I have an HTML5 video with a MediaSource for its source. When I call SourceBuffer.appendBuffer() the SourceBuffer throws an error.
I'm able to detect when this error is thrown like so:
sourceBuffer.addEventListener("error", function() {
console.log(arguments);
});
However the "arguments" in this case contain an ErrorEvent with no meaningful message or data. Just a reference to the SourceBuffer that failed.
I need to figure out why it failed, so I can fix it.
Yes, this is one of the most frustrating parts about MSE... it's really difficult to debug. I think the issue is that the errors have to be standardized from browser to browser. Since there's a lot of
If you're using Chrome, the best place to look is chrome://media-internals.
As of Chrome 91 the media-internals pane is removed. Instead. go to
DevTools / hamburger menu (=== Customize and Control DevTools) / More Tools ▶ / Media
Then you'll have a Media tab on DevTools.
I'm using chrome Version 55.0.2883.87 m (64-bit) on Windows 10.
The following simple html file reproduces the problem and is extracted from my more complex app. It is supposed to speak the 3 words on page load. It works on MS Edge and Firefox but does not work on chrome. This code was working for me on Chrome no problem a couple weeks back.
<html>
<head>
<script lang="javascript">
window.speechSynthesis.speak(new SpeechSynthesisUtterance("cat"));
window.speechSynthesis.speak(new SpeechSynthesisUtterance("dog"));
window.speechSynthesis.speak(new SpeechSynthesisUtterance("bark"));
</script>
</head>
<body></body>
</html>
I may never know for sure, because this problem was intermittent, but it seemed to go away after I started to cancel right before speak.
utter = new window.SpeechSynthesisUtterance("cat");
window.speechSynthesis.cancel();
window.speechSynthesis.speak(utter);
I don't think the cancel necessarily has to come between the utterance object creation and use. Just that it come before every speak. I may have had a different problem as I was only creating one utterance object, not a bunch. I did only see it on Chrome 78. Using Windows 7, 64-bit. Never saw the problem on Firefox or Edge.
EDIT 2 weeks later. No recurrences after several dozen tries. It seems .cancel() solved my problem. My symptoms were: calling speechSynthesis.speak() in Chrome would sometimes not start the speech. There were no immediate indications of a problem in the code, speechSynthesis.speaking would be true and .pending would be false. There would be no events from the utterance object. Normally, when speech would work, I'd get a 'start' event about 0.1 seconds after calling .speak().
speechSynthesis.speak() is no longer allowed without user activation in Google's Chrome web browser since 2018. It violates autoplay policy of Google Chrome. Thus Google Chrome has managed to revoke it's autoplay functionality but you can make use of it by adding a button to make a custom call.
You can visit here to check the status provided by chrome itself also below is the image attached which clearly shows that speechSynthesis.speak() call is prohibited without user's permission.
Link to image
Link to article by Google Chrome
To add to this, the issue for me was the playback rate on the instance of SpeechSynthesisUtterance was above 2. I discovered it must be set to 2 or less in chrome (although it works with higher rates in other browsers like safari).
In chrome, if the utterance rate is above 2, it causes the window.speechSynthesis to be stuck, and needs window.speechSynthesis.cancel() before it will play audio again (at a valid rate below 2) via .speak().
Did your text to voice tryout work only once? Here is why.
In chrome you have to cancel the speechSynthesis, otherwise its not compliant to googles autoplay policy. So you should start your script with:
window.speechSynthesis.cancel()
To cancel any speech synthesis that happened before.
resultsDisplay = document.getElementById("rd");
startButton = document.getElementById("startbtn");
stopButton = document.getElementById("stopbtn");
recognition = new (window.SpeechRecognition || window.webkitSpeechRecognition || window.mozSpeechRecognition || window.msSpeechRecognition)();
recognition.lang = "en-US";
recognition.interimResults = false;
recognition.maxAlternatives = 5;
recognition.onresult = function(event) {
resultsDisplay.innerHTML = "You Said:" + event.results[0][0].transcript;
};
function start() {
recognition.start();
startButton.style.display = "none";
stopButton.style.display = "block";
}
function stop() {
recognition.stop();
startButton.style.display = "block";
stopButton.style.display = "none";
}
.resultsDisplay {width: 100%; height: 90%;}
#stopbtn {display: none;}
<div class="resultsDisplay" id="rd"></div>
<br/>
<center>
<button onclick="start()" id="startbtn">Start</button>
<button onclick="stop()" id="stopbtn">Stop</button>
</center>
Try
utterance = new SpeechSynthesisUtterance("cat, dog, bark");
speechSynthesis.speak(utterance);
I made a Weave at LiveWeave.
