This can be fustrating when querying against RSS feeds or JSON/XML web APIs that don’t define any of these files. The workaround for this issue is to use some sort of proxying service. In this article I’ll be showing how to use Google App Engine to create a simple proxy that will forward these requests for free – within a reasonable daily load.

Google App Engine Overview

The reason I’ve chose to implement the proxy using Google App Engine is because it has a free daily quota and getting started using it is really simple: all you need is a Google account and to download and install the Google App Engine SDK.

Google App Engine supports developing in both Java and Python. In my example I’ll be using Python. In order to use and deploy the code yourself as well, follow these steps:

• If you don’t yet have a Google account, register for one.
• Download and install Python from python.org
• Download and install the Google App Engine Python SDK
• Register an application on Google App Engine (click on Create Application). When done, run the Google App Engine Launcher and create an application with the same name you’ve just created.

Creating the Proxy In Python

Creating a simple proxy is pretty straightforward by using the urlfetch library:

try:
	response = urlfetch.fetch(url)
except (urlfetch.Error, apiproxy_errors.Error):
	# an error occured

In order to make the proxy a bit smarter, I’ll implement some caching using the App Engine’s memcache:

proxiedContent = memcache.get(memcacheKey)
proxiedContentInMemcache = True
if proxiedContent is None:
   # not in memcache: execute request
   # ...
   # add the result content to memcache for CACHE_TIME minutes
   memcache.add(memcacheKey,proxiedContent,CACHE_TIME)
else:
   # use the content from memcache

Using these snipplets, here’s how the main file of the proxy webapp will look like:

import datetime
import hashlib
import logging
import pickle
import urllib
import re
import time
import urllib
import wsgiref.handlers
 
from google.appengine.api import memcache
from google.appengine.api import urlfetch
from google.appengine.ext import db
from google.appengine.ext import webapp
from google.appengine.ext.webapp import template
from google.appengine.runtime import apiproxy_errors
 
CACHE_TIME = 1 # number of minutes to cache content for
 
URL_PREFIXES = ["http://www.google.com/finance"] # only allow URLs to be queried from certain domain(s)
 
def getMemcacheKey(url):
  url_hash = hashlib.sha256()
  url_hash.update(url)
  return "hash_" + url_hash.hexdigest()
 
class ProxyHandler(webapp.RequestHandler):
  def get(self):
    url = self.request.get('url')
    url = urllib.unquote(url)
    # only allow urls that start with prefixes defined in URL_PREFIXES to be used
    if not self.isUrlAllowed(url):
      self.response.out.write("The URL passed can not be proxied due to security reasons.")
      return
    memcacheKey = getMemcacheKey(url) 
 
    # Use memcache to store the request for CACHE_TIME
    proxiedContent = memcache.get(memcacheKey)
    proxiedContentInMemcache = True
    if proxiedContent is None:
      proxiedContentInMemcache = False
      try:
        response = urlfetch.fetch(url)
      except (urlfetch.Error, apiproxy_errors.Error):
        return self.error(404)
      proxiedContent = response.content
    if proxiedContent is None:
      return self.error(404)
 
    # Add the fetched content to memcache
    if (not proxiedContentInMemcache):
      memcache.add(memcacheKey,proxiedContent,CACHE_TIME)
    self.response.out.write(proxiedContent)
 
def isUrlAllowed(self, url):
  for urlPrefix in URL_PREFIXES:
    if url.startswith(urlPrefix):
      return True
  return False
 
app = webapp.WSGIApplication([
  ("/proxy", ProxyHandler),
], debug=True)
 
def main():
  wsgiref.handlers.CGIHandler().run(app)
 
if __name__ == "__main__":
  main()

The proxy can now be called by passing the URL to be proxied via the url parameter. So to proxy e.g. http://google.com, the following request should be made: http://myapplication.appspot.com/proxy?url=http://google.com.

