module: add preImport loader hook

Author: guybedford

doc/api/esm.md

@@ -726,6 +726,51 @@ A hook that returns without calling `next<hookName>()` _and_ without returning
 `shortCircuit: true` also triggers an exception. These errors are to help
 prevent unintentional breaks in the chain.
 
+#### `preImport(specifier, context)`
+
+<!-- YAML
+changes:
+  - version: REPLACEME
+    pr-url: REPLACEME
+    description: Add support for preImport hook
+-->
+
+> The loaders API is being redesigned. This hook may disappear or its
+> signature may change. Do not rely on the API described below.
+
+* `specifier` {string}
+* `context` {Object}
+  * `conditions` {string\[]} Resolution conditions of the current environment,
+    as defined for the `package.json` imports and exports fields
+  * `dynamic` {boolean} Whether this import is a dynamic `import()`
+  * `importAssertions` {Object}
+  * `parentURL` {string|undefined} The module importing this one, or undefined
+    if this is the Node.js entry point
+
+The `preImport` hook allows for tracking and asynchronous setup work for every
+top-level import operation.
+
+The `preImport` hook is called for each top-level import operation by the
+module loader, both for the host-called imports (ie for the main entry) and for
+dynamic `import()` imports. These are distinguished by the `dynamic` context.
+
+All `preImport` hooks for all loaders are run asynchronously in parallel,
+and block any further load operations (ie resolve and load) for the module graph
+being imported until they all complete successfully.
+
+Multiple import calls to the same import specifier will re-call the hook
+multiple times. The first error thrown by the `preImport` hooks will be directly
+returned to the specific import operation as the load failure.
+
+```js
+export async function preImport (specifier, context) {
+  if (context.topLevel)
+    console.log(`Top-level load of ${specifier}`);
+  else
+    console.log(`Dynamic import of ${specifier} in ${context.parentURL}`);
+}
+```
+
 #### `resolve(specifier, context, nextResolve)`
 
 <!-- YAML
@@ -747,7 +792,8 @@ changes:
 
 * `specifier` {string}
 * `context` {Object}
-  * `conditions` {string\[]} Export conditions of the relevant `package.json`
+  * `conditions` {string\[]} Resolution conditions of the current environment,
+    as defined for the `package.json` imports and exports fields
   * `importAssertions` {Object}
   * `parentURL` {string|undefined} The module importing this one, or undefined
     if this is the Node.js entry point
@@ -837,7 +883,8 @@ changes:
 
 * `url` {string} The URL returned by the `resolve` chain
 * `context` {Object}
-  * `conditions` {string\[]} Export conditions of the relevant `package.json`
+  * `conditions` {string\[]} Resolution conditions of the current environment,
+    as defined for the `package.json` imports and exports fields
   * `format` {string|null|undefined} The format optionally supplied by the
     `resolve` hook chain
   * `importAssertions` {Object}

lib/internal/modules/cjs/loader.js

@@ -1025,7 +1025,7 @@ function wrapSafe(filename, content, cjsModuleInstance) {
         const loader = asyncESM.esmLoader;
         return loader.import(specifier,
                              loader.getBaseURL(normalizeReferrerURL(filename)),
-                             importAssertions);
+                             importAssertions, true);
       },
     });
   }
@@ -1042,7 +1042,7 @@ function wrapSafe(filename, content, cjsModuleInstance) {
         const loader = asyncESM.esmLoader;
         return loader.import(specifier,
                              loader.getBaseURL(normalizeReferrerURL(filename)),
-                             importAssertions);
+                             importAssertions, true);
       },
     });
   } catch (err) {

lib/internal/modules/esm/loader.js

@@ -4,18 +4,15 @@
 require('internal/modules/cjs/loader');
 
 const {
-  Array,
-  ArrayIsArray,
   ArrayPrototypeJoin,
   ArrayPrototypePush,
   FunctionPrototypeBind,
   FunctionPrototypeCall,
   ObjectAssign,
   ObjectCreate,
+  ObjectFreeze,
   ObjectSetPrototypeOf,
-  PromiseAll,
   RegExpPrototypeExec,
-  SafeArrayIterator,
   SafeWeakMap,
   StringPrototypeStartsWith,
   globalThis,
@@ -59,7 +56,7 @@ const {
   fetchModule,
   inFetchCache,
 } = require('internal/modules/esm/fetch_module');
-
+const { strictEqual } = require('internal/assert');
 
