public Blog<BenR>

Blog moved to codebetter.com

reic0113 — Tue, 01 Feb 2005 23:17:00 GMT

I moved my blog a few days ago to codebetter.com, I'm sure most of you know the story by now.

This is just a notice to anyone to subscribes to redirect your aggregator to my new url:

http://codebetter.com/blogs/ben.reichelt/Rss.aspx

I would also encourage you to subscribe to the main feed at codebetter.com

http://codebetter.com/blogs/MainFeed.aspx

thanks for reading!

Thanks Brendan

reic0113 — Fri, 28 Jan 2005 22:16:00 GMT

Visual Blogger test post on the new blog :-)

Scrollable divs and Firefox

reic0113 — Fri, 28 Jan 2005 14:18:00 GMT

Firefox does not support scrollable divs with the mouse wheel.

I poked around in the extensive bug database trying to find this listed, but the number of bugs is just too daunting. This really sucks, because we develop our stuff to work on ie as well as Firefox, so whenever we run into a div with a scroll bar, the screen just kind of jumps around. I'm surprised that I haven't heard much about this problem from other people, as the mouse wheel has become invaluable to me, I hate having to move my mouse to the scroll bar, and manually move it down......

Goodger leaves Mozilla for Google

reic0113 — Mon, 24 Jan 2005 16:57:00 GMT

Hmm, is this the sound of a Google web browser rumor starting up again?? Ben Goodger has left the Mozilla Foundation where he was a lead engineer of the Firefox browser. Coincidence? I think not.......

CSci 101 Part III

reic0113 — Fri, 21 Jan 2005 11:52:00 GMT

Now we've seen a slow way to generate the fibonacci numbers, and we've seen why its so slow, now we'll try to do better. Here's my proposed improvement to our recursive function that we made the first day:

static long Fibo(long n)

{

// check the base cases

if (n <= 1)

return 0;

if (n == 2)

return 1;

// initialize our array of fibonacci numbers

long[] numbers = new long[n];

numbers[0] = 0;

numbers[1] = 1;

// calculate the 2nd through nth

// fibonacci numbers in the array

for (long i = 2; i < n; i++)

{

numbers[i] = numbers[i - 1] + numbers[i - 2];

}

return numbers[n - 1];

}

So what we're doing here is that instead of making recursive calls to the Fibo function, we are just storing the fibonacci numbers in an array of length n. We don't have the recursion tree at all now, as there are no recursive calls. All we have is the overhead of the array, and a loop that will loop n times. During each loop interation, we simply add the two previous numbers in the array, and put the result in the current array position. Using this function we can easily get the 50th, 100th, or 1000000th fibonacci number, no problem. Go ahead, try it! The response is nearly immediate! However, there is one more little thing we can change to make the function even better.

In the code above, we maintain an array of the 0 to nth fibonacci numbers, when really, all we need is the current number, and the previous number, in order to generate the next one. So we can remove the overhead costs that are associated with the array:

static long Fibo(long n)

{

// check the base cases

if (n <= 1)

return 0;

if (n == 2)

return 1;

// initialize our loop varibles

long x = 0;

long y = 1;

long current = 0;

// calculate the 2nd through nth

// fibonacci numbers, but without

// the array overheard

for (long i = 2; i < n; i++)

{

current = x + y;

x = y;

y = current;

}

return current;

}

Here, instead of an array of length n, we simply have x, y, and current. We still have the loop that will iterate n times, but each time the loop adds x and y and then puts the value of y into x, and the new value into y. We can store all the values we need in just three variables!

I ran some tests and compared the results of the 3 different functions, heres what I found:

	10	20	30	40	50
Slow	125000	62500	312500	37484375	4.18E+09
Better	140625	78125	31250	31250	0
Best	93750	109375	46875	78125	0

So you can see that the difference in the functions is trivial when finding the 10th or 20th fibonacci number. However, the difference becomes very noticeable when generating the 40th number, and once we get to the 50th, the difference is staggering. I went and ate dinner while the slow function was processing the 50th number! The numbers provided are the number of "ticks" the functions took. I ran each function 10 times for each value. The difference between the faster two functions is negligible, as the only optimization we used there was to reduce the amount of memory needed, not processing time. These results clearly show how much slower tree recursion can be than an iterative function.

