Raygun.com provider for JavaScript
JavaScript CSS Other
Latest commit 10920e6 Feb 14, 2017 @Filip-Dimitrievski Filip-Dimitrievski committed on GitHub Merge pull request #213 from MindscapeHQ/offline-errors-send-fix
Calling makePostCorsRequest so to not over JSON stringify the POST data

README.md

Raygun4js

Raygun.io provider for client-side JavaScript

Build Status

Getting Started

Step 1

Add this snippet to your markup, before the closing </head> tag:

<script type="text/javascript">
    !function(a,b,c,d,e,f,g,h){a.RaygunObject=e,a[e]=a[e]||function(){
    (a[e].o=a[e].o||[]).push(arguments)},f=b.createElement(c),g=b.getElementsByTagName(c)[0],
    f.async=1,f.src=d,g.parentNode.insertBefore(f,g),h=a.onerror,a.onerror=function(b,c,d,f,g){
    h&&h(b,c,d,f,g),g||(g=new Error(b)),a[e].q=a[e].q||[],a[e].q.push({
    e:g})}}(window,document,"script","//cdn.raygun.io/raygun4js/raygun.min.js","rg4js");
</script>

This will fetch the raygun4js script from our CDN asynchronously, so it doesn't block other scripts from being loaded. It will also catch errors that are thrown while the page is loading, and send them when the script is ready.

Step 2

Add the following lines to your JS site code and paste in your API key (from your Raygun dashboard), to set up the provider to automatically send errors to your Raygun:

<script type="text/javascript">
    rg4js('apiKey', 'paste_your_api_key_here');
    rg4js('enableCrashReporting', true);
</script>

This will configure the provider to send to your Raygun app, and to automatically send all unhandled errors.

That's it for the basic setup! See Usage below for more info on how to send errors.

Alternative setup options

Note: This library can now be interacted with in two ways, the V1 API and the V2 API. The V1 API is available as 'public' functions on the global Raygun object, and is intended to be used to control the provider during runtime. Legacy setup methods remain on this API for backwards compatibility with 1.x releases. The V2 API is made available when using the snippet (above), and is used to asynchronously configure the provider during onLoad. This is the recommended approach for new setups.

If you are installing the provider locally using a package manager or manually, you can either use the V2 API by adding the snippet and replace the second-last parameter with the URL of your hosted version of the script, or use the V1 API. The snippet/V2 approach does not support the script being bundled with other vendor scripts, but the V1 API does.

Snippet without page load error handler

If you do not want errors to be caught while the page is loading, use this snippet here.

Synchronous methods

Note that using these methods will not catch errors thrown while the page is loading. The script needs to be referenced before your other site/app scripts, and will block the page load while it is being downloaded/parsed/executed.

This will also disrupt Pulse timings, making them erroneous. For Pulse, it is especially importing that the async snippet method above is used, instead of one of the following.

With Bower

Run bower install raygun4js

With NPM

npm install raygun4js --save

This lets you require the library with tools such as Webpack or Browserify.

From NuGet

Visual Studio users can get it by opening the Package Manager Console and typing Install-Package raygun4js

React Native/as a UMD module

React Native and other bundled app frameworks that uses packaging/module loading libraries can use Raygun4JS as a UMD module:

// Install the library

npm install raygun4js --save

// In a central module, reference and install the library

var rg4js = require('raygun4js');

rg4js('enableCrashReporting', true);
rg4js('apiKey', 'paste_your_api_key_here');

All unhandled errors will then be sent to Raygun. You can also require('raygun4js') in any other modules and use the rest of the V2 API below - including rg4js('send', error) for manual error sending.

If you use this approach, we appreciate your feedback as this is a new feature for the library.

Manual download

Download the production version or the development version. You can also download a version without the jQuery hooks if you are not using jQuery or you wish to provide your own hooks. Get this as a production version or development version.

Usage

To send errors manually:

try {
  throw new Error('oops');
}
catch(e) {
  rg4js('send', e);
}

In order to get stack traces, you need to wrap your code in a try/catch block like above. Otherwise the error hits window.onerror handler and may only contain the error message, line number, and column number.

You also need to throw errors with throw new Error('foo') instead of throw 'foo'.

To automatically catch and send unhandled errors, you can attach the automatic window.onerror handler callback:

rg4js('enableCrashReporting', true);

If you need to detach it (this will disable automatic unhandled error sending):

rg4js('detach');

IE8

If you are serving your site over HTTP and want IE8 to be able to submit JavaScript errors then you will need to set the following setting which will allow IE8 to submit the error over HTTP. Otherwise the provider will only submit over HTTPS which IE8 will not allow while being served over HTTP.

rg4js('options', {
  allowInsecureSubmissions: true
});

Documentation

Legacy V1 documentation

The old documentation for the V1 API (Raygun.send() etc) is available here.

