jQuery Mobile and Google Maps API v3 memory leak - memory-leaks

I have a mobile web app using jQuery Mobile 1 and Google Maps API v3. It appears there is a major memory leak when loading Google Maps inside jQuery Mobile. The memory taken by the Google Maps does not get released when the user leaves the map page. So, when the user returns to the map page the memory usage keep growing. This is especially a big issue on mobile web browsers. I test the following on Android and on a desktop version of Chrome. The leak can be triggered as follows:
Navigate to: http://www.mycoursewalk.com/mobile_course_walk/show/25
Click "View Course" and let the map load fully.
Then click the browser back button or the back button on the lower left corner.
Reclick "View Course" and let the map load fully.
Repeat. On every map load you see about 30MB added to the process of Chrome or the mobile browser.
How can I fix this? I can't find a Map API unload.
Thanks,
Nick,

I think you are calling createMap every time instead of checking whether or not the map already exists and re-using the object.
Try something like this:
var mapDisplay = {
mapCreated : false,
init: function() {
//some init stuff
},
showMap: function() {
if(!mapDisplay.mapCreated) {
mapDisplay._createMap();
mapDisplay.mapCreated = true;
}
$("#my_map").show();
$('#my_map').gmap('refresh');
},
_createMap : function() {
//options would be a json config object
$('#my_map').gmap(options);
},
//all the other methods you need
}
which would work with the libraries used in this jQuery gmap demo application

Related

Google analytics stores username and password as a part of url

Issue Context:
I am using meteor js for a mobile app.
I have hooked it up with google analytics calls and basically I am using two type of calls:
Screen views
Events
Screen views are just fine, but I'm facing an issue with the events.
When I go to Behavior -> Events -> Screens, in the google analytics dashboard, I can see the URL of every page that has triggered an event under the Screen Name column. My problem is that the page URLs for my login page look something like this:
meteor.local/login?username=*******&password=+++++++&rememberMe=on
Where ******* is an actual username and +++++++ is the corresponding password!
Reason:
Since I have to share this analytics account with multiple people, I do not want this information to be available over here.
Clues:
CLUE 1:
I used to do GET http calls, but I have changed them all to POST and it still has not fixed the issue as I expected it not to pass plain parameters through URL anymore.
CLUE 2:
I've noticed that the default google analytics js framework is working with http and not https. I was wondering if it is calling the analytics server with a GET as well. If so, is there anyway to change that?
CLUE 3:
Here is how I am initiating the GA instance:
(function (i, s, o, g, r, a, m) {
i['GoogleAnalyticsObject'] = r;
i[r] = i[r] || function () {
(i[r].q = i[r].q || []).push(arguments)
}, i[r].l = 1 * new Date();
a = s.createElement(o),
m = s.getElementsByTagName(o)[0];
a.async = 1;
a.src = g;
m.parentNode.insertBefore(a, m)
})(window, document, 'script', '//www.google-analytics.com/analytics.js', 'ga');
ga('create', googleKey, 'auto');
CLUE 4:
I have also noticed that these URLs are getting captured very occasionally. E.g. in the pas 12,500 unique events (about 30,000 total events) it has captured just 9 URLs with the username and password. The remaining 12,491 events have
meteor.local/login
OR
meteor.local/--
OR
localhost/--
as the Screen Name.
CLUE 5:
I have also put 4 "search and replace" global filters on the analytics account to search for this string
meteor.local/.*
and replace it with this one
meteor.local/concealedURI
This does not seem to be working either.
I have added this filter on 4 different fields (Since I still really don't know where the URLs are coming from):
Host Name
Page Title
Referral
Request URI
CLUE 6:
This is how I am calling the GA instance to send the event:
ga('send', 'event', 'button', 'click', eventName);
Okay. So, I had to run a lot of experiments and try out different things to solve this issue.
After trying all the things that I have described in the question, I finally found a way to address this problem.
The main cause of this problem was that I was using a google analytics account set to track an App, to capture the data from an app that was built with meteor js (which basically utilizes cordova).
Using meteor means that my app's screens are actually web pages rendered as a mobile app. It seems like meteor uses URLs to navigate through these screens.
On the other hand, google analytics looks at (and captures) the screen name of an app's page, when an event is triggered from that page. In native apps this screen name will be something similar to "About us", "Contact Us", "Home", etc.
Now since a meteor app is not the same, the screen name returned by meteor is actually the URL of the page that has triggered the event.
This does not have anything to do with the http calls (Whether or not they are GET or POST), because it is the local URL used by meteor for navigating that is being passed down to google analytics and not any http calls.
Solutions
1.
If I had the google analytics account set as a web page tracker, I could have access to "Exclude URL Query Parameters" field and I could potentially exclude username and password as was suggested by #Mike and #PhilipPryde in the comments.
However, I needed to use google analytics set as an app tracker. So, this did not work for me.
Failed
2.
I did put a filter on the whole view in the google analytics and searched for meteor.local/.* and replaced that with hiddenURL. The filters on
Host Name
Page Title
Referral
Request URI
did not work.
But when I put the same filter on
Screen Label
field, it worked.
However, this only looked at the screen names returned by screen view hits and not the event. Thus, this did not actually solve my issue either.
Failed
Finally, I had to do this:
There is a method call on GA instance that lets you set different options up. I ended up using this:
ga('set', 'screenName', 'hiddenURL');
This changed the screen name to "hiddenURL". So, I used this before every event and it worked for me.
My code for sending events to google analytics looked like this:
ga('set', 'screenName', 'hiddenURL');
ga('send', 'event', 'button', 'click', eventName);
PS:
This changes the screen name that was showing up in real-time reports of google analytics to "hiddenURL", whenever someone triggered an event. But, it changes back to a screen name as soon as they go to another page. So, it would not also mess with any of your screen view data either, since it is not being captured as a screen view.
Of course that is because, I pass the screen name to my GA instance every time I send a screen view. So it looks like this:
sendScreenViewToGA = function (screenName) {
ga('send', 'screenview', {
'appName': 'Something',
'screenName': screenName,
'appVersion': x.x
});
}
If I had used the screen name, that is being set on the environment tight now, I would have ended up with all my screen names in analytics set to "hiddenURL".
I really hope this post will help others with same issues and save them some time.