Adding crossdomain.xml to the Application

Now that the proxy class is ready, all we need to do is wire it into the web application and include a crossdomain.xml static file which will allow requests from all hosts (it is advisable to change this to the URL the requests are actually made). Based on this, here is how the crossdomain.xml would look like:

<!DOCTYPE cross-domain-policy SYSTEM "http://www.macromedia.com/xml/dtds/cross-domain-policy.dtd"> 
<cross-domain-policy> 
  <allow-access-from domain="*" /> 
</cross-domain-policy>

And here is how the modified app.yaml descriptior will look like:

application: yourappname
version: proxyv1
runtime: python
api_version: 1

handlers:
- url: /crossdomain.xml
  static_files: crossdomain.xml
  upload: crossdomain.xml
- url: /.*
  script: proxy.py
  secure: optional

Download the Code

The code for the application can be downloaded here: Proxy Using Google App Engine.zip. After unzipping be sure to change the “yourappname” name in app.yaml in order to be able to deploy it on Google App Engine.

Update: Security Issues

Andrew noted in the comment that the original solution raised security concerns as it would have been easy for someone to hijack this proxy and use it for their purpose as no authentication or authorization is done. I’ve added a small fix where the proxy only forwards to URLs that start from a list of prefixes. This is probably a sufficient solution for common cases, however more sophisticated authorization methods may be needed in other cases.

The Problem

The type of the StrokeDashArray (that is defined on Shape) is a DoubleCollection. So naturally I tried setting a DoubleCollection instance as the Value of the Setter on the Style I created:

Style ellipse2Style = new Style(typeof(Ellipse));
// Setting a DoubleCollection as the Value of the Setter for StrokeDashArrayProperty doesn't work!
ellipse2Style.Setters.Add(new Setter(Ellipse.StrokeDashArrayProperty,new DoubleCollection() { 2, 2 }));
Ellipse2.Style = ellipse2Style;

This is interesting enough as setting the StrokeDashArray property directly seems to work just fine:

// Works fine
Ellipse1.StrokeDashArray = new DoubleCollection() { 2, 2 };

This issue definitely suggests that this is a bug that is still present in Silverlight 4. A little searching revealed that this issue has been reported in version 2 (almost 2 years ago!), but is still not fixed.

The Workaround

After some playing around I’ve found a workaround: the Setter property needs to be passed as a string, rather then a DoubleCollection.

The fact that this helps solve the issue is quite surprising as it suggests that there is some logic built in to the styling mechanism that in case of the StrokeDashArray property does some funky conversion from string to DoubleCollection. Personally I would’ve thought that only the XAML parser would do something similar, but it looks like this isn’t a case.

So the (strange) workaround is to set the value of the Setter for the StrokeDashArrayProperty to a string, just like one would do when declaring the Style in XAML. The above example can be made working using this code:

Style ellipse3Style = new Style(typeof(Ellipse));
// Setting a string as the Value of the Setter for StrokeDashArrayProperty works - odd!
ellipse3Style.Setters.Add(new Setter(Ellipse.StrokeDashArrayProperty, "2,2"));
Ellipse3.Style = ellipse3Style;

As for a demo of the bug, see the following application:

The code behind for the application is as follows:

// Works fine
Ellipse1.StrokeDashArray = new DoubleCollection() { 2, 2 };
 
Style ellipse2Style = new Style(typeof(Ellipse));
// Setting a DoubleCollection as the Value of the Setter for StrokeDashArrayProperty doesn't work!
ellipse2Style.Setters.Add(new Setter(Ellipse.StrokeDashArrayProperty,new DoubleCollection() { 2, 2 }));
Ellipse2.Style = ellipse2Style;
 
Style ellipse3Style = new Style(typeof(Ellipse));
// Setting a string as the Value of the Setter for StrokeDashArrayProperty works - odd!
ellipse3Style.Setters.Add(new Setter(Ellipse.StrokeDashArrayProperty, "2,2"));
Ellipse3.Style = ellipse3Style;

Download the source of the example here: StrokeDashArray Bug in Silverlight.zip

Hope this helps anyone coming across this issue!

