AzureDev

Getting productive with WAMS: How to handle erroneous push channels

WAMS

As I wrote already in my former article ‘Getting productive with WAMS- about the mpns object (push data to the user’s Windows Phone)’, I needed to add some more detailed error handling to the mpns object on my Mobile Service.

There are two error codes that appear frequently: 404 (Not Found) and 412 (Precondition Failed).

The 404 error code

happens when the push channel gets invalid. Reasons for that can be uninstall or reinstall of the app, or also a hard reset of the users device. In this case, we are following the recommendation of Microsoft to stop sending push notifications via this channel.

There might be several ways, but I prefer to work with SQL queries.  This is how I delete those channels within error: function(error):

if (error.statusCode === 404)
{
 var sqlDelInvalidChannel = "DELETE from pushChannel WHERE id = " + channel.id;
 mssql.query(sqlDelInvalidChannel, {
 success: function(){
	console.log("deleted invalid push channel with id: " + channel.id);
	},
	error: function(err)
	 {
           console.log("there was a problem deleting push channel with id: " + channel.id + ", " + err)
         }
});

As you can see, this is a very simple approach to get rid of those invalid channels.

The 412 error code

needs some more advanced handling. Microsoft recommends to send the push notification as normal, but the code recommends a delay of 61 minutes to resend. Also, with some research on the web, I found out that often the 412 will turn into a 404 after those 61 minutes (at least I found a lot of developers stating this). This is why I went with a different approach. I am going to wait those 61 minutes, and do not send a push notification to those devices.

For that, I do a simple trick. I am saving the time the error shows up in my push channel table, as well as the  time after those 61 minutes. On top, I use a Boolean to determine if the channel is in the delay phase.

Here is the simple code for that, again within error: function(error):

if (error.statusCode === 412)
{
	var t = new Date();
	var tnow = t.getTime();
	var tnow61 = tnow + 3660000;

	var sqlSave412TimeStamp = "UPDATE pushChannel SET Found412Time=" +  tnow + ", EndOf412Hour=" + tnow61 + ", IsPushDelayed = 'true' WHERE id=" + channel.id;

	mssql.query(sqlSave412TimeStamp);
}

Now if my script runs the next time over this push channel, I need to check if the push channel is within the delay phase. Here’s the code:

if (channel.IsPushDelayed === true)
{
	var t = new Date();
	var tnow = t.getTime();
	var EndofHour = channel.EndOf412Hour;
	var tdelay = EndofHour - tnow;

	if (tdelay > 0)
	{
	 console.log("push delivery on id: " + channel.id + " is delayed for: " + (Math.floor(tdelay/1000/60)) + " minutes");
	}
	else if (tdelay < 0)
	{
	 var sqlDelete412TimeStamp = "UPDATE pushChannel SET Found412Time=0 , EndOf412Hour= 0, IsPushDelayed = 'false' WHERE id=" + channel.id;
	 mssql.query(sqlDelete412TimeStamp);
	}
}

As you can see, I am checking my Boolean that I added before. If it is still true, I am writing a log entry. If the time has passed already, I am resetting those time values to 0 respective the Boolean to false.

The script will now check if the push channel is still a 412 or turned in to a 404, and so everything starts over again.

Other error codes

There might be other error codes as well. I did not see any other than those two in my logs, but for the case there would be another, I simply added this code to report them:

else
{
 console.error("error in Toast Push Channel: " + channel.twitterScreenName, channel.id, error)
}

This way, you can easily handle push channel errors in your Mobile Service.

If you have other error codes in your logs, check this list from Microsoft to determine what you should do: http://msdn.microsoft.com/en-us/library/windowsphone/develop/ff941100(v=vs.105).aspx#BKMK_PushNotificationServiceResponseCodes

Note: there might be better ways to handle those errors. If you are using such a way, feel free to leave a comment with your approach.

Otherwise, I hope this post will be helpful for some of you.

Happy coding!

Posted by msicc in Azure, Dev Stories, 0 comments

Getting productive with WAMS: How to call Twitter REST API 1.1 from a scheduled script

WAMS.png

Like I promised in my first post about Windows Azure Mobile Services, I will show you how to call the Twitter Rest API 1.1 from a scheduled script. The documentation of the http request object does only use Twitter API 1.0 (which is no longer available).

First, you will need a Consumer key and a Consumer secret for your app. Just go to dev.twitter.com, register with your Twitter account and then add a new application.