Detect My page rendering in Chrome App

I have followed this link ChromeApp for my chromeApp
I want to detect that Is my HTML page rendering on ChromeApp?
if(chromeApp){
//do this
}
else{
//do this
}
To answer the general question: Outside of the webview you can detect if you are rendering in a Chrome App by:
if (chrome && chrome.app && chrome.app.runtime)
// chrome app.
else
// open web.
(Taken from gapi-chrome-apps.js)
To answer the intent of the question, "How can detect and not show alert messages for content in a webview", you may wish to change the user agent and use that to detect. See this test code for more. Here's the idea, though:
webview.setUserAgentOverride(webview.getUserAgent() + ' in a webview');
Also, you can support alert dialogs with the change 19679002: : Implement dialog API (not quite in Chrome stable I think). The following should illustrate:
From the host:
webview.addEventListener('dialog', function(e) {
// Check e.messageType, e.g. it may be 'alert'.
// Use e.messageText
// Unblock the guest content wity e.dialog.ok();
});

chrome extension - alternative to externally_connectable?

It seems like the externally_connectable feature that allows a website to communicate with an extension is still in the dev channel and not yet stable. Are there any other ways to allow a specific website to communicate with my extension, while I wait for this feature to become stable? How have chrome extension developers traditionally done it?
Thanks Rob W for pointing me in the direction of HTML5 messaging. For the benefit of other chrome extension developers, I'm writing about the general problem I was trying to solve and the solution that worked in the end.
I am making a chrome extension that can control music playback on a tab via a popup player. When a user clicks on play/pause/etc on the popup player, the extension should be able to convey that message to the webpage and get back a response stating whether the action was accomplished.
My first approach was to inject a content script into the music player page. The problem is, though, that content scripts operate in a "sandbox" and cannot access native javascript on the page. Therefore, the content script was pretty useless (on its own), because while it could receive commands from the extension, it could not effect any change on the webpage itself.
One thing that worked in my favor was that the website where the music was playing belongs to me, so I could put whatever javascript I wanted there and have it be served from the server. That's exactly what I used to my advantage: I created another javascript file that would reside on the website and communicate with the content script mentioned above, via the window object of the page (i.e. HTML5 messaging). This only works because the content script and the javascript file both exist in the same webpage and can share the window object of the page. Thanks Rob W for pointing me to this capability. Here is an example of how the javascript file on the page can initiate a connection with the content script via the window object:
content_script.js (injected by extension into xyz.com):
window.addEventListener("message", function(event) {
if(event.data.secret_key &&
(event.data.secret_key === "my_secret_key") &&
event.data.source === "page"){
if(event.data.type){
switch(event.data.type) {
case 'init':
console.log("received connection request from page");
window.postMessage({source: "content_script", type: 'init',
secret_key: "my_secret_key"}, "*");
break;
}
}
}
}, false);
onpage.js (resides on server and served along with xyz.com):
window.postMessage({source: "page", type: 'init',
secret_key: "my_secret_key"}, "*");
window.addEventListener("message", function(event) {
if(event.data.secret_key &&
(event.data.secret_key === "my_secret_key") &&
event.data.source === "content_script"){
if(event.data.type){
switch(event.data.type) {
case 'init':
console.log("connection established");
break;
}
}
}
}, false);
I check the secret key just to make sure that the message originates from where I expect it to.
That's it! If anything is unclear, or if you have any questions, feel free to follow up!
You could have an extension inject a content script alongside a web page, and use that to pass messages back and forth between the website and the background page of the extension.
It's tedious, though, and externally connectable is a lot nicer.