Thanks for reading, I hope you've enjoyed the series!

CSci 101 Part II

reic0113 — Thu, 20 Jan 2005 13:03:00 GMT

Okay, so yesterday we saw one way to generate the "nth" fibonacci number using recursion. The recursion that we used is also called "tree recursion", the reason for this being that we made two recursive calls inside the function. If you were to draw a diagram of the function calls, the diagram would take on the shape of a tree, because every node would have two branches, like this:

If you tried to use this function to find the 35th fibonacci number, you probably realized that it can take quite a bit of time. The reason for this is that we are computing the same value many times over. Consider what happens when you call Fibo(4). The function is going to add Fibo(3) + Fibo(2). Fibo(3) will, in turn, call Fibo(2) + Fibo(1). So you can see that we are calling Fibo(2) twice, which is a waste. Its not a big deal for only the 4th fibonacci number, but when you put in 50, the extra work can definitely start to make a difference.

See if you can come up with a way to speed up the function, and I'll post my ideas tomorrow, along with a performance comparison of a few different functions. Thanks for reading!

Gut feeling followup

reic0113 — Thu, 20 Jan 2005 12:10:00 GMT

Joe posted a comment about my entry on the new nofollow attribute. He has a good idea to fix the problem I had, it would allow blog owners to selectively promote links in comments, so valid links would get their due "Google juice." I think this is a good idea, but I dont think many blog owners would be diligent in promoting peoples links, especially people who get a lot of comments.

What I would like is some sort of keyword list that I could maintain and links that contain these keywords would have the rel="nofollow" attribute. Then meaningful links to real blogs would still their credit. I would also want to be able to retroactively apply the keyword filtering, so if I get a bunch of comment spam one day, and I dont happen to have any keywords for the new spam, I would want to add the keywords, and then have the blogging application check my comments against my updated keywords list.

Gut feeling

reic0113 — Thu, 20 Jan 2005 02:37:00 GMT

Its nice to see some people are not so giddy over the new attribute that we can add to links. I've heard people lauding this as the end of comment spam, and I agree that it will be a big help, however, one important way that people get "known" in the blog world is by getting moved up in the Google rank through meaningful comments. I like the fact that when I make a comment, its contributing to the conversation, but that I'm also getting a link to my blog out there. Now, when I leave comments, it won't be counted if the blog owner applies this attribute to all incoming comments.

This was my first impression when I first learned of the tag, and maybe I'm way off base, but this is just my knee jerk reaction.

CSci 101 Part I

reic0113 — Wed, 19 Jan 2005 03:11:00 GMT

I'd like to do a little educational series a la Raymond Chen and Larry Osterman. Now, this series will NOT be nearly as technical, nor as difficult or in depth as these two can get, its just something that I thought I would try out. If I like it, maybe someday I will have some really cool stuff to give a tutorial on, and I can do that, but for now, we're going to keep it fairly academic, being as I just graduated from college, this seems appropriate.

I'm going to talk about the fibonacci numbers, and a few different ways to go about generating them. The fibonacci numbers are a sequence of numbers that are calculated by adding the previous two numbers in the sequence. The first two numbers are given, they are 0 and 1, respectively. So the 3rd number in the sequence is 0 + 1 = 1, the 4th is 1 + 1 = 2, and so on.

In my first computer science class at school, one of the first concepts they try to beat into you is recursion. A recursive function is one that makes a call to itself. A classic example of recursion is the way that you can write a factorial function.

