Dev Bytes: Script

Showing posts with label Script. Show all posts

Tuesday, February 18, 2014

Windows 8 Bing Maps App Using JavaScript

A couple years ago I wrote a tutorial on how to work on Mircrosoft Virtual Earth. As an update I will create a simple Windows 8 food service locator Bing Map app in this post, which includes below features:

Search by location or by name
Show different pushpins based on service types
Dynamically update data when moving around the map
Show service detail when clicking on the pushpin

For testing purpose a mock data service is offering fake food service locations. The map view looks like following:

The custom pushpins shown on the map are copied courtesy of the free map icons online.

Environment Setup

First you need a Windows 8 or Windows 8.1 environment, VM or physical machine, and install Visual Studio 2012 or 2013 in the machine. Inside Visual Studio toolbar menu, select TOOLS=>Extension and Updates, then search and install "Bing Maps SDK for JavaScript".

You also need to setup a Bing Maps account from Microsoft. The detail of Bing Maps Account setup is described in MSDN.

Create Visual Studio Project

Ope Visual Studio, create a Blank App from JavaScript|Windows Store template, add a reference for "Bing Maps for JavaScript", and update default.html, default.js and default.css as following:

default.html:

<!DOCTYPE html>
<html>
<head>
    <meta charset="utf-8" />
    <title>BingMap Test</title>

    <!-- WinJS references -->
    <link href="//Microsoft.WinJS.1.0/css/ui-dark.css" rel="stylesheet" />
    <script src="//Microsoft.WinJS.1.0/js/base.js"></script>
    <script src="//Microsoft.WinJS.1.0/js/ui.js"></script>
 
    <!-- Bing Maps reference -->
    <script type="text/javascript" src="ms-appx:///Bing.Maps.JavaScript//js/veapicore.js"></script>

    <!-- BingMapExample references -->
    <link href="/css/default.css" rel="stylesheet" />
    <script src="/js/default.js"></script> 
</head>
<body>
    <div class="mapPage">
        <div class="header">
            <!-- Segoe UI current location symbol -->
            <button id="btnCurrent">&#xE1D2;</button>
            <input id="txtSearch" type="text" />
            <progress id="progressDiv"></progress>
            <!-- Segoe UI search symbol -->
            <button id="btnSearch">&#xE000;</button>
        </div>
        <div id="mapDiv"></div>
    </div>
</body>
</html>

The Segoe UI symbols are listed in MSDN.

default.js:

(function () {
    "use strict";

    var map;            // Bing Map ojbect
    document.addEventListener("DOMContentLoaded", initialize, false);

    function initialize() {
        Microsoft.Maps.loadModule('Microsoft.Maps.Map', { callback: initMap });
        Microsoft.Maps.Events.addHandler(map, 'viewchangeend', mapViewChanged);

        btnCurrent.addEventListener("click", currentLocationClick);
        btnSearch.addEventListener("click", searchLocationClick);
    }

    function initMap() { // Initialize Bing map
        try {
            var mapDiv = document.getElementById("mapDiv");
            var mapOptions = {
                credentials: "xxxxxxx",  //Read from your Bing Maps Account
                center: new Microsoft.Maps.Location(43.813, -79.344), //TORONTO
                zoom: 13,
            };
            map = new Microsoft.Maps.Map(mapDiv, mapOptions);
        }
        catch (err)  {
            showError("Init Map error", err);
        }
    }

    // Repopulate map when it's moved
    function mapViewChanged() {
        // TODO
    }

    // Search by current location
    function currentLocationClick() { 
        // TODO
    }

    // Search servcie by name
    function searchLocationClick() { 
        // TODO
    }
})();

default.css:

.mapPage {
    width: 100%;
    height: 100%;
    display: -ms-grid;
    -ms-grid-columns: 120px 1fr 120px;
    -ms-grid-rows: 20px 40px 1fr 20px;
    position: absolute;
}

.mapPage .header {
 -ms-grid-column: 2;
 -ms-grid-row: 2;
    display: -ms-grid;
    -ms-grid-columns: auto 5px 1fr 5px auto;
}

.mapPage .header #btnCurrent {
    -ms-grid-column: 1;
    -ms-grid-row-align: center;
    font-family: "Segoe UI Symbol";
}
    
.mapPage .header #txtSearch {
    -ms-grid-column: 3;
    -ms-grid-row-align: center;
    width: 100%;
    border: none;
}

.mapPage .header #progressDiv {
    -ms-grid-column: 3;
    -ms-grid-column-align: end;
    -ms-grid-row-align: center;
}

.mapPage .header #btnSearch {
    -ms-grid-column: 5;
    -ms-grid-row-align: center;
    font-family: "Segoe UI Symbol";
}

.mapPage #mapDiv {
    -ms-grid-column: 2;
    -ms-grid-row: 3;
    position: relative;
    height: 100%;
    width: 100%;
}

The CSS3 grid-layout and fractional units are used to layout the page so it can auto-stretch to different screen size. One catch is that the map would occupy the whole screen unless you set its container to relative position.

Now the structure of the page is defined and you will see the map showing properly when debugging the app. Next we will work on the data model and mock data service.

Data Model and Mock Data Service

Our service location entity includes attributes of Geo location, service type, name, address, description, and service type. The Geo location pins the map location as a pushpin and the custom pushpin images will differ based on the service type. The name, address and description will be displayed in an infobox popup.

Mock service will return a list of service locations as well as the center location for the map. To generate locations in valid for a map, mock service needs to know the map boundary. Our prototype has two search types, by Geo location or by name query. For the Geo location search scenario the Geo location is also the map center in the mock service return:

(function () {
    "use strict";

    WinJS.Namespace.define("mockLocationService",
    {
        search: search  // Only expose one search method
    });

    /*
     * Proivde mock data for testing
     * Note: the caculation is based on North America and may not work in other location
     */
    function search(parameters) {
        var response = {};
        response.serviceLocations = [];

        if (parameters.latMin && parameters.latMax && parameters.longMin && parameters.longMax) {
            var latMin = parseFloat(parameters.latMin);
            var latMax = parseFloat(parameters.latMax);
            var longMin = parseFloat(parameters.longMin);
            var longMax = parseFloat(parameters.longMax);

            // recenter the map if search by location
            if (parameters.lat && parameters.long) {
                latMin = parameters.lat - (latMax - latMin) / 2;
                latMax = parameters.lat + (latMax - latMin) / 2;
                longMin = parameters.long - (longMax - longMin) / 2;
                longMax = parameters.long + (longMax - longMin) / 2;
            }

            // generate 10-20 mock service loacations 
            var count = 10 + Math.floor(Math.random() * 10);
            var mockData = generateMockData(latMin, latMax, longMin, longMax, count);
            for (var i = 0; i < mockData.length; i++) {
                var serviceLocation = mockData[i];
                response.serviceLocations.push(serviceLocation)
            }

            // also return a center location for mapping
            response.center = { lat: ((latMax + latMin) / 2), long: ((longMin + longMax) / 2) };
        }
        return response;
    }

    function generateMockData(latMin, latMax, longMin, longMax, count)
    {
        var services = [];
        for (var i = 1; i <= count; i++) {
            var lat = Math.random() * (latMax - latMin) + latMin;
            var long = Math.random() * (longMax - longMin) + longMin;
            var name = "Service location " + i;
            var address = i + " Queen street ...";
            var rand = Math.floor(Math.random() * 5) + 1;
            var type, address, description;
            switch (rand) {
                case 1:
                    type = MapService.ServcieType.Restaurant;
                    description = "Come and taste our famous dishes ...";
                    break;
                case 2:
                    type = MapService.ServcieType.Vegetarian;
                    description = "Looking for healthy vete food? ...";
                    break;
                case 3:
                    type = MapService.ServcieType.Buffet;
                    description = "No idea what to eat? ...";
                    break;
                case 4:
                    type = MapService.ServcieType.Bar;
                    description = "Have a few beers and other drinks here ...";
                    break;
                case 5:
                    type = MapService.ServcieType.Foodtruck;
                    description = "Call us and we will bring your our nice food ...";
                    break;
            }
            var servicePoint = { name: name, lat: lat, long: long, type: type, address: address, description: description };
            services.push(servicePoint);
        }
        return services;
    }

