names of the generated module. `execute` is a function that receives
+    // an object with the same keys as `exports`, whose values are get/set
+    // functions for the actual exported values.
     this.dynamicInstantiate = undefined;
   }
 
   hook({ resolve = ModuleRequest.resolve, dynamicInstantiate }) {
+    // Use .bind() to avoid giving access to the Loader instance when it is
+    // called as this.resolver(...);
     this.resolver = resolve.bind(null);
     this.dynamicInstantiate = dynamicInstantiate;
   }
 
+  // Typechecking wrapper around .resolver().
   async resolve(specifier, parentURL = this.base) {
     if (typeof parentURL !== 'string') {
       throw new errors.TypeError('ERR_INVALID_ARG_TYPE',
@@ -48,10 +71,11 @@ class Loader {
     const { url, format } = await this.resolver(specifier, parentURL,
                                                 ModuleRequest.resolve);
 
-    if (typeof format !== 'string') {
+    if (!Loader.validFormats.includes(format)) {
       throw new errors.TypeError('ERR_INVALID_ARG_TYPE', 'format',
-                                 ['esm', 'cjs', 'builtin', 'addon', 'json']);
+                                 Loader.validFormats);
     }
+
     if (typeof url !== 'string') {
       throw new errors.TypeError('ERR_INVALID_ARG_TYPE', 'url', 'string');
     }
@@ -72,14 +96,20 @@ class Loader {
     return { url, format };
   }
 
+  // May create a new ModuleJob instance if one did not already exist.
   async getModuleJob(specifier, parentURL = this.base) {
     const { url, format } = await this.resolve(specifier, parentURL);
     let job = this.moduleMap.get(url);
     if (job === undefined) {
       let loaderInstance;
       if (format === 'dynamic') {
+        const { dynamicInstantiate } = this;
+        if (typeof dynamicInstantiate !== 'function') {
+          throw new errors.Error('ERR_MISSING_DYNAMIC_INSTANTIATE_HOOK');
+        }
+
         loaderInstance = async (url) => {
-          const { exports, execute } = await this.dynamicInstantiate(url);
+          const { exports, execute } = await dynamicInstantiate(url);
           return createDynamicModule(exports, url, (reflect) => {
             debug(`Loading custom loader ${url}`);
             execute(reflect.exports);
@@ -100,5 +130,6 @@ class Loader {
     return module.namespace();
   }
 }
+Loader.validFormats = ['esm', 'cjs', 'builtin', 'addon', 'json', 'dynamic'];
 Object.setPrototypeOf(Loader.prototype, null);
 module.exports = Loader;

lib/internal/loader/ModuleJob.js

@@ -1,27 +1,35 @@
 'use strict';
 
+const { ModuleWrap } = internalBinding('module_wrap');
 const { SafeSet, SafePromise } = require('internal/safe_globals');
+const assert = require('assert');
 const resolvedPromise = SafePromise.resolve();
 
+const enableDebug = (process.env.NODE_DEBUG || '').match(/\besm\b/) ||
+    process.features.debug;
+
+/* A ModuleJob tracks the loading of a single Module, and the ModuleJobs of
+ * its dependencies, over time. */
 class ModuleJob {
-  /**
+  // `loader` is the Loader instance used for loading dependencies.
-   * @param {module: ModuleWrap?, compiled: Promise} moduleProvider
+  // `moduleProvider` is a function
-   */
   constructor(loader, url, moduleProvider) {
     this.loader = loader;
     this.error = null;
     this.hadError = false;
 
-    // linked == promise for dependency jobs, with module populated,
+    // This is a Promise<{ module, reflect }>, whose fields will be copied
-    // module wrapper linked
+    // onto `this` by `link()` below once it has been resolved.
-    this.moduleProvider = moduleProvider;
+    this.modulePromise = moduleProvider(url);
-    this.modulePromise = this.moduleProvider(url);
     this.module = undefined;
     this.reflect = undefined;
-    const linked = async () => {
+
+    // Wait for the ModuleWrap instance being linked with all dependencies.
+    const link = async () => {
       const dependencyJobs = [];
       ({ module: this.module,
          reflect: this.reflect } = await this.modulePromise);
+      assert(this.module instanceof ModuleWrap);
       this.module.link(async (dependencySpecifier) => {
         const dependencyJobPromise =
             this.loader.getModuleJob(dependencySpecifier, url);
@@ -29,63 +37,57 @@ class ModuleJob {
         const dependencyJob = await dependencyJobPromise;
         return (await dependencyJob.modulePromise).module;
       });
+      if (enableDebug) {
+        // Make sure all dependencies are entered into the list synchronously.
+        Object.freeze(dependencyJobs);
+      }
       return SafePromise.all(dependencyJobs);
     };
-    this.linked = linked();
+    // Promise for the list of all dependencyJobs.
+    this.linked = link();
 
     // instantiated == deep dependency jobs wrappers instantiated,
     // module wrapper instantiated
     this.instantiated = undefined;
   }
 