We can rewrite n! as n * (n-1)!, which we can rewrite as n * (n - 1) * (n - 2)!, etc. The way this would be represented in our code would be:

static int fact(int n)

{

if (n == 1)

return 1;

else

return n * fact(n - 1);

}

Notice how the fact function calls itself. Its important to have a base case so that the recursion ends at some point; in this example the base case occurs when n == 1, at that point we simply return 1 because 1! = 1, so we can stop recursing.

The first way that we will generate the fibonacci numbers is using recursion.

The base cases for our function will be when n == 0 or n == 1, because those fibonacci numbers are given to us. If n > 1, then we will calculate the 2 previous fibonacci numbers, by recursively calling the same function, and adding them together. Here's the code:

static long Fibo(long n)

{

// check the base cases

if (n <= 1)

return 0;

if (n == 2)

return 1;

// calculate the 2 previous

// numbers and add them together

return Fibo(n - 1) + Fibo(n - 2);

}

This code is extremely simple, there are more lines of comments than of actual code! Try this code out in a console application, start out small by calling Fibo(4), Fibo(5), etc. See if you notice anything as you put in higher and higher numbers :-) Next, we'll inspect this function a little, and improve upon it a little bit.

Thanks for reading!

Date.AddMinutes()

reic0113 — Fri, 07 Jan 2005 20:35:00 GMT

I really wanted an AddMinutes() method to use in javascript like C# has. So I set out to write one myself.

I soon realized how complicated it could get, for instance, the minutes could move the date to the next day, month, or year. I started writing down all the conditions to check, to make sure that I could handle all situations. Since javascript uses the setMinutes(), setHours(), setMonth(), etc.. style of altering a date, I was going to parse the minutes into the number of months, days, hours, and minutes that were represented by the minutes. So 10080 minutes would equal 7 days. After I did that I could add all the appropriate fields and then get my new date.

Then I realized a MUCH easier way to get it done. There is a Date method called getTime() that returns the number of milliseconds since 1/1/1970. There is also a Date constructor that takes in an integer, which also represents the number of milliseconds since 1/1/1970. So all I had to do was use the getTime() method, convert the minutes that I want to add into milliseconds, add the milliseconds and the create a new date instance with this new value. Much, much easier and I don't have the headache of managing all those situations mentioned above, whew!! Here's the code I used:

function addMinutes(dateTimeObj, iMinutes) {
var length = iMinutes * 60 * 1000;
var iseconds = dateTimeObj.getTime();
iseconds += length;
var dt = new Date(iseconds);
return dt;
}

Maximum length of url in winamp

reic0113 — Fri, 07 Jan 2005 00:05:00 GMT

Winamp currently has a maximum string length of 260 for urls that point to streaming audio. After I finally got the streaming to work, I wanted to encrypt the querystring, so users couldn't mess with the url. After using the encryption, I kept getting a bunch of errors, most of them were "Invalid length for a base 64 char array." So I naturally assumed that my encryption was getting a little goofed up because of the browser url encoding the string. So I messed around with some diffent encryption types, but nothing worked.

Then I noticed that the url in winamp was different from the one that I was initially sending to the user, it was truncated. Only then did I do some digging and find that winamp uses the MAX_PATH variable in the windows api headers, which is set to 260. This really sucks, because winamp handles streams much better than Window Media Player, because you can set the default action to "enqueue" rather than "clear my playlist and play the new stream", also, WMP doesn't seem to read the track names correctly, so instead of a playlist that has my list of songs, it just has the word "Play" for every track, not real easy to tell whats playing next.

Well, at least now I know that it wasn't something that I did wrong, but I guess I should have verified the url before wasting all that time.....

More on o/r

reic0113 — Thu, 06 Jan 2005 01:21:00 GMT

Brock Allen refers to his object relational mapping tool as an RO mapper, as opposed to an OR mapper.

He does this because he builds the database first, and then maps the database into objects, so its a relational-object mapper instead. I've never heard it called that before, but I like it, because thats exactly the process that we follow too, build the database how you want it, and then make the objects fit the data model.

Alternative NUnit GUI

reic0113 — Wed, 05 Jan 2005 11:20:00 GMT

I just read about this alternative gui for nunit over at Roy Osherove's.