Instead of specifying the text while calling new, you could try specifying an object with rate, volume, and text separately, and then converting it to voice.
I've earlier successfully used the JavaScriptAudioNode in the Web Audio API to synthesize and mix audio both in Chrome and Safari 6.0. However, the latest version of Safari no longer appears to work, because it does not call onaudioprocess to fill the source buffers.
This is a simplified example which plays only silence and appends text to the document body on each call to onaudioprocess:
<html>
<head>
<script src="http://ajax.googleapis.com/ajax/libs/jquery/1.9.0/jquery.min.js"></script>
<script type="text/javascript">
$(document).ready(function() {
$("a").click(function() {
var context = new webkitAudioContext();
var mixerNode=context.createJavaScriptNode(2048, 0, 2);
mixerNode.onaudioprocess=function(ape) {
var buffer=ape.outputBuffer;
for(var s=0;s<buffer.length;s++)
{
buffer.getChannelData(0)[s]=0;
buffer.getChannelData(1)[s]=0;
}
$("body").append("buffering<br/>");
};
$("body").html("");
mixerNode.connect(context.destination);
return false;
});
});
</script>
</head>
<body>
start
</body>
</html>
The above example works in Chrome as expected, but not in desktop Safari. The iOS version of Safari does not work either, but it never did work for me in the first place.
Calling context.createJavaScriptNode does return a proper object of type JavaScriptAudioNode and connecting it to the destination node does not throw any exceptions. context.activeSourceCount remains at zero, but this is also the case in Chrome as it apparently only counts active nodes of type AudioBufferSourceNode. context.currentTime also increments as expected.
Am I doing something wrong here or is this an actual bug or a missing feature in Safari? The Apple documentation has no mention of JavaScriptAudioNode (nor the new name, ScriptProcessorNode) but it did work before on the first release of Safari 6. The iOS Safari requirement for user input doesn't seem to help, as the example above should take care of that.
The simple example can be found here and a more complex one is my Protracker module player which exhibits the same behaviour.
There are a couple bugs in Safari's implementation of the Web Audio API that you'll need to look out for. The first is in the createJavaScriptNode constructor... it seems to have problems with the "input channels" param being set to 0. Try changing it to this:
createJavaScriptNode(2048, 1, 2)
The second issue has to do with garbage collection (I think); once your mixerNode variable is out of scope, Safari seems to stop firing the onaudioprocess callback. One solution is to introduce mixerNode at the top-level scope (i.e. declaring var mixerNode; at the top of your script) and then store your JavaScriptNode in that top-level variable. If you plan on dynamically creating multiple mixerNodes, you can achieve the same effect by storing references to them in a top-level array variable.
If you make these two changes (input channel param set to 1, maintaining a reference to the mixerNode) then your script should work in Safari as expected.
First of all, I know there's libraries that provide polyfills for location.pushState/popState (History.js, Hash.js, jQuery hashchange), so please don't just link to those.
I need a more powerful library to achieve the following in a RIA:
User clicks a link
library is notified and loads context via Ajax (no complete reload!)
All <a> elements are leveraged with a click handler that
prevents page reloads in 2. (preventDefault) and
calls location.pushState instead / sets location.hash for older browsers
loaded content is inserted in page and replaces current content
Continue with 1.
Also, previously loaded content should be restored as the user navigates back.
As an example, klick through Google+ in Internet Explorer <10 and any other browser.
Is there anything that comes even close? I need support for IE8, FF10, Safari 5 and Chrome 18. Also, it should have a permissive license like MIT or Apache.
I believe Sammy.js ( http://sammyjs.org) (MIT-licenced) has the best focus on what you want to do, with its 2 main pillars being:
Routes
Events
I could quote from the docs but it's pretty straightforward:
setup clientside routes that relate to stuff to be done, e.g: update the view through ajax
link events to call routes, e.g: call the route above when I click an link. (You would have to make sure e.preventDefault is called in the defined event I believe, since this is an app decision really, so that can't be abstracted away by any library that you're going to use imho)
Some relevant docs
http://sammyjs.org/docs
http://sammyjs.org/docs/routes
http://sammyjs.org/docs/events
Example for a route: (from http://sammyjs.org/docs/tutorials/json_store_1)
this.get('#/', function(context) {
$.ajax({
url: 'data/items.json',
dataType: 'json',
success: function(items) {
$.each(items, function(i, item) {
context.log(item.title, '-', item.artist);
});
}
});
});
Or something like
this.get('#/', function(context) {
context.app.swap(''); ///the 'swap' here indicates a cleaning of the view
//before partials are loaded, effectively rerendering the entire screen. NOt doing the swap enables you to do infinite-scrolling / appending style, etc.
// ...