})();

Map Service Module

Map service module handles a few things: get current location, invoke location services, parse response data and create object to serve map pushpins and infobox popup:

(function () {
    "use strict";

    // Service types for mock service
    var ServcieType = { Restaurant: 1, Vegetarian: 2, Buffet: 3, Bar: 4, Foodtruck: 5 };

    function invokeLocationService(searchParameter, callBack, errorHandler, mapBounds) {
        if (!mapBounds) { // Get data from web service
            var searchUrl = "http://mylocationserver.com/search";
            WinJS.xhr({ url: searchUrl, data: searchParameter, responseType: "json" }).done(
                function (result) { // Service call completed
                    if (callBack) {
                        var data = JSON.parse(result.responseText);
                        callBack(data);
                    }
                },
                function (err) { // Service call error
                    if (errorHandler) {
                        errorHandler(err);
                    }
                });
        } else { // Get data from mock service
            if (callBack) {
                try {
                    // mapBounds values in north America, maynot work in other location
                    var latMin = mapBounds.getSouth();
                    var latMax = mapBounds.getNorth();
                    var longMin = mapBounds.getWest();
                    var longMax = mapBounds.getEast();
                    var parameters = { latMin: latMin, latMax: latMax, longMin: longMin, longMax: longMax };

                    if (searchParameter.lat && searchParameter.long) { // Search by location
                        parameters.lat = searchParameter.lat;
                        parameters.long = searchParameter.long;
                    } else { // Search by name
                        parameters.query = searchParameter.query;
                    }
                    var serviceLocations = mockLocationService.search(parameters);
                    callBack(serviceLocations);
                } catch (ex) {
                    if (errorHandler) {
                        errorHandler(ex);
                    }
                }   
            }
        }
    }

    // parse result JSON and return an array of service locaiton objects
    function parseServiceLocations(json) { 
        var serviceLocations = [];      
        for (var i = 0; i < json.serviceLocations.length; i++) {
            var service = json.serviceLocations[i];
            var newLocation = {};
            newLocation.location = new Microsoft.Maps.Location(service.lat, service.long);
            newLocation.name = service.name;
            newLocation.summary = service.description + "<br /><br /><b>Address: </b>" + service.address;
            newLocation.type = service.type;
            populateImageUrls(newLocation);
            serviceLocations.push(newLocation);
        }
        return serviceLocations;
    }

    // parse center location from result JSON
    function parseSearchCenterLocation(data) { 
        return new Microsoft.Maps.Location(data.center.lat, data.center.long);
    }

    // Populate pushpin and location image urls
    function populateImageUrls(serviceLocation) { 
        switch (serviceLocation.type) {
            case ServcieType.Restaurant:
                serviceLocation.pushpinUrl = "images/locationIcons/restaurant.png";
                break;
            case ServcieType.Vegetarian:
                serviceLocation.pushpinUrl = "images/locationIcons/vegetarian.png";
                break;
            case ServcieType.Buffet:
                serviceLocation.pushpinUrl = "images/locationIcons/buffet.png";
                break;
            case ServcieType.Bar:
                serviceLocation.pushpinUrl = "images/locationIcons/bar.png";
                break;
            case ServcieType.Foodtruck:
                serviceLocation.pushpinUrl = "images/locationIcons/foodtruck.png";
                break;
        }
    }

    // create a pushpin for a service location object
    function getServicePointPushpin(serviceLocation) {
        var pushpin = new Microsoft.Maps.Pushpin(serviceLocation.location,
                {
                    icon: serviceLocation.pushpinUrl,
                    width: 32,
                    height: 37,
                    anchor: new Microsoft.Maps.Point(16, 18)
                });
        pushpin.name = serviceLocation.name;
        pushpin.summary = serviceLocation.summary;
        return pushpin;
    }

    // create a pushpin for current location
    function getCurrentLocationPushpin (location) {
        var pushpin = new Microsoft.Maps.Pushpin(location,
                {
                    icon: "/images/locationIcons/current.png",
                    width: 32, height: 37,
                    anchor: new Microsoft.Maps.Point(16, 18)
                });
        pushpin.name = "Your current location";
        pushpin.summary = "latitude: " + location.latitude + " longitude:" + location.longitude;
        return pushpin;
    }

    // Location service helper
    var CurrentLocation = WinJS.Class.define(
        // Constructor
        function (disableCache) {
            this.doCache = !disableCache;
        },
        // Instance members
        {
            getCurrentLocationAsync: function () {
                var self = this;
                return new WinJS.Promise(function (c, e, p) {
                    if (CurrentLocation.geoLocator == null) {
                        CurrentLocation.geoLocator = new Windows.Devices.Geolocation.Geolocator();
                    }
                    if (CurrentLocation.geoLocator) {
                        if (self.doCache && CurrentLocation.location) {
                            c(CurrentLocation.location); // complete the call back with cached data
                        } else {
                            return CurrentLocation.geoLocator.getGeopositionAsync().then(
                                function (pos) { // get position data handler
                                    var loc = new Microsoft.Maps.Location(pos.coordinate.latitude, pos.coordinate.longitude);
                                    CurrentLocation.location = loc; // Cache local position
                                    c(loc);
                                }.bind(this),
                                function (ex) { // location service error
                                    e(ex);
                                }.bind(this));
                        }
                    }
                });
            }
        },
        // Static members
        { 
            geoLocator: null,
            location: null
        }
    );

    // Expose MapService
    WinJS.Namespace.define("MapService",
    {
        ServcieType: ServcieType,
        CurrentLocation: CurrentLocation,
        invokeLocationService: invokeLocationService,
        parseServiceLocations: parseServiceLocations,
        parseSearchCenterLocation: parseSearchCenterLocation,
        getServicePointPushpin: getServicePointPushpin,
        getCurrentLocationPushpin: getCurrentLocationPushpin,
    });

})();

A couple of notes about the MapService module:

The mapbounds parameter is unnecessary for real service calls, it's only used by mock service so that it has a coordinate to generate valid service points inside the map.
Locaiton service uses native Geolocator to retrieve current location, and caches the data for future usage to improve the performace. The cache feature can turn off when instantiating the CurrentLocation object if that's desired.
The Bing Maps' default pushpin is a 25x35 size image pining at the middle bottom of the image. We need to use pushpin's anchor property to adjust the pin point if using custom images with different size or different pining location.
Infobox popup supports some HTML tags inside the message, e.g. <br /> line break and <b> bold fonts for "Address" are used in our example

UI Interaction

Now we are ready to refine the UI and implement the user interactions. First add mock service and map service javascript references in the default.html. Following is the screen-shot of the final solution inside Visual Studio:

In general there are three user actions: click current location button to search services around; move map to search nearby locations; type some keyword and click search button to do search name. The first two cases are basically the same as they both search by location. The updated default.js is listed below and it's quite self explanatory:

<script type="text/javascript">
(function () {
    "use strict";

    var map;            // Bing Map ojbect
    var currentLocPin;  // Pushpin for current location
    var locationServie; // Location serverice instance 
    var pushpinLayer;   // A layer to contain all pushpins
    var infoboxLayer;   // A layer to contain the infobox
    var resetMapCenter; // Reset the map center location if is true

    document.addEventListener("DOMContentLoaded", initialize, false);

    function initialize() {
        Microsoft.Maps.loadModule('Microsoft.Maps.Map', { callback: initMap });
        Microsoft.Maps.Events.addHandler(map, 'viewchangeend', mapViewChanged);

        btnCurrent.addEventListener("click", currentLocationClick);
        btnSearch.addEventListener("click", searchLocationClick);
        showProgress();
    }

    function initMap() { // Initialize Bing map
        try {
            var mapDiv = document.getElementById("mapDiv");
            var mapOptions = {
                credentials: "xxxxxxx",  //Read from your Bing Maps Account
                center: new Microsoft.Maps.Location(43.813, -79.344), //TORONTO
                zoom: 13,
            };
            map = new Microsoft.Maps.Map(mapDiv, mapOptions);

            // layers to maintain the pushpins and infobox popup
            pushpinLayer = new Microsoft.Maps.EntityCollection();
            infoboxLayer = new Microsoft.Maps.EntityCollection();
            map.entities.push(pushpinLayer);
            map.entities.push(infoboxLayer);
        }
        catch (err)  {
            showError("Init Map error", err);
        }
    }

    // Repopulate map when its moved
    function mapViewChanged() {
        resetMapCenter = false;
        if (map.getZoom() > 10) {          
            var mapCenter = map.getCenter();
            var parameters = { lat: mapCenter.latitude, long: mapCenter.longitude };
            searchLocationService(parameters);
        }
    }

    // Search by current location
    function currentLocationClick() { 
        resetMapCenter = true;
        if (!locationServie) {
            locationServie = new MapService.CurrentLocation();
        }
        locationServie.getCurrentLocationAsync().then(
            function (loc) {
                currentLocPin = MapService.getCurrentLocationPushpin(loc);
                searchLocationService({ lat: loc.latitude, long: loc.longitude });
            },
            function (err) {
                showError("Location service error", err);
            });
    }

    // Search servcie by name
    function searchLocationClick() { 
        resetMapCenter = true;
        var searchInput = txtSearch.value;
        if (searchInput) {
            var parameters = { query: escape(searchInput) };
            searchLocationService(parameters);
        }
    }

    // Invoke map service by MapService module
    function searchLocationService(searchParameter) { 
        showProgress();
        try {
            var mapBounds = map.getBounds(); // Map Bounds are only used by mock service
            MapService.invokeLocationService(searchParameter, handleSearchResult, handleSearchError, mapBounds);
        } catch (err) {
            showError("Error occurs", err);
        }
    }

    // Show error message dialog
    function showError(title, err) { 
        var msg =  (err && err.message) ? err.message : err;
        var msgBox = new Windows.UI.Popups.MessageDialog(msg, title);
        msgBox.showAsync();
    }

    // Handle search error
    function handleSearchError(err) {
        showError("Map Service error", err);
        hideProgress();
    }

    // Search Result handling call back
    function handleSearchResult(data) { 
        var serviceLocations = MapService.parseServiceLocations(data);

        if (resetMapCenter) {
            var searchCenter = MapService.parseSearchCenterLocation(data);
            var locations = [];
            for (var i = 0; i < serviceLocations.length; i++) {
                locations.push(serviceLocations[i].location);
            }
            var origCenter = map.getCenter();
            var origBound = map.getBounds();
            var mapBound = Microsoft.Maps.LocationRect.fromLocations(locations);
            map.setView({ center: searchCenter, bounds: mapBound });
        }

        updatePushpins(serviceLocations);

        hideProgress();
    }
    
    // Shwo progress
    function showProgress() {
        progressDiv.style.display = "block";
    }

    // Hide progress
    function hideProgress() {
        progressDiv.style.display = "none";
    }

    // Update pushpins on the map
    function updatePushpins(serviceLocations) { 
        infoboxLayer.clear();
        pushpinLayer.clear();

        for (var i = 0; i < serviceLocations.length; i++) {
            var loc = serviceLocations[i];
            var pin = MapService.getServicePointPushpin(loc);
            pin.name = loc.name;
            pin.summary = loc.summary;
            pushpinLayer.push(pin);
            Microsoft.Maps.Events.addHandler(pin, 'click', showInfobox);
        }

        if (currentLocPin) {
            pushpinLayer.push(currentLocPin);
        }
    }

    // Display infobox when pushpin is clicked
    function showInfobox(e) { 
        if (e.targetType == "pushpin") {
            infoboxLayer.clear();
            var pin = e.target;
            var pinLocation = pin.getLocation();
            var infoboxOptions = {
                showCloseButton: true,
                //offset: new Microsoft.Maps.Point(0, 5),
                showPointer: true,
                zIndex: 10,
                title: pin.name,
                description: pin.summary,
            };

            var infobox = new Microsoft.Maps.Infobox(pinLocation, infoboxOptions);
            infoboxLayer.push(infobox);
        }
    }
})();
</script>

The whole project can be downloaded from GitHub.

Wednesday, November 20, 2013

Scheduled Task to Add Colleagues in SharePoint

Here is the use case in SharePoint 2010/2013: HR wants all employees to add some key people in your organization as colleagues, so that everyone inside the company would be able to see their posts and feeds. A simple powershell script can do the job:

## Add SharePoint Snap-in
Add-PSSnapin Microsoft.SharePoint.PowerShell -ErrorAction SilentlyContinue

## Get profiles
$site = Get-SPSite("https://my_company_portal")
$context = Get-SPServiceContext $site
$profilemanager = New-object Microsoft.Office.Server.UserProfiles.UserProfileManager($context)
 
## People who will be the common colleagues for all employees
$vipNames = "domain\user1", "domain\user2", "domain\user3"
$vipMembers = @()
foreach($vipName in $vipNames) {
    try {      
        $vipUser = $profilemanager.GetUserProfile($ecName)
        $vipMembers += $vipUser
    } catch {
        write-output $_.Exception.Message
   }   
}
 
## Colleague setting
$group = "General"
$groupType = [microsoft.office.server.userprofiles.colleaguegrouptype]::General
$privacy = [microsoft.office.server.userprofiles.privacy]::Public
$isInWorkGroup = $false
 
## Go through all users and add colleagues
$profiles = $profilemanager.GetEnumerator()
foreach($user in $profiles) {
    foreach($vipMember in $vipMembers) {
        if (($user.ID -ne $vipMember.ID) -and !($user.Colleagues.IsColleague($vipMember[0].ID))) {
            try {      
                $user.Colleagues.CreateWithoutEmailNotification($vipMember, $groupType, $group, $isInWorkGroup, $privacy)
            } catch {
                write-output $_.Exception.Message
            }
        }
    }
}

How about the new hires? We can setup a scheduled task to run the powershell script so the profile update will be applied weekly or daily. Open Task Scheduler from Administrative Tools or Control Panel, create a Task with following settings:

General tab:
- Name: "Add colleagues"
- Make sure the account running the task having sufficient permission to update the SharePoint Profile database
- Check (enable) "Run whether user is logged on or not"
- Check (enable) "Run as highest privileges"
Triggers tab:
- Create a new trigger: Weekly or daily at midnight
Actions tab:
- Select "Start a program" type
- Program/script: powershell
- Add arguments (optional): -Command "C:\SharePointScheduledTasks\addColleagues.ps1"