Opening and writing to a new window from a Google Chrome extension sandbox page

(Cross posted here)
Hi,
I have a sandboxed page (specified in my manifest) which is loaded into an iframe in my extension's background page. From within my sandboxed page, I'd like to open a new window and write to it, i.e.:
var win = window.open(); win.document.write('<p&gtHello!</p>');
This works from my extension's background page and from regular web pages, but when invoked from either content scripts or my sandboxed page, the window opens, but I cannot access the win object (it's defined, but empty---console.log outputs "Window {}").
I assume this is due to same-origin policies (with every window being given a uinque-origin within the sandboxed environment). However, since the window opens an about:blank page, I'm confused why this would matter.
Is this a feature? Is there a parameter I can add to my manifest to avoid this? And does anyone know of work-arounds that don't involve using postMessage back to my background page? My ideal solution is to have my sandboxed script open a new window and interact with it directly, not with message passing.
I can provide a full example if necessary, but I'm hoping someone might just know off the top of their head. I'm running Chrome 24.0.1312.57 on Mac and 24.0.1312.68 on Ubuntu if that helps.
Thanks,
Hank
Content scripts, by definition, are part of external regular web pages you load so I'm not really sure how your script could work on a "regular web page" but not the content script. Do you mean the code works when you embed it in your own pages, but not in other pages via the content script?
Regardless, if the script is working properly from your background page, you could always try messaging. (more here: http://developer.chrome.com/extensions/messaging.html)
Script for your sandbox/contentscript:
//send message to background page
chrome.extension.sendMessage({todo: "newWindow"});
In background page:
//create a listener
chrome.extension.onMessage.addListener(
function(request, sender) {
if (request.todo === "newWindow") {
//do your stuff here
var win = window.open(); win.document.write('<p&gtHello!</p>');
}
});
Per the cross-post here, the issue is indeed that the opened window is given a unique origin. This was intentional as the members of the standards working group (SWG) felt that it would be more secure to not make an exception for about:blank pages where they inherit the sandbox's origin.
However, to get around this issue, at least for my purposes, I can use the following method. First, forget sandboxing. This workaround uses an iframe embedded in a background page with the src url set to data:text/html,.... This gives a unique origin to the iframe so it's no longer in extension space. That means eval can be used and chrome apis cannot be accessed. Unlike in a sandbox, windows opened from the iframe share that same origin as the iframe, allowing created windows to be accessed. For example:
In a background html page:
<html>
<head>
...
<script src="background.js"></script>
...
</head>
<body>
...
<iframe id="myframe"></iframe>
...
</body>
</html>
In background.js:
...
document.getElementById('myframe').setAttribute('src', 'data:text/html,'+
encodeURI('<html><head>'+
'<script src='+chrome.extension.getURL('jquery.js')+'></script>'+
'<script src='+chrome.extension.getURL('myscript.js')+'></script>'+
...
'</head><body></body></html>'
));
...
In myscript.js
jQuery(document).ready(function(){
...
// To receive messages from background.js.
window.addEventListener('message', function(e){ ... } );
// To send messages to background.js.
parent.postMessage({...}, '*');
// To open AND ACCESS a window.
var win = window.open();
win.document.write('Hello'); // Fails in sandbox, works here.
// Eval code, if you want. Can't do this from an extension
// page loaded normally unless you allow eval in your manifest.
// Here, it's okay!
eval( 'var x = window.open(); x.document.write("Hi!")' );
// No chrome apis.
chrome.log( chrome.extension ); // -> undefined
chrome.log( chrome.windows ); // -> undefined
// No direct access to background page (e.g., parent).
chrome.log( parent ); // -> Window with no properties.
...
});