});
Of course other clientside MVC-frameworks could be an option too, which take away even more plumbing, but might be overkill in this situation.
a pretty good (and still fairly recent) comparison:
http://codebrief.com/2012/01/the-top-10-javascript-mvc-frameworks-reviewed/
( I use Spine.js myself ) .
Lastly, I thought it might be useful to include an answer I've written a while ago that goes into detail to the whole best-practice (as I see it) in client-side refreshes, etc. Perhaps you find it useful:
Accessibility and all these JavaScript frameworks
I currently use PathJS in one of my applications.
It has been the best decision that i have made.
For your particular usecase take a look at HTML5 Example.
The piece of code that that makes the example work (from the source):
<script type="text/javascript">
// This example makes use of the jQuery library.
// You can use any methods as actions in PathJS. You can define them as I do below,
// assign them to variables, or use anonymous functions. The choice is yours.
function notFound(){
$("#output .content").html("404 Not Found");
$("#output .content").addClass("error");
}
function setPageBackground(){
$("#output .content").removeClass("error");
}
// Here we define our routes. You'll notice that I only define three routes, even
// though there are four links. Each route has an action assigned to it (via the
// `to` method, as well as an `enter` method. The `enter` method is called before
// the route is performed, which allows you to do any setup you need (changes classes,
// performing AJAX calls, adding animations, etc.
Path.map("/users").to(function(){
$("#output .content").html("Users");
}).enter(setPageBackground);
Path.map("/about").to(function(){
$("#output .content").html("About");
}).enter(setPageBackground);
Path.map("/contact").to(function(){
$("#output .content").html("Contact");
}).enter(setPageBackground);
// The `Path.rescue()` method takes a function as an argument, and will be called when
// a route is activated that you have not yet defined an action for. On this example
// page, you'll notice there is no defined route for the "Unicorns!?" link. Since no
// route is defined, it calls this method instead.
Path.rescue(notFound);
$(document).ready(function(){
// This line is used to start the HTML5 PathJS listener. This will modify the
// `window.onpopstate` method accordingly, check that HTML5 is supported, and
// fall back to hashtags if you tell it to. Calling it with no arguments will
// cause it to do nothing if HTML5 is not supported
Path.history.listen();
// If you would like it to gracefully fallback to Hashtags in the event that HTML5
// isn't supported, just pass `true` into the method.
// Path.history.listen(true);
$("a").click(function(event){
event.preventDefault();
// To make use of the HTML5 History API, you need to tell your click events to
// add to the history stack by calling the `Path.history.pushState` method. This
// method is analogous to the regular `window.history.pushState` method, but
// wraps calls to it around the PathJS dispatched. Conveniently, you'll still have
// access to any state data you assign to it as if you had manually set it via
// the standard methods.
Path.history.pushState({}, "", $(this).attr("href"));
});
});
</script>
PathJS has some of the most wanted features of a routing library:
Lightweight
Supports the HTML5 History API, the 'onhashchange' method, and graceful degredation
Supports root routes, rescue methods, paramaterized routes, optional route components (dynamic routes), and Aspect Oriented Programming
Well Tested (tests available in the ./tests directory)
Compatible with all major browsers (Tested on Firefox 3.6, Firefox 4.0, Firefox 5.0, Chrome 9, Opera 11, IE7, IE8, IE9)
Independant of all third party libraries, but plays nice with all of them
I found the last too points most attractive.
You can find them here
I hope you find this useful.
i'd like to suggest a combination of
crossroads.js as a router
http://millermedeiros.github.com/crossroads.js/
and hasher for handling browser history and hash urls (w/ plenty of fallback solutions):
https://github.com/millermedeiros/hasher/
(based on http://millermedeiros.github.com/js-signals/)
This will still require a few lines of code (to load ajax content etc.), but give you loads and loads of other possibilities when handling a route.
Here's an example using jQuery (none of the above libraries require jQuery, i'm just lazy...)
http://fiddle.jshell.net/Fe5Kz/2/show/light
HTML
<ul id="menu">
<li>
foo
</li>
<li>
bar/baz
</li>
</ul>
<div id="content"></div>
JS
//register routes
crossroads.addRoute('foo', function() {
$('#content').html('this could be ajax loaded content or whatever');
});
crossroads.addRoute('bar/{baz}', function(baz) {
//maybe do something with the parameter ...
//$('#content').load('ajax_url?baz='+baz, function(){
// $('#content').html('bar route called with parameter ' + baz);
//});
$('#content').html('bar route called with parameter ' + baz);
});
//setup hash handling
function parseHash(newHash, oldHash) {
crossroads.parse(newHash);
}
hasher.initialized.add(parseHash);
hasher.changed.add(parseHash);
hasher.init();
//add click listener to menu items
$('#menu li a').on('click', function(e) {
e.preventDefault();
$('#menu a').removeClass('active');
$(this).addClass('active');
hasher.setHash($(this).attr('href'));
});
Have you looked at the BigShelf sample SPA (Single Page Application) from Microsoft? It sounds like it covers how to achieve most of what you're asking.