• Pinning variables when debugging
• Box selection
• On-the-fly search
• Zooming
• View call hierarchy
• Sequence diagrams
• Dependency graphs
• IntelliTrace and dump debugging
• Multi-monitor support
• Intellisense improvements

For a detailed description of each of these features read my blog post Ten Cool Things You Didn’t Know About Visual Studio at my Scott Logic blog.

In my model I’ve had a simple many-to-many relation: Entries that had multiple Categories each and Categories that belonged to multiple Entries as well.

Categories and Entries: a many-to-many relationship

In the underlying SQL model the many-to-many relationship was implemented via a relationship table:

Categories and Entries table structure

Now I wanted to find all entites within a given category. Pretty simple, right? In SQL it would have been a simple join:

SELECT Entry.* FROM Entry JOIN Category_rel_Entry ON Category_rel_Entry.Entry_ID = Entry.ID WHERE Category_rel_Entry.Category_ID = @CategoryId

In Nhibernate this query is a bit more tricky, let me share how it can be done.

Note: this post is also a good example of how to define a many to many relationship schema in NHibernate.

My NHibernate model looked like the following:
The Category entity:

The Enity entry:

To do query all entites of a given category, the Restrictions class has to be used for the categories property of the Entity class. For me it took a while to figure out this solution so here is the code:

Category category; // The category of which its entities wanted to  be retreived
IList nodes = NHHttpModule.CurrentSession.CreateCriteria(typeof(Entity))
                            .CreateCriteria("categories")
                           .Add(NHibernate.Criterion.Restrictions.IdEq(category.ID))
                           .List();

In the end the NHibernate code seems cleaner than the SQL one regarding semantics. However figuring out the syntax of these not so everyday queries is not so self explanatory.

Update (2009. 04. 09.) to have the example working you need to code the the NHttpModule class as well. The code is as follows:

using System;
using System.Collections.Generic;
using System.Text;
using System.Web;
using NHibernate;
 
namespace NHibernateApp
{
    /// 
    /// NHibernateHttpModule
    /// 
    public class NHHttpModule : IHttpModule
    {
        public const string KEY = "_TheSession_";
        private static ISession _session;
 
        private void context_BeginRequest(object sender, EventArgs e)
        {
            HttpApplication application = (HttpApplication)sender;
            HttpContext context = application.Context;
            context.Items[KEY] = NHSessionHelper.OpenSession();
        }
 
        public static ISession CurrentSession
        {
            get
            {
                if (HttpContext.Current == null)
                {
                    if (_session != null && _session.IsOpen)
                        return _session;
 
                    _session = NHSessionHelper.OpenSession();
                    return _session;
                }
                else
                {
                    HttpContext currentContext = HttpContext.Current;
                    ISession session = currentContext.Items[KEY] as ISession;
                    if (session == null || !session.IsOpen)
                    {
                        session = NHSessionHelper.OpenSession();
                        currentContext.Items[KEY] = session;
                    }
                    return session;
                }
            }
        }
 
        private void context_EndRequest(object sender, EventArgs e)
        {
            HttpApplication application = (HttpApplication)sender;
            HttpContext context = application.Context;
 
            ISession session = context.Items[KEY] as ISession;
            if (session != null)
            {
                try
                {
                    session.Flush();
                    session.Close();
                }
                catch { }
            }
            context.Items[KEY] = null;
        }
 
        public void Init(System.Web.HttpApplication app)
        {
        }
 
        public void Dispose()
        {
        }
    }
 
}

And you will also need to create the NSessionHelper class as well. In this class replace MyNHibernateAppAssembly with the name of your assembly that contains NHibernate definitions.