-  instantiate() {
+  async instantiate() {
     if (this.instantiated) {
       return this.instantiated;
     }
-    return this.instantiated = new Promise(async (resolve, reject) => {
+    return this.instantiated = this._instantiate();
-      const jobsInGraph = new SafeSet();
+  }
-      let jobsReadyToInstantiate = 0;
+
-      // (this must be sync for counter to work)
+  // This method instantiates the module associated with this job and its
-      const queueJob = (moduleJob) => {
+  // entire dependency graph, i.e. creates all the module namespaces and the
-        if (jobsInGraph.has(moduleJob)) {
+  // exported/imported variables.
-          return;
+  async _instantiate() {
-        }
+    const jobsInGraph = new SafeSet();
-        jobsInGraph.add(moduleJob);
+
-        moduleJob.linked.then((dependencyJobs) => {
+    const addJobsToDependencyGraph = async (moduleJob) => {
-          for (const dependencyJob of dependencyJobs) {
+      if (jobsInGraph.has(moduleJob)) {
-            queueJob(dependencyJob);
+        return;
-          }
+      }
-          checkComplete();
+      jobsInGraph.add(moduleJob);
-        }, (e) => {
+      const dependencyJobs = await moduleJob.linked;
-          if (!this.hadError) {
+      return Promise.all(dependencyJobs.map(addJobsToDependencyGraph));
-            this.error = e;
+    };
-            this.hadError = true;
+    try {
-          }
+      await addJobsToDependencyGraph(this);
-          checkComplete();
+    } catch (e) {
-        });
+      if (!this.hadError) {
-      };
+        this.error = e;
-      const checkComplete = () => {
+        this.hadError = true;
-        if (++jobsReadyToInstantiate === jobsInGraph.size) {
+      }
-          // I believe we only throw once the whole tree is finished loading?
+      throw e;
-          // or should the error bail early, leaving entire tree to still load?
+    }
-          if (this.hadError) {
+    this.module.instantiate();
-            reject(this.error);
+    for (const dependencyJob of jobsInGraph) {
-          } else {
+      // Calling `this.module.instantiate()` instantiates not only the
-            try {
+      // ModuleWrap in this module, but all modules in the graph.
-              this.module.instantiate();
+      dependencyJob.instantiated = resolvedPromise;
-              for (const dependencyJob of jobsInGraph) {
+    }
-                dependencyJob.instantiated = resolvedPromise;
+    return this.module;
-              }
-              resolve(this.module);
-            } catch (e) {
-              e.stack;
-              reject(e);
-            }
-          }
-        }
-      };
-      queueJob(this);
-    });
   }
 
   async run() {

lib/internal/loader/ModuleWrap.js

@@ -10,39 +10,49 @@ const createDynamicModule = (exports, url = '', evaluate) => {
     `creating ESM facade for ${url} with exports: ${ArrayJoin(exports, ', ')}`
   );
   const names = ArrayMap(exports, (name) => `${name}`);
-  // sanitized ESM for reflection purposes
+  // Create two modules: One whose exports are get- and set-able ('reflective'),
-  const src = `export let executor;
+  // and one which re-exports all of these but additionally may
-  ${ArrayJoin(ArrayMap(names, (name) => `export let $${name}`), ';\n')}
+  // run an executor function once everything is set up.
-  ;(() => [
+  const src = `
-    fn => executor = fn,
+  export let executor;
-    { exports: { ${
+  ${ArrayJoin(ArrayMap(names, (name) => `export let $${name};`), '\n')}
-  ArrayJoin(ArrayMap(names, (name) => `${name}: {
+  /* This function is implicitly returned as the module's completion value */
-        get: () => $${name},
+  (() => ({
-        set: v => $${name} = v
+    setExecutor: fn => executor = fn,
-      }`), ',\n')
+    reflect: {
-} } }
+      exports: { ${
-  ]);
+  ArrayJoin(ArrayMap(names, (name) => `
-  `;
+        ${name}: {
+          get: () => $${name},
+          set: v => $${name} = v
+        }`), ', \n')}
+      }
+    }
+  }));`;
   const reflectiveModule = new ModuleWrap(src, `cjs-facade:${url}`);
   reflectiveModule.instantiate();
-  const [setExecutor, reflect] = reflectiveModule.evaluate()();
+  const { setExecutor, reflect } = reflectiveModule.evaluate()();
   // public exposed ESM
-  const reexports = `import { executor,
+  const reexports = `
+  import {
+    executor,
     ${ArrayMap(names, (name) => `$${name}`)}
   } from "";
   export {
     ${ArrayJoin(ArrayMap(names, (name) => `$${name} as ${name}`), ', ')}
   }
-  // add await to this later if top level await comes along
+  if (typeof executor === "function") {
-  typeof executor === "function" ? executor() : void 0;`;
+    // add await to this later if top level await comes along
+    executor()
+  }`;
   if (typeof evaluate === 'function') {
     setExecutor(() => evaluate(reflect));
   }
-  const runner = new ModuleWrap(reexports, `${url}`);
+  const module = new ModuleWrap(reexports, `${url}`);
-  runner.link(async () => reflectiveModule);
+  module.link(async () => reflectiveModule);
-  runner.instantiate();
+  module.instantiate();
   return {
-    module: runner,
+    module,
     reflect
   };
 };