 /**
  * @typedef {object} ExportedHooks
@@ -123,6 +120,8 @@ class ESMLoader {
     },
   ];
 
+  #preImporters = [];
+
   /**
    * Phase 1 of 2 in ESM loading.
    * @private
@@ -184,6 +183,7 @@ class ESMLoader {
    */
   static pluckHooks({
     globalPreload,
+    preImport,
     resolve,
     load,
     // obsolete hooks:
@@ -232,6 +232,9 @@ class ESMLoader {
       acceptedHooks.globalPreloader =
         FunctionPrototypeBind(globalPreload, null);
     }
+    if (preImport) {
+      acceptedHooks.preImporter = FunctionPrototypeBind(preImport, null);
+    }
     if (resolve) {
       acceptedHooks.resolver = FunctionPrototypeBind(resolve, null);
     }
@@ -259,6 +262,7 @@ class ESMLoader {
       } = customLoaders[i];
       const {
         globalPreloader,
+        preImporter,
         resolver,
         loader,
       } = ESMLoader.pluckHooks(exports);
@@ -272,6 +276,12 @@ class ESMLoader {
           },
         );
       }
+      if (preImporter) {
+        ArrayPrototypePush(
+          this.#preImporters,
+          preImporter
+        );
+      }
       if (resolver) {
         ArrayPrototypePush(
           this.#resolvers,
@@ -308,7 +318,7 @@ class ESMLoader {
         importModuleDynamically: (specifier, { url }, importAssertions) => {
           return this.import(specifier,
                              this.getBaseURL(url),
-                             importAssertions);
+                             importAssertions, true);
         }
       });
 
@@ -472,48 +482,47 @@ class ESMLoader {
    * This method must NOT be renamed: it functions as a dynamic import on a
    * loader module.
    *
-   * @param {string | string[]} specifiers Path(s) to the module.
-   * @param {string} parentURL Path of the parent importing the module.
+   * @param {string} specifiers Imported specifier
+   * @param {string} parentURL URL of the parent importing the module.
    * @param {Record<string, string>} importAssertions Validations for the
    *                                                  module import.
-   * @returns {Promise<ExportedHooks | KeyedExports[]>}
-   *  A collection of module export(s) or a list of collections of module
-   *  export(s).
+   * @param {boolean} dynamic Whether the import is a dynamic `import()`.
+   * @returns {Promise<ModuleNamespace>}
    */
-  async import(specifiers, parentURL, importAssertions) {
-    // For loaders, `import` is passed multiple things to process, it returns a
-    // list pairing the url and exports collected. This is especially useful for
-    // error messaging, to identity from where an export came. But, in most
-    // cases, only a single url is being "imported" (ex `import()`), so there is
-    // only 1 possible url from which the exports were collected and it is
-    // already known to the caller. Nesting that in a list would only ever
-    // create redundant work for the caller, so it is later popped off the
-    // internal list.
-    const wasArr = ArrayIsArray(specifiers);
-    if (!wasArr) { specifiers = [specifiers]; }
-
-    const count = specifiers.length;
-    const jobs = new Array(count);
-
-    for (let i = 0; i < count; i++) {
-      jobs[i] = this.getModuleJob(specifiers[i], parentURL, importAssertions)
-        .then((job) => job.run())
-        .then(({ module }) => module.getNamespace());
-    }
-
-    const namespaces = await PromiseAll(new SafeArrayIterator(jobs));
-
-    if (!wasArr) { return namespaces[0]; } // We can skip the pairing below
-
-    for (let i = 0; i < count; i++) {
-      const namespace = ObjectCreate(null);
-      namespace.url = specifiers[i];
-      namespace.exports = namespaces[i];
-
-      namespaces[i] = namespace;
-    }
+  async import(specifier, parentURL, importAssertions = ObjectCreate(null), dynamic = false) {
+    await this.preImport(specifier, parentURL, importAssertionsForResolve, dynamic);
+    const job = await this.getModuleJob(specifier, parentURL, importAssertions);
+    this.getModuleJob(specifier, parentURL, importAssertions);
+    const { module } = await job.run();
+    return module.getNamespace();
+  }
 