The second thing you will need, is the so called Access token and Access token secret. Both are user dependent, without them Twitter will give you an error that your app is not authorized to use this account for anything on Twitter.

There are several ways to obtain these values. As I am registering the user within my phone app, I am uploading these values from phone and store it in my Mobile Services database.

To generate the requested data, we need several additional data for our request to Twitter:

  • a timestamp for the oAuth Header and the signature string
  • a random number to secure the request (= nonce)
  • an oAuth signature (signed array of the user’s data)
  • a HMAC encoded Hash string

These data is used for our request to Twitter.

Let’s start with the “simple” things:

generate Timestamp:

//generating the timestamp for the OAuth Header and signature string
var timestamp  = new Date() / 1000;
timestamp = Math.round(timestamp);

generate nonce

function generateNonce() {
    var code = "";
    for (var i = 0; i < 20; i++) {
        code += Math.floor(Math.random() * 9).toString();
    }
    return code;
}

oAuth signature

//generating the oAuth signatured array for the Twitter request
function generateOAuthSignature(method, url, data) {
    //remove query string parameters
    var index = url.indexOf('?');
    if (index > 0)
        url = url.substring(0, url.indexOf('?'));

    var signingToken = encodeURIComponent(ConsumerSecret) + "&" + encodeURIComponent(twitterAccessTokenSecret);

    var keys = [];
    for (var d in data) {
        if (d != 'oauth_signature') {
            //console.log('data:', d);
            keys.push(d);
        }
    }

    keys.sort();
    var output = "GET&" + encodeURIComponent(url) + "&";
    var params = "";
    keys.forEach(function (k) {
        params += "&" + encodeURIComponent(k) + "=" + encodeURIComponent(data[k]);
    });
    params = encodeURIComponent(params.substring(1));

    return hashString(signingToken, output + params, "base64");
}

generate the HMAC encoded hash string

//generate Hash-string, encoded in HMAC-SHA1 as required by Twitter's API v1.1
function hashString(key, str, encoding) {
    //console.log('basestring:', str);
    var hmac = crypto.createHmac("sha1", key);
    hmac.update(str);
    return hmac.digest(encoding);
}

Now we have prepared all of these functions, we are well prepared to call the Twitter API. In this example we are calling the user’s profile data:

function requestToTwitter()
{

    //the url declaration has to be in this function to make the request working!
    //declaring it in another function would cause an error 401 from Twitter's API
    url = 'https://api.twitter.com/1.1/users/show.json?user_id=' + twitterId;

    //generate data for sending the request to Twitter
    //this is the data used in the signature string as well as in the Authorization header
    var oAuthData = {oauth_consumer_key: ConsumerKey, oauth_nonce: nonce, oauth_signature: null, oauth_signature_method: "HMAC-SHA1", oauth_timestamp: timestamp, oauth_token: twitterAccessToken, oauth_version: "1.0"};
    var sigData = {};
    for (var k in oAuthData) {
        sigData[k] = oAuthData[k];
    }
    sigData['user_id'] = twitterId;

    var sig = generateOAuthSignature('GET', url, sigData);
    oAuthData.oauth_signature = sig;

    var oAuthHeader = "";
    for (k in oAuthData) {
        oAuthHeader += "," + encodeURIComponent(k) + "=\"" + encodeURIComponent(oAuthData[k]) + "\"";
    }
    oAuthHeader = oAuthHeader.substring(1);
    //very important to not miss the space after OAuth!
    authHeader = 'OAuth '+oAuthHeader;

    var reqOptions = {
            uri: url,
            headers: { 'Accept': 'application/json', 'Authorization': authHeader }
    };

    var httpRequest = require('request');
        httpRequest(reqOptions,callback );

}

var callback = function(err, response, body) {
    //console.log("in requestToTwitter = callback"); 
            if (err) {
            console.log(err)
            } else if (response.statusCode !== 200) {
                console.log("from twitter callback " + response.statusCode + " response: " + response.body);
            } else {
                var userProfile = JSON.parse(body);
                UserIdFromTwitter = userProfile.id;
                twitterScreenName = userProfile.screen_name;

}
}

You may have noticed that there are several variables that are not declared within these functions. Just declare them globally in your scheduled script.

You can read more about the oAuth authorization process at http://oauth.net/.

