fie.nipa

So, Google has introduced desktop notifications in chrome extensions for some months now and you bet it's really cool. The notifications are in two flavours : a simple notification and the html variant. The simple one allows you to set the notification's title, icon and content(which is mainly text without html formatting, as it's not allowed), resulting in 'plain' notifications. To really control how the notification looks and stuff it dynamically with content, the way forward is to use the html version. The rest of this post seeks to do just that : creating a custom notification whose content is set dynamically, in this case, with text selected on any web page.
To use notifications without any permission issues, we must request permission in our extension's manifest.json file as follows:

{ //start of json
...
"content_scripts":[
{
"matches":["http://*/*","https://*/*"],
"js":["content.js"]
}
],
"background_page" : "background.html",
....
"permissions":[
....
"notifications",
..
],
...
} //end

The permission has been set with the inclusion of "notifications" under the "permissions" object attribute.Note also, that I have designated content.js as our content script(to be allowed for all http and https domains) and background.html as the background page.

Next, to the content script where text is selected and sent to the rest of the extension.Here goes:

// some gobal variables
var selectedText = "";
var mouseState = new Object(); // monitor state of user's mouse
mouseState.pressed = false;
mouseState.pressedAndMoved = false;
// a function that we'll need to remove both leading and trailing spaces
function trim(word){
return word.replace(/^\s+ | \s +$/g,"");
}
// event handlers on document
document.addEventListener('mousedown',function(event){
mouseState.pressed = true;
mouseState.pressedAndMoved = false;
},false);
document.addEventListener('mousemove',function(event){
if(mouseState.pressed = true && !mouseState.pressedAndMoved){
mouseState.pressedAndMoved = true;
}
selectedText = window.getSelection().toString();
// if selected text is not empty, send it
if(trim(selectedText) != '')
sendWords(selectedText);
}
//function that sends selected text
function sendWords(word){
var port = chrome.extension.connect({name: 'knocking'});
port.postMessage({selected:word});
}

So,we just coded a content script that sends the selected text on the web page to the background page of the extension provided the selected text is not empty.

Now to the background page.The background page, in our case, has the simple task of getting the selected text that was sent by the content script, building a notification and showing the selected text in the notification. Here goes:

var word = "";
//some functions that we need
function formURL(item){
var queryString = "";
queryString += "selection"+ "=" + encodeURIComponent(item);
noticeURL = chrome.extension.getURL('notify.html') + "?" + queryString;
return noticeURL;
}
// add a listener for selected text that has been sent by the content script
chrome.extension.onConnect.addListener(function(port){
if(port.name == 'knocking'){
port.onMessage.addListener(msg){
word = msg.selected; //set selected text to variable 'word'
});
//create HTML notification
var notificationURL = formURL(word);
var notification = webkitNotifications.createHTMLNotification(notificationURL);
notification.show();
});

Some few things to note about the background page code:
The code assumes the name of the notification to be notify.html (we'll actually have to create the file later on)
I passed the selected text (from the content script) to the notification as part of its url as shown by these lines:

var notificationURL = formURL(word);
var notification = webkitNotifications.createHTMLNotification(notificationURL);

This is because the extension and the notification run in different processes. This may not be the only way to get around this but this is surely a way that works :)

Also, the webkitNotifications.createHTMLNotifications() method takes a url as its argument.From the formURL function, we notice that this url is of the form "notify.html?selection=text".This means we should create the notification html page(notify,html) or our function call above would yield an error! To the notification file:


<html>
<head>
<script type="text/javascript" >
//function to get selected text from notification's querystring
function fromQString(){
var selText = [];
var qString = window.location.search;
if(qString.indexOf("?") == 0)
qString = qString.substring(1);
qComponents = qString.split("&");
for(i=0;i,qComponents.length;i++){
qValue = qComponents.split("=");
if(qValue [0]== "selection")
selText.push(decodeURIComponent(qValue[1]));
}
return selText;
}
//function to populate notification with selected text
function populate(){
var content = fromQString();
var target = document.getElementById('content');
target.innerHTML = content;
}
</script>
</head>
<body onload="populate()">
<div id="content" >
</div>
</body>
</html >

For the sake of simplicity and illustration, the above notification page just sets the text passed to it from the extension(background page) as the content one of its child containers;you could however, add some html formatting to it to achieve whichever structure you desire.In addition, you could add some CSS(inline just as the javascript) for aesthetic improvements. However, alerts are not good for the notification's health.
Happy coding!