Initialization Options

To configure the provider, call this and pass in an options object:

rg4js('options', {
  // Add some or all of the options below
});

The second parameter should contain one or more of these keys and a value to customize the behavior:

allowInsecureSubmissions - posts error payloads over HTTP. This allows IE8 to send JS errors

ignoreAjaxAbort - User-aborted Ajax calls result in errors - if this option is true, these will not be sent.

ignoreAjaxError - Ajax requests that return error codes will not be sent as errors to Raygun if this options is true.

debugMode - Raygun4JS will log to the console when sending errors.

wrapAsynchronousCallbacks - if set to false, async callback functions triggered by setTimeout/setInterval will not be wrapped when attach() is called. Defaults to true

ignore3rdPartyErrors - ignores any errors that have no stack trace information. This will discard any errors that occur completely within 3rd party scripts - if code loaded from the current domain called the 3rd party function, it will have at least one stack line and will still be sent.

excludedHostnames - Prevents errors from being sent from certain hostnames (domains) by providing an array of strings or RegExp objects (for partial matches). Each should match the hostname or TLD that you want to exclude. Note that protocols are not tested.

excludedUserAgents - Prevents errors from being sent from certain user agents by providing an array of strings. This is very helpful to exclude errors reported by certain browsers or test automation with CasperJS, PhantomJS or any other testing utility that sends a custom user agent. If a part of the client's navigator.userAgent matches one of the given strings in the array, then the client will be excluded from error reporting.

disableErrorTracking - Prevent uncaught errors from being sent.

disablePulse - Prevent Pulse real user monitoring events from being sent.

apiEndpoint - A string URI containing the protocol, domain and port (optional) where all payloads will be sent to. This can be used to proxy payloads to the Raygun API through your own server. When not set this defaults internally to the Raygun API, and for most usages you won't need to set this.

pulseMaxVirtualPageDuration - The maximum time a virtual page can be considered viewed, in milliseconds (defaults to 30 minutes).

pulseIgnoreUrlCasing - Ignore URL casing when sending data to Pulse.

An example:

rg4js('options', {
  allowInsecureSubmissions: true,
  ignoreAjaxAbort: true,
  ignoreAjaxError: true,
  debugMode: true,
  ignore3rdPartyErrors: false,
  wrapAsynchronousCallbacks: true,
  excludedHostnames: ['\.local'],
  excludedUserAgents: ['Mosaic'],
  disableErrorTracking: false,
  disablePulse: false,
  pulseMaxVirtualPageDuration: 1800000,
  pulseIgnoreUrlCasing: false
});

Pulse API

Tracking Single Page Application (SPA) events

Raygun Pulse supports client-side SPAs through the trackEvent function:

rg4js('trackEvent', {
    type: 'pageView',
    path: '/' + window.location.pathname // Or perhaps window.location.hash
});

When a route or view change is triggered in your SPA, this function should be called with type being pageView and path set to a string representing the new view or route. Pulse will collect up all timing information that is available and send it to the dashboard. These are then viewable as 'virtual pages'.

The following are a couple of configuration examples that you can use or adapt for your client-side view library/framework. Naturally, if you are using a more full-featured routing system, you should trigger a pageView inside there when the route changes.

jQuery

$(window).hashchange(function() {
  rg4js('trackEvent', {
      type: 'pageView',
      path: '/' + location.hash
  });
});

AngularJS

$scope.$on('$routeChangeSuccess', function () {
  rg4js('trackEvent', {
      type: 'pageView',
      path: '/' + $scope.area
  });
});

Multiple Raygun objects on a single page

You can have multiple Raygun objects in global scope. This lets you set them up with different API keys for instance, and allow you to send different errors to more than one application in the Raygun web app.

To create a new Raygun object and use it call:

var secondRaygun = rg4js('getRaygunInstance').constructNewRaygun();
secondRaygun.init('apikey');
secondRaygun.send(...);

Only one Raygun object can be attached as the window.onerror handler at one time, as onerror can only be bound to one function at once. Whichever Raygun object had attach() called on it last will handle the unhandle errors for the page. Note that you should use the V1 API to send using the second Raygun object, and it should be created and called once the page is loaded (for instance in an onload callback).