using System;
using System.Collections.Generic;
using System.Text;
using NHibernate;
using NHibernate.Cfg;
 
namespace NHibernateApp
{
 
    /// 
    /// Summary description for SessionHelper
    /// 
    public class NHSessionHelper
    {
        private static ISessionFactory _sessionFactory = null;
        private static ISessionFactory SessionFactory
        {
            get
            {
                if (_sessionFactory == null)
                {
                    Configuration config = new Configuration();
                    config.AddAssembly("VizsgTervNHibernate");
                    _sessionFactory = config.BuildSessionFactory();
                    if (_sessionFactory == null)
                        throw new InvalidOperationException("Could not build SessionFactory");
                }
                return _sessionFactory;
            }
 
        }
 
        public static ISession OpenSession()
        {
            ISession session;
            session = SessionFactory.OpenSession();
            if (session == null)
                throw new InvalidOperationException("Cannot open session");
 
            return session;
        }
 
    }
 
}

In Sense/Net 6.0, the open source ECMS I am working on I came across the exact same problem: I wanted to resize images, this time on the fly. This time however I decided to do it a bit less complicated as in previous cases and create a clean and simple solution. After a good deal of googling, tutorial reading and planning I came up with a fairly simple and good code snipplet.

The way I intended the thumbnail generation to work is to always fit the thumbnail into the target dimensions. So if there was a 600×400 picture and the target thumbnail size was 200×200 the generated picture would be 200×133. Similarly if the image size was 100×500 and the target thumbnail size 200×200 the generated picture’s width and height would be 40×200.

The idea behind the algorithm was to first determine whid dimension (width or height) is the “more critical” dimension. In the above examples at the 600×400 picture this dimension is the first one, width. In the second case (100×500) it is the second dimension, the height.

In the algorithm the first step is to determine this “critical” dimension by comapring current size/target size splits. Which ever is larger will be the dimension that will be resized to the thumbnails same dimension size and the other dimension will be shrinked accordingly. This might sound a little confusing so let the code speak for itself.

The C# code:

private static void GetRealXY(double imgX, double imgY, double targetX, double targetY, out double realX, out double realY)
{
	double xScale = imgX / targetX;
	double yScale = imgY / targetY;
 
	// Do not enlarge image
	if (yScale < 1)
		yScale = 1;
	if (xScale < 1)
		xScale = 1;
 
	if (yScale > xScale) // Image has to be shrinked based on height
	{
		realX = imgX * 1 / yScale;
		realY = imgY * 1 / yScale;
	}
	else // xScale > yScale // Image has to be shrinked based on width
	{
		realX = imgX * 1 / xScale;
		realY = imgY * 1 / xScale;
	}
}

And here’s how the same code in PHP would look like:

function getRealXY($imgX,$imgY,$targetX,$targetY)
{
	$xScale = $imgX / $targetX;
	$yScale = $imgY / $targetY;
 
	// Do not enlarge image
	if ($yScale < 1)
		$yScale = 1;
	if ($xScale < 1)
		$xScale = 1;
 
	if ($yScale > $xScale) // Image has to be shrinked based on height
	{
		$realX = $imgX * 1 / $yScale;
		$realY = $imgY * 1 / $yScale;
	}
	else // $xScale > $yScale // Image has to be shrinked based on width
	{
		$realX = $imgX * 1 / $xScale;
		$realY = $imgY * 1 / $xScale;
	}
	return array("x"=>$realX,"y"=>$realY);
}

Hope this snipplet helps. If you’re interested for the whole image resampling process you will be able to access it in the Sense/Net 6.0 Beta 3 source code which will be coming out in a few days.

ctl00_WebPartManager1_184c801b_7c67_4967_be37_3ab10406f967_ctl00_linkVote1
.

Then the name sent on postback (which identified the control on the server side) was:

ctl00$WebPartManager1$184c801b$7c67$4967$be37$3ab10406f967$ctl00$linkVote1
.

But since some time the _ to $ conversion is not true any more. Instead the _ signs are randomly replaced by dollar signs. Recently the name sent back on postback is:

ctl00$WebPartManager1$184c801b_7c67_4967_be37_3ab10406f967$ctl00$linkVote1
.

This change screwed up quite a lot of client side manual __doPostback() codes I’ve written – most of the making these invokes due to JQuery addons.