-    return namespaces;
+  /**
+   * Run the prepare hooks for a new import operation.
+   *
+   * Internally, this behaves like a backwards iterator, wherein the stack of
+   * hooks starts at the top and each call to `nextResolve()` moves down 1 step
+   * until it reaches the bottom or short-circuits.
+   *
+   * @param {string} specifier The import specifier.
+   * @param {string} parentURL The URL of the module's parent.
+   * @param {ImportAssertions} [importAssertions] Assertions from the import
+   *                                              statement or expression.
+   * @param {boolean} dynamic Whether the import is a dynamic `import()`.
+   */
+   async preImport(
+    specifier,
+    parentURL,
+    importAssertions,
+    dynamic
+  ) {
+    const context = ObjectFreeze({
+      conditions: DEFAULT_CONDITIONS,
+      dynamic,
+      importAssertions,
+      parentURL
+    });
+    await Promise.all(this.#preImporters.map(preImport => preImport(specifier, context)));
   }
 
   /**

lib/internal/modules/esm/translators.js

@@ -105,7 +105,7 @@ function errPath(url) {
 async function importModuleDynamically(specifier, { url }, assertions) {
   return asyncESM.esmLoader.import(specifier,
                                    asyncESM.esmLoader.getBaseURL(url),
-                                   assertions);
+                                   assertions, true);
 }
 
 // Strategy for loading a standard JavaScript module.

lib/internal/process/esm_loader.js

@@ -66,11 +66,14 @@ async function initializeLoader() {
   const internalEsmLoader = new ESMLoader();
 
   // Importation must be handled by internal loader to avoid poluting userland
-  const keyedExportsList = await internalEsmLoader.import(
-    customLoaders,
-    pathToFileURL(cwd).href,
-    ObjectCreate(null),
-  );
+  const parentURL = pathToFileURL(cwd).href;
+  const importAssertions = ObjectCreate(null);
+
+  const keyedExportsList = await Promise.all(customLoaders.map(url => {
+    const exports = await internalEsmLoader.import(url, parentURL,
+                                                   importAssertions);
+    return { exports, url };
+  }));
 
   // Hooks must then be added to external/public loader
   // (so they're triggered in userland)

lib/internal/process/execution.js

@@ -79,7 +79,7 @@ function evalScript(name, body, breakFirstLine, print) {
       [kVmBreakFirstLineSymbol]: !!breakFirstLine,
       importModuleDynamically(specifier, _, importAssertions) {
         const loader = asyncESM.esmLoader;
-        return loader.import(specifier, baseUrl, importAssertions);
+        return loader.import(specifier, baseUrl, importAssertions, true);
       }
     }));
   if (print) {

lib/repl.js

@@ -462,7 +462,7 @@ function REPLServer(prompt,
               displayErrors: true,
               importModuleDynamically: (specifier, _, importAssertions) => {
                 return asyncESM.esmLoader.import(specifier, parentURL,
-                                                 importAssertions);
+                                                 importAssertions, true);
               }
             });
           } catch (fallbackError) {
@@ -506,7 +506,7 @@ function REPLServer(prompt,
             displayErrors: true,
             importModuleDynamically: (specifier, _, importAssertions) => {
               return asyncESM.esmLoader.import(specifier, parentURL,
-                                               importAssertions);
+                                               importAssertions, true);
             }
           });
         } catch (e) {