In case you don't like powershell script, following code is the equivalent c# version doing the exact same task:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
 
using Microsoft.SharePoint;
using Microsoft.Office.Server;
using Microsoft.Office.Server.UserProfiles;
 
class Program
{
    static void Main(string[] args)
    {
        UpdateUsersProfile("https://my_company_portal");
    }
 
    private static void UpdateUsersProfile(string siteUrl)
    {
        using (SPSite spSite = new SPSite(siteUrl))
        {
            SPServiceContext context = SPServiceContext.GetContext(spSite);
            UserProfileManager profileManager = new UserProfileManager(context);
 
            List<UserProfile> vipMeMembers = GetVipMembers(profileManager);
 
            foreach (UserProfile profile in profileManager)
            {
                AddColleagues(profile, vipMeMembers);
            }
        }
    }
 
    private static List<UserProfile> GetVipMembers(UserProfileManager profileManager)
    {
        List<string> vipUserNames = new List<string>() { "domain\\user1", "domain\\user2", "domain\\user3" };
        List<UserProfile> vipMembers = new List<UserProfile>();
 
        foreach (string name in vipUserNames)
        {
            try
            {
                UserProfile vipMember = profileManager.GetUserProfile(name);
                vipMembers.Add(vipMember);
            }
            catch (Exception ex)
            {
                // User doesn't exit
            }
        }
        return vipMembers;
    }
 
    private static void AddColleagues(UserProfile user, List<UserProfile> colleagues)
    {
        foreach (UserProfile colleague in colleagues)
        {
            if (user.ID != colleague.ID && !user.Colleagues.IsColleague(colleague.ID))
            {
                user.Colleagues.CreateWithoutEmailNotification(colleague, 
                    ColleagueGroupType.General, "General", false, Privacy.Public);
            }
        }
    }
}

Thursday, November 07, 2013

A Simple Logging Server Implementation Using Node.js and MongoDB

In my previous post I built a logging & reporting module for Windows 8 store apps. In this post I am going to build a simple while fast and scalable logging server.

Open source node.js is used to serve as http logging server. Node.js is a JavaScript server-side solution. It's light, efficient, and using event-driven and non-blocking I/O. These features make it a good fit for logging JSON objects sent by the client.

Traditionally the logging service would just save logs to file system. However it's hard to do data analysis and poor performance to do query over big size of logged data if using plain text file logs. Alternatively we can use document type NoSQL as the data store. In our use case, each log is an independent information unit, and document stores works extremely well in such scenario. Data insert and query would be faster and more efficient than traditional SQL database. We pick MongoDB as it's easy to use in different platform, well documented and has native support to JSON data.

Node.js requires a driver or module to connect to MongoDB database. We use mongojs in our example:

npm install mongojs

The whole logging server implementation is less than 100-line of code:

var http = require('http');
var url = require('url');

var mongojs = require('mongojs');
var db = mongojs('Data');
var collection = db.collection('Logs');

var contentType = {'Content-Type': 'text/html'};
var OK = 'OK', BAD = 'BAD REQUEST';
var MAX_URL_LENGTH = 2048, MAX_POST_DATA = 10240;

function createLog(request) {
    var logObj = {};
    try {
        logObj.time = new Date();
        logObj.method = request.method;
        logObj.host = request.headers.host;
        logObj.url = request.url;
        var queryStrings = url.parse(request.url, true).query;
        // Check if it's an empty object
        if (Object.keys(queryStrings).length) {
            logObj.query = queryStrings;
            for (var attrname in queryStrings) { 
                logObj[attrname] = queryStrings[attrname]; 
            }
        }
    } catch (ex) {}
    return logObj;
}

http.createServer(function (request, response) {
    // Bypass request with URL larger than 2K
    if (request.url.length > MAX_URL_LENGTH) {
        response.writeHead(414, contentType);
        response.end(BAD);
    } else if (request.method == 'GET') {  // HTTP GET
        var logObj = createLog(request);
        if (logObj.query) {
            delete logObj.query;
            collection.save(logObj);
            response.writeHead(200, contentType);
            response.end(OK);
        } else {  // Empty request
            response.writeHead(400, contentType);
            response.end(BAD);
            request.connection.destroy();
        }
    } else if(request.method == 'POST') {  // HTTP POST
        var logObj = createLog(request);
        var postData = '';
        request.on('data', function(data) {
            postData += data;
            // Bypass request with POST data larger than 10K
            if(postData.length > MAX_POST_DATA) {
                postData = "";
                response.writeHead(413, contentType);
                response.end(BAD);
                request.connection.destroy();
            }
        });
        request.on('end', function() {
            if (!postData && !logObj.query) {  // Empty request
                response.writeHead(400, contentType);
                response.end(BAD);
                request.connection.destroy();
            } else {
                if (postData) {
                    try {
                        postObjs = JSON.parse(postData);
                        for (var attrname in postObjs) { 
                            logObj[attrname] = postObjs[attrname]; 
                        }    
                    } catch (ex) {
                        logObj.postData = postData;
                    }
                }
                if (logObj.query) {
                    delete logObj.query;
                }
                collection.save(logObj);
                response.writeHead(200, contentType);
                response.end(OK);
            } 
        });
    }
}).listen(process.env.PORT || 8080);

Friday, November 01, 2013

WinJSLog - A Logging and Reporting JavaScript Module For Windows 8 Store Apps

Background

Usually developers use console.log() or WinJS.Log() function to log error and debug information. But that only helps during debugging in local machine. How about those users who installed your app and had errors or experienced crashes? Microsoft has Quality reports in Windows 8 app's dashboard if its telemetry service is enabled, but that function is quite limited, and it only collects crash data when the user accepts the so called Customer Experience Improvement Program (CEIP). It would be great to have errors and custom data reported to your own server so you can do some investigation at real-time. In this post I will have some discussion on this topic and present an implementation to log and report crashes, errors, and potentially any info you want for a Windows 8 store app.

Why logging and reporting are important for mobile apps?

Is your Mobile app working nicely in different devices? Are there any unexpected issues happening in different environment? The customer feedback is likely your main source of investigation. But that's not something you as a developer should rely on. Some issues may never surface, and people would just uninstall the app when they are not happy on it. Unless for some royal users, no many people would spend their precious time to write feedback or report to help you to improve your application.

Even you are lucky enough to have a nice user having a polite comment like this: "This's a good app! However sometimes it crashes when I am on editing mode. Please fix it.", you have no idea what's the real problem is. You may end up spending countless time just to reproduce the issue in specific situation.

So you want to monitor your app's running issues in real-time, and you want to get as much info when crash or error occurs. When logging service is on board, you may want more data to analyze the customer usage. Then it becomes an analytics framework and you have another communication channel to your app (carefully not to abuse using it).

All that sound to be essential to maintain the app quality for a distributed mobile app. But surprisingly I couldn't find a public available solution online. So I decided to write my own, and I called it WinJSLog. The goal of the implementation is light, effective, and easy to use.

Why an app crashes?

I have discussed WinJS exceptions in a previous post. Basically the unhandled exception will result in app crash by default. Crashes is the worst user experience ever because the running app is suddenly disappear without any prompt and reason to end user. A better experience would be redirect the user to a landing page. People may get confused too "why I am switched to a different place?", but at least the app maintains the running condition, giving user a chance to do other stuff or retry the previous action.

Unhandled exception in Windows 8 can be caught in Windows.Application.onerror global event. Following code snippet shows how to register an event handler to redirect user to a specific page when unhandled exception occurs:

    Windows.Application.onerror = function (e) {
        showErrorPage();
        return true; // telling WinJS that the exception was handled, don't terminate
    }

What data to collect?