I was notified by a client that recently a voting control I developed for them is not working any more. The control itself was a list of radio buttons transformed to stars with some javascript magic. On the client side I had to code that when the user clicks any radio button a the form gets submitted by imitating a LinkButton being clicked.

So this is the code I had in the aspx page that worked like a charm so far:

function SubmitVote() {
__doPostBack('','');
}

Now after this change no events were fired on the server side because of the control name not being valid.

Next try: instead of transforming the ID lets just get the name property from the control:

function SubmitVote() {
__doPostBack($("#").attr("name"),'');
}

This solution didn’t work either as the control did not have a name attribute. So the way I dot the thing solved was by invoking the click event of the LinkButton:

function SubmitVote() {
eval($("#").attr("href"));
}

And this was the way I managed to successfully invoke the __doPostback() method from client side. Actually it wasn’t really invoking but only immitating the user clicking.

If you’ve information on when (and why) this server control naming convention has been changed feel free to share within the comments.

In the database schema I had two tables parentTable and childTable. ParentTable had an ID column and some other columns. The childTable had an ID primary key column, some other columns and a parentId NOT NULL column which had a foreign key constraint against the ID column in the parentTable.

The NHibernate schema describing the relationship was the following:

• Parent schema (the relevant part):

• Child schema (the relevant part):

I was trying to delete the child the following way:

ParentClass parent;
ChildClass child;
parent.children.Remove(child);
session.Delete(child);
session.Save(parent);
session.Flush();

When executing the code I got an exception saying

Cannot insert the value NULL into column 'parentId', table 'childTable'; column does not allow nulls. UPDATE fails.
The statement has been terminated.

It took me some googling to find out what I was doing wrong. The NHibernate schema was incomplete. The errors were:

• I did not specify that the parentId column in the childTable cannot be NULL. This can be done using the not-null=”true” attribute
• I did not specify that the one-to-many relation was inverse even though it was: the child knew its parent but the parent did not directly know its children in the database schema. (See this article on what this inverse attribute means in more of depth).
• It is obvious that no child can exist without a parent. NHibernate will take care of deleting the child once it is removed from the children collection of the parent if the on-delete=”cascade” is set on the key element within the children collection.

The NHibernate schema that solved my problem finally is the following:

• Parent schema (the relevant part):

• Child schema (the relevant part):

However when trying to re-map a project to the same localtion as before (only from the new TFS server) I got the following error message: “The path [yourpath] is already mapped in workspace [yourworkspace]“. It took me a while to figure out what the resolution of the problem was.

Attempts not working

It was obvios that a conflict occured with the previos workspace. However I had no access to the previos TFS server and could not view or delete the previos workspace. (When having access to the existing TFS server the workspace uses it is easy to delete it either with the Team Foundation Sidekick – get latest version here).

When not being able to connect to the TFS server workspaces associated to that server cannot be accessed with the GUI (or at least I found no way) and cannot be deleted. However this can be done with command line. I borrowed the idea from a post by Mahesh Mitkari and first checked if the workspace is indeed mapped. To list the workspaces mapped in the system use this command:

tf workspaces

(Note that tf.exe is either located within C:\Program Files\Microsoft Visual Studio 8\Common7\IDE for VS2005 or C:\Program Files\Microsoft Visual Studio 9.0\Common7\IDE for VS2008).

To delete a workspace simply type

tf workspace /delete [/server:servername] workspacename[;workspaceowner]

After typing this command I got the error “Unable to connect to the remote server. Command cancelled”. Something was definitely not right: why can’t I delete any workspace without connecting to it?

The solution

It turned out that I was on the wrong track all the time. After searching I ran into the solution by Buck Hodges in a comment on a blog. Turned out that getting rid of disconnected workspaces isn’t done by deleting but by removing them with the following command:

tf workspaces /remove:*

This command cleans out all cached workspaces. And this was the thing that solved my problem. (Actually it didn’t make that much sense why there are these two distinct commands… here’s a good blog post on this topic)