Unable to use getBackgroundPage() api from a devtools extension

I am trying to write an extension that adds functionality to the Chrome devtools.
According to the devtools documentation, it says that the pages in devtools support very limited apis. Any API that is not supported can be access by accessing it through the background page, just as what contentscripts does.
Here is the relevant documentation snippet:
The tabId property provides the tab identifier that you can use with the chrome.tabs.* API calls. However, please note that chrome.tabs.* API is not exposed to the Developer Tools extension pages due to security considerations — you will need to pass the tab ID to the background page and invoke the chrome.tabs.* API functions from there.
Here is the source url: http://developer.chrome.com/extensions/devtools.inspectedWindow.html
However, when I try to do that, I get the following error in the console:
uncaught Error: "getBackgroundPage" can only be used in extension processes. See the content scripts documentation for more details.
Here is my code in my devtools.js script:
chrome.extension.getBackgroundPage().getLocation();
What am I doing wrong?
EDIT
I should describe my scenario first, and show how I am implementing it.
What I want to do is to display extra data in a devtools panel related to a webpage. In order to get that data, I will need to send a HTTP request in the same session as the page being debugged, because it requires authentication.
Use Case:
User browses to a particular URL. He is authenticated to the site. He then invokes devtools. The devtools panel opens up and a new panel shows up that has extra data related to the page.
Implementation:
1) DevTools script finds out the url of the page being inspected. If the url matches the site base hostname, then it opens a panel. In the callback of the panel creation, it sends a message to a background page, asking it to download a JSON payload from a debug endpoint on the same site, and then sends it to the devtools extension, wh ich then displays it.
Problems:
1) The background page gets the request, and downloads the URL. However the download is not using the same session as the user, so the download request fails.
2) From devtools window, I got the tabId of the inspected window. I send this tabId to the background page so that it can parse some stuff out of the url. However, chrome.tabs.get(tabId) does not return the tab.
To summarize, I need to
1) Get the background page to download data in the same session as the user's tab that is being debugged.
2) I need to have the background page be able to get access to the user's tab.
The APIs available to extension pages within the Developer Tools window include all devtools modules listed above and chrome.extension API. Other extension APIs are not available to the Developer Tools pages, but you may invoke them by sending a request to the background page of your extension, similarly to how it's done in the content scripts.
I guess the documentation is little ambiguous, By chrome.extension API they mean the Supported API's for content scripts.
So, you can use long lived communication for communication between inspected page and background page
Demonstration:
The following code illustrate scenario where a devtools page need some information from background page, it uses messages for communication.
manifest.json
Ensured permissions are all available in manifest file
{
"name":"Inspected Windows Demo",
"description":"This demonstrates Inspected window API",
"devtools_page":"devtools.html",
"manifest_version":2,
"version":"2",
"permissions":["experimental"],
"background":{
"scripts" : ["background.js"]
}
}
devtools.html
A trivial HTML File
<html>
<head>
<script src="devtools.js"></script>
</head>
<body>
</body>
</html>
devtools.js
Used Long lived Communication API's
var port = chrome.extension.connect({
name: "Sample Communication"
});
port.postMessage("Request Tab Data");
port.onMessage.addListener(function (msg) {
console.log("Tab Data recieved is " + msg);
});
background.js
Responded to communication request and passed trivial information using tab API()'s
chrome.extension.onConnect.addListener(function (port) {
port.onMessage.addListener(function (message) {
chrome.tabs.query({
"status": "complete",
"currentWindow": true,
"active": true
}, function (tabs) {
port.postMessage(tabs[0].id);
});
console.log("Message recived is "+message);
});
});
Sample Output received for trivial devtools.js here
Let me know if you need more information
EDIT 1)
For your question 1)
Can you make you call(s) from browser extension HTML Page\Content Script so same session is shared, i have tried both the ways in a sample and it is working form me, instead of code in background page- make the code in content script or browser action HTML Page.
Let me know if you are still facing problems.
For your question 2)
The following code always fetches current window user is browsing
manifest.json
Ensure you have tabs permission in your manifest.
{
"name":"Inspected Windows Demo",
"description":"This demonstrates Inspected window API",
"manifest_version":2,
"version":"2",
"permissions":["tabs"],
"background":{
"scripts" : ["background.js"]
}
}
background.js
chrome.tabs.query({
"status": "complete", // Window load is completed
"currentWindow": true, // It is in current window
"active": true //Window user is browsing
}, function (tabs) {
for (tab in tabs) { // It returns array so used a loop to iterate over items
console.log(tabs[tab].id); // Catch tab id
}
});
Let me know if you are still unable to get tab id of current window.

Resources