Errors like "Object reference is null" is not so helpful. As a development you want as much info as possible to identify an issue, such as the stack trace which provides more context about the error. The device and environment variance could make some issue super hard to figure out. A problem may only happen in one specific device, in one specific language or in one specific screen size.

My implementation includes the error message detail, as well as following context info:

application version
language
device maker
device model
screen size
suspend and resume history
page navigation history
method invocation history

Above context information could help us to pin point the issues occurring in certain environment. Ideally the memory usage info should be included but I have not figured out how to get that data from WinJS libraries yet.

What's the data format?

Data are collected as key-value JavaScript objects, and JSON is used to transfer the logged data to different media, like file system or cross the network. JSON was born for such usage and it's a total no-brainer. Following JSON data is an example of crash log sent from a Windows 8 test app running in a VMWare virtual machine:

{
    "os" : "Windows 8",
    "version" : "1.0.0.1",
    "manufacturer" : "VMware, Inc.",
    "model" : "VMware Virtual Platform",
    "lang" : "en-CA",
    "screen" : "1680x1050",
    "orientation" : "landscape",
    "timezone" : "-5", 
    "latitude" : "43.666698",
    "longitude" : "-79.416702",
    "logtime" : "November 1, 2013 9:42:56 AM", 
    "pagetrace" : "home[9:42:27AM]=>user[9:42:41AM]=>test[9:42:48AM]",
    "level" : "crash", 
    "log": [ 
            {
             "source" : "promise", 
             "message" : "'username' is undefined", 
             "description" : "'username' is undefined",
             "stacktrace" : "ReferenceError: 'username' is undefined\n 
                at ready (ms-appx://testApp/pages/test/test.js:85:13)\n 
                at Pages_ready (ms-appx://microsoft.winjs.1.0/js/base.js:4511:29)\n
                at notifySuccess (ms-appx://microsoft.winjs.1.0/js/base.js:1404:21)\n 
                at enter (ms-appx://microsoft.winjs.1.0/js/base.js:1091:21)\n
                at _run (ms-appx://microsoft.winjs.1.0/js/base.js:1307:17)\n
                at _completed (ms-appx://microsoft.winjs.1.0/js/base.js:1275:13)\n
                at notifySuccess (ms-appx://microsoft.winjs.1.0/js/base.js:1404:21)\n
                at enter (ms-appx://microsoft.winjs.1.0/js/base.js:1091:21)\n
                at _run (ms-appx://microsoft.winjs.1.0/js/base.js:1307:17)\n
                at _completed (ms-appx://microsoft.winjs.1.0/js/base.js:1275:13)", 
             "level" : "crash", 
             "time":"2013-11-01T14:42:56.438Z"
           }
        ]
}

In next post I will present a simple logging server implementation to handle such requests.

When to send the logged data? What happen if no Internet access?

Crash or unhandled exception is considered serious issue, and the error detail should be sent to the logging server immediately when it happens. For regular error messages or debugging info we could accumulate them and send them out once current page is completed or a different page is visited. There's also a timer process to check and cleanup the pending logs in background (2 minutes by default and is configurable), so the logged message could be delivered even the user stay on one page for a long time.

The logged data will be stored to file system if Internet service is not available. When the application starts up, it will check if such log file(s) exist, if so the log file(s) will be read and send to logging server if Internet is available. Usually there're many operations during the application booting up. To avoid the action congestion, it's a good idea to defer to run the check-and-send action a little bit later than the application launch (30 seconds by default). If the Internet access is not available, the background process will also periodically recheck the network status (2 minutes by default) and send out the file logs when Internet access becomes available.

Other consideration

You app may target to some specific features such as map and camera stuff, and the location and camera info would be good for you to diagnose the issue. In such case, the location and camera data should be obtained right in the spot when they are used, and store them in a global variable like sessionState, so the logging code can read it directly without a redundant code executed again, which makes the logging service lighter and less intrusive.

What happens if logging server requires authentication? As I discussed in WinJS Authentication, it would be better to allow anonymous access for such general logging service. Hard-coded logging username/password is not safe, and it's not feasible to let user to type the username and password. Would it be a potential DOS/DDOS attach when anonymous access is open? No service in Internet can avoid that, but firewall in front of your servers should take care of that threat. However we can still do a bit more to protect ourself in the code level, such as not logging abnormal data with extremely large size of URL or message.

How about the errors inside logging service? In general any errors or potential errors within logging and reporting code should be silently skipped. Optionally those errors can be sent to the logging server, or send to a different server if errors occur during sending data to logging server.

How to use the code?

The WinJSLog module exposes following functions:

    function fatal(error)                // log crash
    function error(description, error)   // log error
    function warning(error, description) // log warning
    function page(pageName)              // navigate to a page
    function method(methodName)          // invoke a method
    function registerLogging()           // register logging
    function unregisterLogging()         // unregister logging

First include the winjslog.js file inside default.html or your custom page, then register the logging service before app.start():

(function () {
    "use strict";

    WinJS.Binding.optimizeBindingReferences = true;

    var app = WinJS.Application;
    var nav = WinJS.Navigation;

    app.addEventListener("activated", function (args) {
      ...// skip for abreviation
    }); 
   
    app.onerror = function (e) {
        Logging.fatal(e);
        Logging.page("homepage");
        nav.navigate("/pages/home/home.html");
        
        return true; // the app will terminate if false
    };

    Logging.registerLogging("http://myloggingserver/");
    
    app.start();

    ...

In above code an onerror event handler is registered to catch the unhandled exception to avoid the app's crash, log the crash by Logging.fatal(e), then redirect to home page. Note that the action of redirecting to home page is also logged as a navigation path by Logging.page() function. Following code snippet shows how to log a function invocation and its potential error message:

    function doComplicatedTask() {
        Logging.method("doComplicatedTask");
        try {
            ... // job with potential error
        } catch (e) {
            logging.error("Error occur!", e);
        }
    }

Alternatively you can log lower level warning, info and debug message by using Logging.warning(), Logging.info() and logging.debug() functions respectively.

Source code of WinJSLog.js

Update: the source code now is available at github and you can get the latest version there.

(function () {
    "use strict";

    var logs = [], pages = [], methods = []; // log data
    var loggingServer, loggingEnabled, debugEnabled; // service setting
    var version, manufacturer, model; // app context
    var currentLevel = 'debug', levels = { 'debug': 1, 'info': 2, 'warning': 3, 'error': 4, 'crash': 5 };

    // register logging service
    function registerLogging(serverUrl, doDebugging, deferRunInSeconds, recheckInSeconds) {
        if (!serverUrl) {
            throw 'registerLogging error: logging server not defined.';
        }

        loggingEnabled = true;
        loggingServer = serverUrl;
        debugEnabled = doDebugging || false;
        deferRunInSeconds = deferRunInSeconds || 30;
        recheckInSeconds = recheckInSeconds || 60;

        // Register event handler for suspend, resume and relaunch (terminated then restarted)
        WinJS.Application.addEventListener("activated", restoreSessionData);
        Windows.UI.WebUI.WebUIApplication.addEventListener("suspending", suspend);
        Windows.UI.WebUI.WebUIApplication.addEventListener("resuming", resume);

        // defer to run the check-log-file-and-send process
        WinJS.Promise.timeout(deferRunInSeconds * 1000).then(function () {
            cleanupLogs(recheckInSeconds);
        });
    }

    // unregister logging service
    function unregisterLogging() {
        loggingEnabled = false;
        WinJS.Application.removeEventListener("activated", restoreSessionData);
        Windows.UI.WebUI.WebUIApplication.removeEventListener("suspending", suspend);
        Windows.UI.WebUI.WebUIApplication.removeEventListener("resuming", resume);
    }