NoConflict mode

If you already have an variable called Raygun attached to window, you can prevent the provider from overwriting this by enabling NoConflict mode:

rg4js('noConflict', true);

To then get an instance of the Raygun object when using V2, call this once the page is loaded:

var raygun = rg4js('getRaygunInstance');

Callback Events

onBeforeSend

rg4js('onBeforeSend', function (payload) {
  return payload;
});

Call this function and pass in a function which takes one parameter (see the example below). This callback function will be called immediately before the payload is sent. The one parameter it gets will be the payload that is about to be sent. Thus from your function you can inspect the payload and decide whether or not to send it.

From the supplied function, you should return either the payload (intact or mutated as per your needs), or false.

If your function returns a truthy object, Raygun4JS will attempt to send it as supplied. Thus, you can mutate it as per your needs - preferably only the values if you wish to filter out data that is not taken care of by filterSensitiveData(). You can also of course return it as supplied.

If, after inspecting the payload, you wish to discard it and abort the send to Raygun, simply return false.

By example:

var myBeforeSend = function (payload) {
  console.log(payload); // Modify the payload here if necessary
  return payload; // Return false here to abort the send
}

rg4js('onBeforeSend', myBeforeSend);

onAfterSend

rg4js('onAfterSend', function (xhrResponse) {
  // Inspect the XHR response here
});

Call this function and pass in a function which takes one parameter (see the example below). This callback function will be immediately called after the XHR request for a Crash Reporting or Pulse event responds successfully, or errors out (its onerror was called). You can inspect the one parameter, which is the XHR object containing the HTTP response data.

onBeforeXHR

rg4js('onBeforeXHR', function (xhr) {
  // Mutate the xhr parameter as per your needs
});

Call this function when you want control over the XmlHttpRequest object that is used to send error payloads to the API. Pass in a callback that receives one parameter (which is the XHR object). Your callback will be called after the XHR object is opened, immediately before it is sent.

For instance, you can use this to add custom HTTP headers.

Custom error grouping

You can control custom grouping for error instances by passing in a callback. This will override the automatic grouping and be used to group error instances together. Errors with the same key will be placed within the same error group. The callback's signature should take in the error payload, stackTrace and options and return a string, ideally 64 characters or less. If the callback returns null or or a non-string the error will be grouped using Raygun's server side grouping logic (this can be useful if you only wish to use custom grouping for a subset of your errors).

var groupingKeyCallback = function (payload, stackTrace, options) {
  // Inspect the above parameters and return a hash derived from the properties you want

  return payload.Details.Error.Message; // Naive message-based grouping only
};

rg4js('groupingKey', groupingKeyCallback);

Sending custom data

On initialization:

Custom data variables (objects, arrays etc) can be added by calling the withCustomData function on the Raygun object:

rg4js('withCustomData', { foo: 'bar' });

During a send:

You can also pass custom data with manual send calls, with an options object. This lets you add variables that are in scope or global when handled in catch blocks. For example:

rg4js('send', {
  error: e,
  customData: [{ foo: 'bar' }]
});

Providing custom data with a callback

To send the state of variables at the time an error occurs, you can pass withCustomData a callback function. This needs to return an object. By example:

var desiredNum = 1;

function getMyData() {
 return { num: desiredNum };
}

rg4js('withCustomData', getMyData);

getMyData will be called when Raygun4JS is about to send an error, which will construct the custom data. This will be merged with any custom data provided on a Raygun.send() call.

Adding tags

The Raygun dashboard can also display tags for errors. These are arrays of strings or Numbers. This is done similar to the above custom data, like so:

On initialization:

rg4js('withTags', ['tag1', 'tag2']);

During a send:

Pass tags in using an options object:

rg4js('send', {
  error: e,
  tags: ['tag3'];
});

Adding tags with a callback function

As above for custom data, withTags can now also accept a callback function. This will be called when the provider is about to send, to construct the tags. The function you pass to withTags should return an array (ideally of strings/Numbers/Dates).