It makes use of History.js, a custom wrapper object to easily control navigation called NavHistory and Knockout.js for click handling.
Here's an extremely abbreviated workflow of how this works: first you'll need to initialize a NavHistory object which wraps history.js and registers a callback which executes when there is a push state or hash change:
var nav = new NavHistory({
params: { page: 1, filter: "all", ... etc ... },
onNavigate: function (navEntry) {
// Respond to the incoming sort/page/filter parameters
// by updating booksDataSource and re-querying the server
}
});
Next, you'll define one or more Knockout.js view models with commands that can be bound to links buttons, etc:
var ViewModel = function (nav) {
this.search = function () {
nav.navigate({ page: 2, filter: '', ... }); // JSON object matching the NavHistory params
};
}
Finally, in your markup, you'll use Knockout.js to bind your commands to various elements:
<a data-bind="click: search">...</a>
The linked resources are much more detailed in explaining how all of this works. Unfortunately, it's not a single framework like you're seeking, but you'd be surprised how easy it is to get this working.
One more thing, following the BigShelf example, the site I'm building is fully cross-browser compatible, IE6+, Firefox, Safari (mobile and desktop) and Chrome (mobile and desktop).
The AjaxTCR Library seems to cover all bases and contains robust methods that I haven't seen before. It's released under a BSD License (Open Source Initiative).
For example, here are five AjaxTCR.history(); methods:
init(onStateChangeCallback, initState);
addToHistory(id, data, title, url, options);
getAll();
getPosition();
enableBackGuard(message, immediate);
The above addToHistory(); has enough parameters to allow for deep hash-linking in websites.
More eye-candy of .com.cookie(), .storage(), and .template() provides more than enough methods to handle any session data requirements.
The well documented AjaxTCR API webpage has a plethora of information with downloadable doc's to boot!
Status Update:
That website also has an Examples Webpage Section including downloadable .zip files with ready to use Front End(Client) and Back End(Server) project files.
Notably are the following ready-to-use examples:
One-way Cookie
HttpOnly Cookies
History Stealing
History Explorer
There are quite a bit other examples that rounds out the process to use many of their API methods, making any small learning curve faster to complete.
Several suggestions
ExtJs, see their History Example, and here are the docs.
YUI Browser History Manager.
jQuery BBQ seem to provide a more advanced feature-set over jQuery.hashcode.
ReallySimpleHistory may also be of help, though it's quite old and possibly outdated.
Note: ExtJs History has been extended to optimize duplicate (redundant) calls to add().
PJAX is the process you're describing.
The more advanced pjax techniques will even start to preload the content, when the user hovers over the link.
This is a good pjax library.
https://github.com/MoOx/pjax
You mark the containers which need will be updated on the subsequent requests:
new Pjax({ selectors: ["title", ".my-Header", ".my-Content", ".my-Sidebar"] })
So in the above, only the title, the .my-header, .my-content, and .my-sidebar will be replaced with the content from the ajax call.
Somethings to look out for
Pay attention to how your JS loads and detects when the page is ready. The javascript will not reload on new pages. Also pay attention to when any analytics calls get called, for the same reason.
I want to execute a chrome script,when the content of webpage is update (E.g like Facebook)
So what method I should use for that?
You can use
$("body").bind("DOMSubtreeModified", function() {
alert("something changed");
});
Note: DOMSubtreeModified is deprecated
You should try it with chrome.* API:
chrome.tabs.onUpdated.addListener(
function(integer tabId, object changeInfo, Tab tab) {
//your code
});
See here for documentation:
http://code.google.com/chrome/extensions/tabs.html#event-onUpdated
It should work for updates like (AJAX) requests etc...
As sachleen said, you can use DOMSubtreeModified listener
document.addEventListener("DOMSubtreeModified", methodToRun);
Note that the Mutation Events are performance hogs which can't really be tamed well (they fire too often and slow down the page a lot). Therefore, they have been deprecated over a year ago and should be used only when really needed. However, they work.
If you want this for a Chrome extension, you could use the new and shiny Mutation Observers from DOM Level 4 (follow the links there, they explain a lot!). Where DOMSubtreeModified fired a thousand times, MutationObserver fires only once with all the modifications contained and accessible.
Works for (as of 2012/06):
Chrome 18+ (prefixed, window.WebKitMutationObserver)
Firefox 14+ (unprefixed)
WebKit nightlies