    // restore logging data after app relaunch from terminated state
    function restoreSessionData(args) {
        if (args && args.detail && args.detail.kind === Windows.ApplicationModel.Activation.ActivationKind.launch) {
            if (args.detail.previousExecutionState === Windows.ApplicationModel.Activation.ApplicationExecutionState.terminated) {
                var sessionState = WinJS.Application.sessionState;
                if (sessionState.logPages)
                    pages = sessionState.logPages;
                if (sessionState.logMethods)
                    methods = sessionState.logMethods;
                if (sessionState.logLogs)
                    logs = sessionState.logLogs;
                if (sessionState.logLevel)
                    currentLevel = sessionState.logLevel;
            }
        }
    }

    // log suspending event and store log data into session
    function suspend() {
        if (loggingEnabled) {
            var pageEntry = { 'time': new Date(), 'page': 'suspending' };
            pages.push(pageEntry);
            var sessionState = WinJS.Application.sessionState;
            sessionState.logPages = pages;
            sessionState.logMethods = methods;
            sessionState.logLogs = logs;
            sessionState.logLevel = currentLevel;
        }
    }

    // log resuming event
    function resume() {
        if (loggingEnabled) {
            var pageEntry = { 'time': new Date(), 'page': 'resuming' };
            pages.push(pageEntry);
        }
    }

    // log an event: navigate to a page
    function pageLog(pageId) {
        if (loggingEnabled) {
            try {
                processMemoryLogs();
            } catch (e) { }
            var pageEntry = { 'time': new Date(), 'page': pageId };
            pages.push(pageEntry);
        }
    }

    // log an event: invoke a method
    function methodLog(methodName) {
        if (loggingEnabled) {
            methods.push(methodName);
        }
    }

    // log a crash; a crash or unhandled exception can be caught by WinJS.Application.onerror event
    function crashLog(err) {
        if (loggingEnabled) {
            setLevel('crash');
            var errWrapper = getErrorObject(err);
            errWrapper.level = "crash";
            errWrapper.time = new Date();
            logs.push(errWrapper);
            try {
                processMemoryLogs();
            } catch (e) { }
        }
    }

    // log an error
    function errorLog(description, err) {
        if (loggingEnabled) {
            setLevel('error');
            logs.push(getLogObject('error', description, err));
        }
    }

    // log a warning message
    function warningLog(description, err) {
        if (loggingEnabled && debugEnabled) {
            setLevel('warning');
            logs.push(getLogObject('warning', description, err));
        }
    }

    // log an info message
    function infoLog(description) {
        if (loggingEnabled && debugEnabled) {
            setLevel('info');
            logs.push(getLogObject('info', description));
        }
    }

    // log a debug message
    function debugLog(description) {
        if (loggingEnabled && debugEnabled) {
            setLevel('debug');
            logs.push(getLogObject('debug', description));
        }
    }

    // build a log object
    function getLogObject(level, description, err) {
        var logObject = getErrorObject(err);
        if (logObject.description) {
            logObject.description = logObject.description + description;
        } else {
            logObject.description = description || '';
        }
        logObject.level = level || 'unknown';
        logObject.time = new Date();
        return logObject;
    }

    // build an error object
    function getErrorObject(err) {
        var errObject = {};
        if (err) {
            if (err.detail && typeof err.detail === 'object') {
                var detail = err.detail;
                if (detail.promise) {
                    errObject.source = "promise";
                }
                if (detail.errorMessage) {
                    errObject.message = detail.errorMessage;
                    if (detail.errorLine)
                        errObject.codeline = detail.errorLine;
                    if (detail.errorUrl)
                        errObject.sourcUrl = detail.errorUrl;
                } else if (detail.error && typeof detail.error === 'object') {
                    errObject.message = detail.error.message || 'unknown';
                    if (detail.error.description)
                        errObject.description = detail.error.description;
                    if (detail.error.stack)
                        errObject.stacktrace = detail.error.stack;
                } else {
                    errObject.message = detail.message || 'unknown';
                    if (detail.description)
                        errObject.description = detail.description;
                    if (detail.number)
                        errObject.codeline = detail.number;
                    if (detail.stack)
                        errObject.stacktrace = detail.stack;
                }
            } else {
                errObject.message = err.message || err.exception || err;
            }
        }
        return errObject;
    }

    // determine the highest log level for current log entry
    function setLevel(level) {
        if (levels[level] > levels[currentLevel]) {
            currentLevel = level;
        }
    }

    // periodically check the memory logs and storage logs, and send logs to server if Internet is available
    function cleanupLogs(recheckInseonds) {
        if (loggingEnabled) {
            processMemoryLogs();
            processFileLogs();
            setTimeout(function () {
                cleanupLogs(recheckInseonds);
            }, recheckInseonds * 1000);
        }
    }

    // construct log message and send to server if Internet is available, otherwise save it to local storage
    function processMemoryLogs() {
        if (logs.length > 0) {
            var data = getContext();
            var date = new Date();
            data.logtime = date.toLocaleString() + ' [' + date.toISOString() + ']';
            if (pages.length > 0) {
                var pagetrace = pages.map(function (item) {
                    if (item.time && item.time.toLocaleTimeString)
                        return item.page + "[" + item.time.toLocaleTimeString().replace(' ', '') + ']';
                    else
                        return item.page + "[" + item.time + ']';
                }).join(' => ');
                data.pagetrace = pagetrace;
            }
            if (methods.length > 0) {
                data.methodtrace = methods.join(' => ');
            }
            data.level = currentLevel;
            data.log = logs.slice(0); //(logs.length == 1) ? logs[0] : logs.slice(0);

            if (isConnectedToInternet()) {
                sendLogsToServer(JSON.stringify(data));
            } else {
                saveLogsToFile(data);
            }
        }

        // clean up the logs
        methods = [];
        logs = [];
        currentLevel = 'debug';
    }

    // read all saved log files and send them to server if Internet is available
    function processFileLogs() {
        if (isConnectedToInternet()) {
            var localFolder = Windows.Storage.ApplicationData.current.localFolder;
            localFolder.getFilesAsync().then(function (files) {
                files.forEach(function (file) {
                    if (file && file.displayName && file.displayName.indexOf("logs") == 0) {
                        Windows.Storage.FileIO.readTextAsync(file).then(function (text) {
                            sendLogsToServer(text);
                        }).then(function () {
                            file.deleteAsync();
                        }).done(function () { }, function (err) { });
                    }
                });
            });
        }
    }

    // save a log entry to file system if Internet is not available
    function saveLogsToFile(obj) {
        var fileName = "logs.txt";
        var content = JSON.stringify(obj);
        var localFolder = Windows.Storage.ApplicationData.current.localFolder;
        var saveOption = Windows.Storage.CreationCollisionOption;
        localFolder.createFileAsync(fileName, saveOption.generateUniqueName).then(
            function (file) {
                return Windows.Storage.FileIO.writeTextAsync(file, content);
            }).done(function () {
                console.log("Log saved");
            }, function (error) {
                console.log("Log saved error");
            });
    }

    // send log message to logging server
    function sendLogsToServer(jsonData) {
        WinJS.xhr({
            type: "post",
            url: loggingServer,
            headers: { "Content-type": "application/json" },
            data: jsonData
        }).done(function completed(c) {
            console.log("log sent");
        },
        function error(e) { // One more try? send to different server? or silently skip?
            console.log("log sent error");
        });
    }