There are more services out there that use the oAuth process, so you should be able to convert this for other requests, like getPocket.com (formerly Read It later) and others.

As always, I hope this post was helpful for some of you.

Happy coding!

Posted by msicc in Azure, Dev Stories, 0 comments

Getting productive with WAMS: respect time zone offset for every single user

time_Azure

In my second post about WAMS I will show you how to respect the time zone of every user.

If you have users that are from all over the world, they have all different time zones. Your Mobile Service script runs always at UTC time, and every user gets the same date & time if you send them push notifications or update a live tile for example. Users don’t want to calculate the time zone differences, so we need to handle that for them.

UTC time is the time since 01/01/1970 00:00 in milliseconds. If we know this, it is somewhat easy to show users their local date and time.

Let’s have a look at the Windows Phone code.

To get the local time zone in our Windows Phone app, we only need three lines of code:

TimeZoneInfo localZone = TimeZoneInfo.Local;
DateTime localTime = DateTime.Now;
TimeSpan offsetToUTC = localZone.GetUtcOffset(localTime);

As you can see, we are getting the local time zone first. This is essential as this one is UTC based. Then we are creating a TimeSpan on our actual DateTime object to get the offset. To make this TimeSpan working on our Azure Mobile Service, which uses JavaScript, we need to convert it to milliseconds. That is the value that has an equal value on all programming languages.

useritemLookUp.TimezoneOffset = offsetToUTC.TotalMilliseconds;

This is the final line of code, which is used to update our user’s item in our SQL table row (for example).

Let’s have a look at the Azure code.

The code is similar to our Windows Phone part. First, we need to fetch the time zone offset from our SQL table:

var sql = "select * from users";

mssql.query(sql, {
        success: function (results) {
            if (results.length > 0) {
                for (var i = 0; i < results.length; i++) {
                    userResult = {
                                        TimeZoneOffset: results[i].TimezoneOffset,
                                        }

TimeZoneOffset = userResult.TimeZoneOffset;

This way, we can run through our whole table on an Azure Script and calculate the correct time, which is pretty easy to achieve:

var d = new Date();
var locald = new Date(d.getTime() + TimeZoneOffset);

These two lines generate the local time in Milliseconds for the specific user entry. You don’t have to worry whether a user is before or after UTC, it will always calculate the correct time.

If you want to use it for example to show the updated time to your users, you can format the time like I described earlier in this post.

That’s all about respecting local time for your users with Windows Azure and Windows Phone.

Happy coding everyone!

Posted by msicc in Azure, Dev Stories, 0 comments

Getting productive with WAMS: how to update data for a specific row in a table

WAMS.png

Like I promised, I will share some of the Azure goodness I learned during creating my last app.

This post is all about how to update a specific table entry (like a user’s data) in a Azure SQL table from an Windows Phone app.

First, we need to make sure that there is some data from the user we want to update. I used the LookupAsync () method to achieve that.

IMobileServiceTable<userItems> TableToUpdate = App.MobileService.GetTable<userItems>();
IMobileServiceTableQuery<userItems> query = TableToUpdate.Where(useritem => useritem.TwitterId == App.TwitterId);

var useritemFromAzure = await query.ToListAsync();
var useritemLookUp = await TableToUpdate.LookupAsync(useritemFromAzure.FirstOrDefault<userItems>().Id);

If we want a specific entry, we need a search criteria to find our user and fetch the id of the user’s table entry. In my case, I used the Twitter Id for the query as every user has this on my project.

Now that we have the table row id, we can easily update the data of this specific row with the UpdateAsync() method.

We need to declare which columns should be updated and asign the values to it first. After that, we simply call the UpdateAsync() method.

useritemLookUp.TwitterId = App.TwitterId;
useritemLookUp.LastCheckedAt = DateTime.Now;
useritemLookUp.OSVersion = "WP8";
useritemLookUp.AppVersion = App.VersionNumber;

await TableToUpdate.UpdateAsync(useritemLookUp);

Please note that you need a items class/model to create the update data (which you should have already before thinking about updating the data).

You should wrap this code in a try{}/catch{} block to be able to react to the Exceptions that possibly can be thrown and display a matching message to the user.

That’s already all about updating a specific row in a WAMS SQL table.

As always I hope this post is helpful for some of you.

Happy coding!