Affected user tracking

By default, Raygun4JS assigns a unique anonymous ID for the current user. This is stored as a cookie. If the current user changes, to reset it and assign a new ID you can call:

rg4js('getRaygunInstance').resetAnonymousUser();

Disabling anonymous user tracking

rg4js('options', { disableAnonymousUserTracking: true });

Rich user data

You can provide additional information about the currently logged in user to Raygun by calling:

rg4js('setUser', {
  identifier: 'user_email_address@localhost.local',
  isAnonymous: false,
  email: 'emailaddress@localhost.local',
  firstName: 'Foo',
  fullName: 'Foo Bar',
  uuid: 'BAE62917-ACE8-ab3D-9287-B6A33B8E8C55'
});

Only identifier or the first parameter is required. This method takes additional parameters that are used when reporting over the affected users. The full method signature of the public function and the options object above is:

setUser: function (user, isAnonymous, email, fullName, firstName, uuid)

user|identifier is the user identifier. This will be used to uniquely identify the user within Raygun. This is the only required parameter, but is only required if you are using user tracking.

isAnonymous is a bool indicating whether the user is anonymous or actually has a user account. Even if this is set to true, you should still give the user a unique identifier of some kind.

email is the user's email address.

fullName is the user's full name.

firstName is the user's first or preferred name.

uuid is the identifier of the device the app is running on. This could be used to correlate user accounts over multiple machines.

This will be transmitted with each message. A count of unique users will appear on the dashboard in the individual error view. If you provide an email address, the user's Gravatar will be displayed (if they have one). This method is optional; if it is not called no user tracking will be performed. Note that if the user context changes (such as in an SPA), you should call this method again to update it.

Resetting the user

You can now pass in empty strings (or false to isAnonymous) to reset the current user for login/logout scenarios.

Version filtering

You can set a version for your app by calling:

rg4js('setVersion', '1.0.0.0');

This will allow you to filter the errors in the dashboard by that version. You can also select only the latest version, to ignore errors that were triggered by ancient versions of your code. The parameter should be a string in the format x.x.x if you want to get the version sorting in Raygun to work nicely, where x is a non-negative integer.

Filtering sensitive data

You can blacklist keys to prevent their values from being sent it the payload by providing an array of key names:

rg4js('filterSensitiveData', ['password', 'credit_card']);

If any key matches one in the input array, its value will be replaced with [removed by filter].

You can also pass RegExp objects in the array to filterSensitiveData, for controllable matching of keys:

var creditCardDataRegex = /credit\D*/; // Remove any keys that begin with 'credit'

rg4js('filterSensitiveData', [creditCardDataRegex]);

Change filter scope

By default this is applied to the UserCustomData object only (legacy behavior). To apply this to any key-value pair, you can change the filtering scope:

rg4js('setFilterScope', 'all'); // Filter any key in the payload
rg4js('setFilterScope', 'customData'); // Just filter the custom data (default)

Source maps support

Raygun4JS now features source maps support through the transmission of column numbers for errors, where available. This is confirmed to work in recent version of Chrome, Safari and Opera, and IE 10 and 11. See the Raygun souce maps documentation for more information.

Offline saving

The provider has a feature where if errors are caught when there is no network activity they can be saved (in Local Storage). When an error arrives and connectivity is regained, previously saved errors are then sent. This is useful in environments like WinJS, where a mobile device's internet connection is not constant.

Offline saving is disabled by default. To change it:

rg4js('saveIfOffline', true);

If an error is caught and no network connectivity is available (the Raygun API cannot be reached), or if the request times out after 10s, the error will be saved to LocalStorage. This is confirmed to work on Chrome, Firefox, IE10/11, Opera and WinJS.

Limited support is available for IE 8 and 9 - errors will only be saved if the request times out.

Errors in scripts on other domains

Browsers have varying behavior for errors that occur in scripts located on domains that are not the origin. Many of these will be listed in Raygun as 'Script Error', or will contain junk stack traces. You can filter out these errors by settings this:

rg4js('options', { ignore3rdPartyErrors: true });

Whitelisting domains

There is also an option to whitelist domains which you do want to allow transmission of errors to Raygun, which accepts the domains as an array of strings:

rg4js('options', { ignore3rdPartyErrors: true });
rg4js('whitelistCrossOriginDomains', ['code.jquery.com']);

This can be used to allow errors from remote sites and CDNs.

The provider will default to attempt to send errors from subdomains - for instance if the page is loaded from foo.com, and a script is loaded from cdn.foo.com, that error will be transmitted on a best-effort basis.

To get full stack traces from cross-origin domains or subdomains, these requirements should be met:

  • The remote domain should have Access-Control-Allow-Origin set (to include the domain where raygun4js is loaded from).

  • For Chrome the script tag must also have crossOrigin="Anonymous" set.

  • Recent versions of Firefox (>= 31) will transmit errors from remote domains will full stack traces if the header is set (crossOrigin on script tag not needed).

In Chrome, if the origin script tag and remote domain do not meet these requirements the cross-origin error will not be sent.

Other browsers may send on a best-effort basis (version dependent) if some data is available but potentially without a useful stacktrace. The provider will cancel the send if no data is available.

Release History

View the changelog here