    // get current application context
    function getContext() {
        if (!version) {
            var appVersion = Windows.ApplicationModel.Package.current.id.version;
            version = appVersion.major + "." + appVersion.minor + "." + appVersion.build + "." + appVersion.revision;
            try {
                var deviceInfo = new Windows.Security.ExchangeActiveSyncProvisioning.EasClientDeviceInformation();
                manufacturer = deviceInfo.systemManufacturer;
                model = deviceInfo.systemProductName;
            } catch (e) {
                manufacturer = 'unknown';
                model = 'unknown';
            }
        }
        var context = {};
        context.version = version;
        context.manufacturer = manufacturer;
        context.model = model;
        context.os = "Windows 8";
        context.lang = navigator.appName == "Netscape" ? navigator.language : navigator.userLanguage;
        context.screen = screen.width + "x" + screen.height;
        context.orientation = getOrientation();
        context.timezone = (-(new Date()).getTimezoneOffset() / 60).toString();
        return context;
    }

    // determine current orientation
    function getOrientation() {
        var orientation = "unknown";
        switch (Windows.Graphics.Display.DisplayProperties.currentOrientation) {
            case Windows.Graphics.Display.DisplayOrientations.landscape:
                orientation = "landscape";
                break;
            case Windows.Graphics.Display.DisplayOrientations.portrait:
                orientation = "portrait";
                break;
            case Windows.Graphics.Display.DisplayOrientations.landscapeFlipped:
                orientation = "landscapeFlipped";
                break;
            case Windows.Graphics.Display.DisplayOrientations.portraitFlipped:
                orientation = "portraitFlipped";
                break;
        }
        return orientation;
    }

    // check if Internet access is available
    function isConnectedToInternet() {
        var connectivity = Windows.Networking.Connectivity;
        var profile = connectivity.NetworkInformation.getInternetConnectionProfile();
        if (profile) {
            var connected = (profile.getNetworkConnectivityLevel() == connectivity.NetworkConnectivityLevel.internetAccess);
            return connected;
        } else {
            return false;
        }
    }

    WinJS.Namespace.define("Logging", {
        registerLogging: registerLogging,
        unregisterLogging: unregisterLogging,
        page: pageLog,
        method: methodLog,
        fatal: crashLog,
        error: errorLog,
        warning: warningLog,
        info: infoLog,
        debug: debugLog
    });
})();

Friday, September 06, 2013

Page Buttons Not Responding After Getting PDF in SharePoint

Today I have been working on a SharePoint 2010 WebPart in which the user can click a button to get a PDF report. Open source iTextSharp library is used to generate the PDF report. The code is quite straightforward:

        void GeneratePDF()
        {
            Document doc = new Document(PageSize.A4);
            MemoryStream pdfStream = new MemoryStream();
            PdfWriter.GetInstance(doc, pdfStream);
            doc.Open();
            // Populate document with business data
            doc.Close();

            Response.Clear();
            Response.ClearHeaders();
            Response.ContentType = "application/pdf";
            Response.AddHeader("Content-Disposition", "attachment;filename=report.pdf");
            Response.BinaryWrite(pdfStream.ToArray());
            Response.Flush();
            Response.End();   
        }

That PDF function works fine, but all other buttons on the same page are not responding (postback doesn't occur) after the PDF button is click. Such behavior only happens in SharePoint environment and everything is okay in a regular ASP.NET page. It looks like some special validation in SharePoint causing the problem. I debugged into the JavaScript and found the setting of "__spFormOnSubmitCalled" variable is the culprit.

ASP.NET validation process triggered by the click of a button includes invocation of JavaScript function called WebForm_OnSubmit. SharePoint overrides this function for each page:

<script type="text/javascript">
//<![CDATA[
    if (typeof(Sys) === 'undefined') 
        throw new Error('ASP.NET Ajax client-side framework failed to load.');
    if (typeof(DeferWebFormInitCallback) == 'function') 
        DeferWebFormInitCallback();
    function WebForm_OnSubmit() {
        UpdateFormDigest('webUrl..', 1440000);
        if (typeof(vwpcm) != 'undefined') {
            vwpcm.SetWpcmVal();
        };
        return _spFormOnSubmitWrapper();
    }
//]]>
</script>

The JavaScript function __spFormOnSubmitWrapper is defined in /_layouts/1033/init.js:

function _spFormOnSubmitWrapper() {
    if (_spSuppressFormOnSubmitWrapper)
    {
        return true;
    }
    if (_spFormOnSubmitCalled)
    {
        return false;
    }
    if (typeof(_spFormOnSubmit) == "function")
    {
        var retval = _spFormOnSubmit();
        var testval = false;
        if (typeof(retval) == typeof(testval) && retval == testval)
        {
            return false;
        }
    }
    _spFormOnSubmitCalled=true;
    return true;
}

The "_spFormOnSubmitCalled" field is false by default when the page is loaded. It's set to true when you click a button on the page. This machanism ensures only the first button click will take action and prevents other clicks from posting back to the server. The "_spFormOnSubmitCalled" field is reset to false once the page is reloaded. A postback will usually result in a page reloading, but not in above PDF out case where the server writes the PDF attachment to the client then ends the interaction. So the "_spFormOnSubmitCalled" field remains true which blocks any future postback.

So theoretically the issue is not limited to PDF output. Directly writing and ending on the Response object in the server side would result in the same problem. There're a few approaches to resolve the problem:

1. Reset "_spFormOnSubmitCalled" to false after the PDF button is clicked. Note that the reset timing is important, and it must be later then the submission process (after the WebForm_OnSubmit method is called), for example:
```
   function resetSharePointSubmitField() {
        setTimeout(function () { _spFormOnSubmitCalled = false; }, 1000); // set the field after 1 second
        return true;
    }
```

2. Overide the WebForm_OnSubmit function and make it always return true:

    function resetSharePointSubmitField() {
        window.WebForm_OnSubmit = function() {return true;};
    }

Apply the JavaScript to a button:

    <asp:Button ID="btnGeneratePDF" runat="server" Text="Get PDF" OnClientClick="resetSharePointSubmitField();" />

The other option is add a client side script "JavaScript: _spFormOnSubmitCalled = false;" for all buttons on the page, but that is not scalable and not recommended.

Bonus tip The regular pdf export function by Response.Write() won't work inside a modal dialog (the popup window open by window.showModalDialog() from JavaScript). To resolve this particular problem you can replace the PDF export button with a hyper link and set its target to an empty iframe:

    <a id="popupPDF" target="pdfTarget" runat="server">Get PDF</a>
    <iframe name="pdfTarget" id="pdfTarget" width="0" height="0" style="display:none;"></iframe>

Then simply assign a query string the anchor which tells the server to generate the PDF:

    protected void Page_Load(object sender, EventArgs e)
    {
        popupPDF.HRef = Request.Url.AbsoluteUri + "&pdf=true";
        if (Request.QueryString["pdf"] == "true")
            GeneratePDF();
    }

Thursday, June 13, 2013

More on Cross-domain Authentication

In my previous post I mentioned that it's impossible to do NTLM or Kerberos authentication with client side JavaScript in cross-domain requests. We know username and password can be used in JavaScript Ajax calls for BASIC authentication, but not for NTLM or Kerberos. However my above statement is not fully true since on top of JavaScript the browser could do some magic for you. Thanks my colleague Antonio for correcting me on this. He has a test environment with Kerberos setup where a test user logged on domainA can visit a page from http://companyB.com in domainB, and inside that page JavaScript Ajax calls successfully grab data from http://companyA.com in domainA using JSONP and Kerberos authentication without user typing in username and password. What he did is simply add http://companyA.com to Local Intranet zone in IE, then IE will automatically logon http://companyA.com with Kerberos authentication seamlessly.