Posted by msicc in Azure, Dev Stories, 1 comment

New Series: Getting productive with Windows Azure Mobile Services (WAMS)

WAMS

Now that my current app project is near to go live, I will start a new series about how to get productive with Windows Azure Mobile Services (WAMS).

I will cover some interesting topics in this series, which are not really documented in the very well written “Getting started” series from the Azure team itself.

These topics are (list is subject to be updated if needed):

First, if you want to get  started, you should check out this link: http://www.windowsazure.com/en-us/develop/mobile/tutorials/get-started-wp8/

The tutorial makes you very easy and fast using WAMS.

What can you expect from this series? As always, I will add some of my personal experiences during my journey of creating my app. There were a lot of small stones in my way, and I will also tell you how to remove them. And of course I hope that my posts will help some of you to get their own WAMS story started.

Happy coding everyone!

Posted by msicc in Azure, Dev Stories, 0 comments

How to format Date and Time on Windows Azure

time_Azure

Phew, my first post about my journey on starting development on Windows Azure. I started a few weeks ago using the Mobile Services from Windows Azure, and I did learn a lot about it.

This post is about formatting Date and Time strings, because Azure uses a different format than my Windows Phone app.

If we upload a DateTime String to Windows Azure from a Windows Phone app, it looks like this: 2013-05-04T06:45:12.042+00:00

If we translate this, you have “YYYY-MM-DD” for the date. The letter “T” declares that the time string  is starting now, formatted “HH:MM:ss.msmsms”. The part “+00:00” is the timezone offset.

So far, probably nothing new for you.

Now let’s get to Azure. Azure by standard uses the GMT time for Date strings (DateTime() in JavaScript = Date()). I have written a scheduler which fetches data from another web service and puts it into my table. Naturally, I wanted to know when the data were last checked, so I added a column for it.

Then I did what everyone that is new to JavaScript has done and added a variable with a new Date(). And now the trouble begins. The output of new Date() is a totally different string: Sat, 04 May 2013 07:02:51 GMT.

Sure, we can parse and convert it within our app, but that would need (although not much) additional resources. So I decided to let to Azure the conversion to a Windows Phone readable string.

How do we  manipulate the Date()-string?

I binged a bit and finally found a very helpful page, that explains all about the JavaScript Date() object: http://www.elated.com/articles/working-with-dates/

I then started off with the following code:

var d = new Date();
var formattedDate = d.getFullYear() + "-" + d.getMonth() + "-" + d.getDate();
var formattedTime  = d.getHours() + ':' d.getMinutes() + ':' + d.getSeconds();
var checkedDateTime = formattedDate + "T" + t;

Those of you that are familiar with JavaScript will immediately see what I did wrong. Let me explain for the newbies:

First thing, date().getMonth is zerobased. So we will always get a result that is one month behind. We have to get it this way for the correct month:

d.getMonth()+1

But that is not all. If you will use the code above, your result will look like this: 2013-5-4T7:2:51

JavaScript does not use leading zeros. If you want to insert it into a date formatted column, you will get the following error from Azure:

Error occurred executing query: Error: [Microsoft][SQL Server Native Client 10.0][SQL Server]Conversion failed when converting date and/or time from character string.

So we need to add the leading zero before inserting it. Luckily we are able to that very easy. Here is my implementation:

var d = new Date();
var formattedDate = d.getFullYear() + "-" + ('0' + (d.getMonth()+1)).slice(-2) + "-" + ('0' + d.getDate()).slice(-2);
var formattedTime  = ('0' + d.getHours()).slice(-2) + ':' + ('0' + d.getMinutes()).slice(-2) + ':' + ('0' + d.getSeconds()).slice(-2);
var checkedDateTime = formattedDate + "T" + t;

What have we done here?

We are adding the leading 0 to each object string. The slice(-2) is for only picking the last two numbers. To make it more clear: if we have 9 as hour, adding the zero in front results in 09. Picking only the last two numbers by .slice(-2) results in still in 09. If we have 10 as hour, adding the leading zero results in 010. But the .slice(-2) operation will cut it back to 10. Easy enough, right?

If we run the code above to get the Date and Time, the result will look like this: 2013-05-04T7:02:51

The timezone offset is automatically added to the date when we update the table. If we now send the data to our Windows Phone or Windows 8 app, no conversion is needed as we already have a correctly formatted string.

I hope this is helpful for some of you and will save you some time.

Happy coding everyone!

Posted by msicc in Azure, Dev Stories, 0 comments