A lightweight polyfill library for Promise-based WebExtension APIs in Chrome
Go to file
Rob Wu 67b3780d38 fix: Lazily initialize API via the original target (#71)
Originally, the polyfill created a Proxy with the original API object as
the target. This was changed to `Object.create(chrome)` because not
doing so would prevent the `browser.devtools` API from working because
the devtools API object is assigned as a read-only & non-configurable
property (#57).

However, that action itself caused a new bug: Whenever an API object
is dereferenced via the `browser` namespace, the original API is no
longer available in the `chrome` namespace, and trying to access the
API through `chrome` returns `undefined` plus the "Previous API
instantiation failed" warning (#58).
This is because Chrome lazily initializes fields in the `chrome`
API, but on the object from which the property is accessed, while
the polyfill accessed the property through an object with the
prototype set to `chrome` instead of directly via chrome.

To fix that, `Object.create(chrome)` was replaced with
`Object.assign({}, chrome)`. This fixes both of the previous issues
because
1) It is still a new object.
2) All lazily initialized fields are explicitly initialized.

This fix created a new issue: In Chrome some APIs cannot be used even
though they are visible in the API (e.g. `chrome.clipboard`), so
calling `Object.assign({}, chrome)` causes an error to be printed to
the console (#70).

To solve this, I use `Object.create(chrome)` again as a proxy target,
but dereference the API via the original target (`chrome`) to not
regress on #58.
Besides fixing the bug, this also reduces the performance impact
of the API because all API fields are lazily initialized again,
instead of upon start-up.

This fixes #70.
2018-03-12 19:23:28 +01:00
src fix: Lazily initialize API via the original target (#71) 2018-03-12 19:23:28 +01:00
test fix: Lazily initialize API via the original target (#71) 2018-03-12 19:23:28 +01:00
.eslintrc test: Run a minimal set of tests from a test extension running on Chrome (#66) 2017-10-23 19:42:01 +02:00
.gitignore test: introduced a test suite for unit testing. 2016-11-07 16:33:42 +01:00
.travis.yml fix: Pass --no-sandbox chrome CLI option when running the integration tests on travis (#85) 2018-01-09 17:49:56 +01:00
Gruntfile.js fix: Lock gruntify-eslint version to fix build errors (#42) 2017-08-20 13:26:43 +02:00
LICENSE Initial commit 2016-10-06 12:02:51 -07:00
README.md introducing table of contents 2017-10-01 11:47:50 +01:00
api-metadata.json feat: add identity.launchWebAuthFlow() in metadata. (#39) 2018-01-09 16:27:33 +01:00
package.json test: Run a minimal set of tests from a test extension running on Chrome (#66) 2017-10-23 19:42:01 +02:00

README.md

WebExtension browser API Polyfill Build Status

This library allows extensions written for the Promise-based WebExtension/BrowserExt API being standardized by the W3 Browser Extensions group to be used without modification in Google Chrome.

Table of contents

Building

To build, assuming you're already installed node >= 6 and npm, simply run:

npm install
npm run build
npm run test

This will install all the npm dependencies and build both non-minified and minified versions of the final library, and output them to dist/browser-polyfill.js and dist/browser-polyfill.min.js, respectively, and finally executes the unit tests on the generated dist files.

Basic Setup

In order to use the polyfill, it must be loaded into any context where browser APIs are accessed. The most common cases are background and content scripts, which can be specified in manifest.json:

{
  // ...

  "background": {
    "scripts": [
      "browser-polyfill.js",
      "background.js"
    ]
  },

  "content_scripts": [{
    // ...
    "js": [
      "browser-polyfill.js",
      "content.js"
    ]
  }]
}

For HTML documents, such as browserAction popups, or tab pages, it must be included more explicitly:

<!DOCTYPE html>
<html>
  <head>
    <script type="application/javascript" src="browser-polyfill.js"></script>
    <script type="application/javascript" src="popup.js"></script>
  </head>
  <!-- ... -->
</html>

And for dynamically-injected content scripts loaded by tabs.executeScript, it must be injected by a separate executeScript call, unless it has already been loaded via a content_scripts declaration in manifest.json:

browser.tabs.executeScript({file: "browser-polyfill.js"});
browser.tabs.executeScript({file: "content.js"}).then(result => {
  // ...
});

Using the Promise-based APIs

The Promise-based APIs in the browser namespace work, for the most part, very similarly to the callback-based APIs in Chrome's chrome namespace. The major differences are:

  • Rather than receiving a callback argument, every async function returns a Promise object, which resolves or rejects when the operation completes.

  • Rather than checking the chrome.runtime.lastError property from every callback, code which needs to explicitly deal with errors registers a separate Promise rejection handler.

  • Rather than receiving a sendResponse callback to send a response, onMessage listeners simply return a Promise whose resolution value is used as a reply.

  • Rather than nesting callbacks when a sequence of operations depend on each other, Promise chaining is generally used instead.

  • For users of an ES7 transpiler, such as Babel, the resulting Promises are generally used with async and await, rather than dealt with directly.

Examples

The following code will retrieve a list of URLs patterns from the storage API, retrieve a list of tabs which match any of them, reload each of those tabs, and notify the user that is has been done:

browser.storage.get("urls").then(({urls}) => {
  return browser.tabs.query({url: urls});
}).then(tabs => {
  return Promise.all(
    Array.from(tabs, tab => browser.tabs.reload(tab.id)));
  );
}).then(() => {
  return browser.notifications.create({
    type: "basic",
    iconUrl: "icon.png",
    title: "Tabs reloaded",
    message: "Your tabs have been reloaded",
  });
}).catch(error => {
  console.error(`An error occurred while reloading tabs: ${error.message}`);
});

Or, using an async function:

async function reloadTabs() {
  try {
    let {urls} = await browser.storage.get("urls");

    let tabs = await browser.tabs.query({url: urls});

    await Promise.all(
      Array.from(tabs, tab => browser.tabs.reload(tab.id)));
    );

    await browser.notifications.create({
      type: "basic",
      iconUrl: "icon.png",
      title: "Tabs reloaded",
      message: "Your tabs have been reloaded",
    });
  } catch (error) {
    console.error(`An error occurred while reloading tabs: ${error.message}`);
  }
}

It's also possible to use Promises effectively using two-way messaging. Communication between a background page and a tab content script, for example, looks something like this from the background page side:

browser.tabs.sendMessage("get-ids").then(results => {
  processResults(results);
});

And like this from the content script:

browser.runtime.onMessage.addListener(msg => {
  if (msg == "get-ids") {
    return browser.storage.get("idPattern").then(({idPattern}) => {
      return Array.from(document.querySelectorAll(idPattern),
                        elem => elem.textContent);
    });
  }
});

or:

browser.runtime.onMessage.addListener(async function(msg) {
  if (msg == "get-ids") {
    let {idPattern} = await browser.storage.get("idPattern");

    return Array.from(document.querySelectorAll(idPattern),
                      elem => elem.textContent);
  }
});

Or vice versa.