In this post I will explore some common scenarios related to browser or AJAX authentication, including Intranet and Internet environment. Note that for those public sites and services that allow anonymous access the cross-domain authentication issue doesn't exist at all.

1. The browser is in the same domain as the server (Windows Authentication).

If the user is in the same domain as the server where the cross-domain requests go to, such as above Kerberos case, the single-sign-on (SSO) would work with NTLM or Kerberos authentication, i.e. IE would auto-authenticate to the server with the current logged in Windows account. No user input is required in this case but the server has to be added to the Local Intranet zone. The key point is the the "Automatic logon only in intranet zone" setting in Local Intranet zone:

How about other browsers? Chrome shares the same network configuration with IE, so the IE settings will apply to Chrome and the auto-logon would just work fine in Chrome. In Firefox you need to open up the system configuration by typing "about:config" in the address bar, then double click the "network.automatic-ntlm-auth.trusted-uris" entry and set the domain names you want to auto-authenticate to, e.g. "google.com, microsoft.com" (not in "*.google.com" format which is used in IE trusted sites).

If the authentication failed, the browser will popup a login window to let you input your username and password.

2. The browser is not in the same domain as the server (Windows Authentication).

If the user is not in the same domain as the server where the cross-domain requests go to, e.g. a user logs into domainA and accesses a web page in which some JavaScript Ajax calls to http://companyB.com in domainB, then the single-sign-on won't work because domainA users can not be authenticated by domainB so a login window will pop up for user to input the username and password. In IE you have the option to save the password in this case:

After you select to save your password then next time IE will automatically authenticate for you (no need to input username and password again) on the same site if the "Automatic logon with current user name and password" is configured for the zone:

You can only save the password in IE but not in Chrome or Firefox. When the password is saved the Chrome will have the same behavior as IE, and you need to configure the same "network.automatic-ntlm-auth.trusted-uris" setting to let Firefox do the auto-authentication.

Where the password is saved by IE? They are stored in Windows secure storage. You can see and update those saved usernames and passwords by clicking "Control Panel => User Account => Advanced => Manage Password" or run "control keymgr.dll" command:

Just like the previous case you will also see a browser login window if the authentication failed.

3. Server is Form authentication.

In Form authentication you submit your username and password to the server. The server will issue an authentication cookie to the client if username and password are valid, then you are authenticated and logged on. The interactive steps are purely HTML and all browsers have the same behavior. After logging in the authentication cookie will be sent by browser in the consequent calls to the same server with same URI domain. The auto-authentication of AJAX calls will be okay if the cookie is valid which means that you have logged in to the server before the AJAX call. In some login forms you can select to "keep me logged in" option where a persistence cookie is generated for long time usage so you can auto-login next time even you close and reopen the browser.

Not like Windows authentication, browser will not popup a login window when authentication failed in AJAX calls, instead any error is silently swallowed by the browser by default. In Form authentication the server will respond to the client to redirect to the login page which is not recognized by AJAX call. Of course you can handle that specific 302 redirect case and popup a login window using JavaScript but that's another story.

4. How about OpenID, SAML, OAuth, Claimed based authentication, Federated authentication...?

In most cases the behavior is the same as Form authentication: access would be okay if you have signed in before the AJAX calls. This is valid as long as the server is using cookie to trace the authentication status, such as Windows and Form authentication discussed above, otherwise the authentication immediately fails. How to know if the server is using authentication cookie? You can check if you are able to access the site assets after successfully authenticated through the browser.

There're some cases where each service request requires authentication. The AJAX call will fail in such case unless you provide required data. One example is the new Twitter API V1.1: every Twitter API service call now requires OAuth authentication. Setting OAuth header using JavaScript directly is impractical and unsafe (however you could calculate the OAuth signature from the web server and JavaScript simply pass it to the Twitter service).

Also like Form authentication, you won't see any popup window whenever the authentication fails, and the AJAX calls are simply failed.

5. What about SSL, SSL client authentication?

SSL, also known as https protocol, encrypts the data in transport layer so the communication between the client and the server is protected. SSL itself doesn't do anything for Authentication. However server certificate is required to establish a SSL connection. The AJAX calls will fail if the server certificate is invalid such as a self-signed certificate. Browser will display a warning message if you visit the server with invalid certificate directly (not from AJAX call):

In order to get rid of the warning message, you could install the certificate and put it into Trusted Root Certification Authorities for IE and Chrome (caution of doing so) or import it to Firefox. After that you can visit the site without seeing the warning message page, and the AJAX calls now will be able to go through to the server for authentication.

SSL client authentication is another story. Basically SSL server also asks client's certificate when setting up the SSL connection. The SSL client authentication can be configured as optional or required. If client authentication is required the SSL connection will fail if browser doesn't provide a certificate, or the certificate is invalid, or the certificate is not accepted. If the SSL client authentication is okay, the SSL connection will be established and the regular server side authentication will come to play in the next step.

In summary, an indicator of a successful AJAX call is that you can browse the resource or service directly without any issue. The statement is okay for same-domain AJAX calls, but that's not enough in cross-domain scenarios where you can visit the site successfully in the browser but the cross-domain AJAX calls fail (such failure is not caused by client-server authentication). Cross-domain AJAX communication requires special implementation in both client side and server side, for example, you can define {crossDomain: true, dataType: 'jsonp'} parameters for jQuery AJAX calls in the client side, and implement JSONP or CORS in the server side. Following html page is created to test the cross-domain request using jQuery. Simply save the html page in your local machine, open the page and run the test:

<html xmlns="http://www.w3.org/1999/xhtml">
<head>
    <meta http-equiv="content-type" content="text/html; charset=utf-8" />          
    <title>Cross-domain AJAX test</title>          
    <script src="http://ajax.googleapis.com/ajax/libs/jquery/1.9.1/jquery.min.js" ></script>
</head>     
<body>         
    <div>
        <div style="text-align:center">
            <h3>Cross-domain AJAX test</h3>
            Server URL:  <input name="txtUrl" type="text" id="txtUrl" style="width:500px;" />
            <button id="btnTest">Send cross-domain request</button>
        </div>
        <div id="result"></div>
    </div>
    <script type="text/javascript">
        $(document).ready(function() {
            $("#btnTest").click(function () {
                testRequest($("#txtUrl").val());    
            });
        });
        
        function testRequest(url) {
            $("#result").html("");
            if (url) {
                $.ajax({url: url, cache: false, crossDomain: true, dataType: 'jsonp'})
                .done(function(data) {
                    var value = getDataString(data);//decodeURIComponent($.param(myObject));
                    $("#result").html("Response:<pre>" + value + "</pre>");})
                .fail(function(xhr, err) {
                    $("#result").html("Error occurs.");
                });    
            } else {
                $("#result").html("Invalid Url.");
            }
        }
        
        // http://www.sitepoint.com/javascript-json-serialization/
        function  getDataString(obj) { 
            var t = typeof (obj);
            if (t != "object" || obj === null) {
                // simple data type
                if (t == "string") 
                    obj = '"'+obj.replace(/>/g,'&gt;').replace(/</g,'&lt;')+'"';
                return String(obj);
            } else {
                // recurse array or object
                var n, v, json = [], arr = (obj && obj.constructor == Array);
                for (n in obj) {
                    v = obj[n]; t = typeof(v);
                    if (t == "string") 
                        v = '"'+v+'"';
                    else if (t == "object" && v !== null) 
                        v = getDataString(v);
                    json.push((arr ? "" : '"' + n + '":') + String(v));
                }
                return (arr ? "[" : "{") + String(json) + (arr ? "]" : "}");
            }
        };
    </script>
</body>
</html>