A module namespace object is an object that describes all exports from a module. It is a static object that is created when the module is evaluated. There are two ways to access the module namespace object of a module: through a namespace import (import * as name from moduleName), or through the fulfillment value of a dynamic import.
The module namespace object is a sealed object with null prototype. This means all string keys of the object correspond to the exports of the module and there are never extra keys. All keys are enumerable in lexicographic order (i.e., the default behavior of Array.prototype.sort()), with the default export available as a key called default. In addition, the module namespace object has a [Symbol.toStringTag] property with the value "Module", used in Object.prototype.toString().
The string properties are non-configurable and writable when you use Object.getOwnPropertyDescriptors() to get their descriptors. However, they are effectively read-only, because you cannot re-assign a property to a new value. This behavior mirrors the fact that static imports create "live bindings" — the values can be re-assigned by the module exporting them, but not by the module importing them. The writability of the properties reflects the possibility of the values changing, because non-configurable and non-writable properties must be constant. For example, you can re-assign the exported value of a variable, and the new value can be observed in the module namespace object.
Each (normalized) module specifier corresponds to a unique module namespace object, so the following is generally true:
import * as mod from "/my-module.js";
import("/my-module.js").then((mod2) => {
console.log(mod === mod2); // true
});
Except in one curious case: because a promise never fulfills to a thenable, if the my-module.js module exports a function called then(), that function will automatically get called when the dynamic import's promise is fulfilled, as part of the promise resolution process.
// my-module.js
export function then(resolve) {
console.log("then() called");
resolve(1);
}
// main.js
import * as mod from "/my-module.js";
import("/my-module.js").then((mod2) => {
// Logs "then() called"
console.log(mod === mod2); // false
});
Warning: Do not export a function called then() from a module. This will cause the module to behave differently when imported dynamically than when imported statically.
This aggressive caching ensures that a piece of JavaScript code is never executed more than once, even if it is imported multiple times. Future imports don't even result in HTTP requests or disk access. If you do need to re-import and re-evaluate a module without restarting the entire JavaScript environment, one possible trick is to use a unique query parameter in the module specifier. This works in non-browser runtimes that support URL specifiers too.
import(`/my-module.js?t=${Date.now()}`);
Note that this can lead to memory leaks in a long-running application, because the engine cannot safely garbage-collect any module namespace objects. Currently, there is no way to manually clear the cache of module namespace objects.
Module namespace object caching only applies to modules that are loaded and linked successfully. A module is imported in three steps: loading (fetching the module), linking (mostly, parsing the module), and evaluating (executing the parsed code). Only evaluation failures are cached; if a module fails to load or link, the next import may try to load and link the module again. The browser may or may not cache the result of the fetch operation, but it should follow typical HTTP semantics, so handling such network failures should not be different from handling fetch() failures.