I just downloaded it and it looks pretty sweet! Haven't had a chance to use it much yet, but I'm excited to see all the statistics that it produces....

Updates, Inserts, and checkboxes

reic0113 — Tue, 04 Jan 2005 03:49:00 GMT

When we were developing our data access objects, we added an Update, and an Insert method, with the corresponding signatures:

Update(System.Collections.Specialized.NameValueCollection form);
Insert(System.Collections.Specialized.NameValueCollection form);

These methods take advantage of the fact that when you post a form, the form values are contained in a NameValueCollection object, so you can loop through the elements of the form, and update the respective data fields, and then update (or insert) the database with the new values.

This scheme is really convenient for when you want to add or remove items from the form, because you can simply remove them from the html, without even worrying about the backend, because those added or removed form elements are inconsequential to the workings of the code behind. You could have a form that contained none of the data fields, and still call these methods without any problem.

The only problem arises when you need to deal with checkboxes.

If the user checks the checkbox, the value of the checkbox gets submitted and the data object can be updated with the checkbox value. However, suppose the checkbox is initially checked, and the user unchecks the checkbox. Now, when the form is submitted, it appears as though the checkbox never existed, because it will not be present in the NameValueCollection, and the data object does not know to update the appropriate value. We typically use checkboxes to denote true/false or 1/0 values in the database, so if a user were to leave a checkbox unchecked it means that the database value should be false or 0, but since there is no form value present, the data object does nothing to that field.

What we did to work around this problem, was to add a hidden form element to our forms. This hidden field is a comma separated list of all the form elements that exist on the form. This hidden field gets submitted with the rest of the form, and as the data object loops the form elements, it records them. When the looping is finished, it compares the list of form elements that it found, to the comma separated list that was submitted with the form. If there are any form elements in the list, that were not included in the form, then the data object knows that a checkbox was left unchecked, and can then update the field accordingly.

Can't stream mp3 file in .net

reic0113 — Sun, 02 Jan 2005 16:23:00 GMT

I am trying to convert an internal application that we have that streams mp3 files to winamp. The app is currently written in asp, and I'm moving it to asp.net. I copied the streaming code almost exactly from the vbscript to c#, but now when I try to listen to an mp3 from the new site, I get the great error message "error syncing to mpeg." Does anyone have experience in streaming mp3 files to winamp? The asp code reads the file from an ADODB.Stream object in 2048 size chunks and buffers the content. I initially did the same thing in c#, but after getting the error, I tried simplifying the code and just using the Response.WriteFile(string path) method. The reason for reading the file in chunks is so that large files don't timeout.

I think that it might have something to do with a bad header or something, but I don't know if asp.net puts in extra headers automatically, in any case, I don't see why these don't work the same way. I can load both examples in a firefox browser window, and the content is the same, its the base64 content of the file, so I know that its actually reading the correct data, winamp just won't recognize the file as an mp3.

asp code (this works):

Response.Buffer = True
Response.Clear
Response.AddHeader "Content-Disposition", "inline; filename=" & FileName
Response.AddHeader "Content-Length", FileSize

Set objStream = Server.CreateObject("ADODB.Stream")
objStream.Open
objStream.Type = adTypeBinary
objStream.LoadFromFile FilePath

lSize = objStream.size

CHUNK=2048
NumberofBlocks=(lSize -(lsize mod chunk))/chunk
For lBlocks = 1 To NumberofBlocks
If Response.IsClientConnected = False Then Exit For
Response.BinaryWrite objStream.Read(CHUNK)
Response.Flush
Next

objStream.Close
Set objStream = Nothing

c# code:

HttpResponse Response = System.Web.HttpContext.Current.Response;

Response.Buffer = true;
Response.Clear();
Response.ClearContent();
Response.ClearHeaders();
Response.AppendHeader("Content-Disposition", "inline; filename=" + System.Web.HttpContext.Current.Server.UrlEncode(SongName));
Response.AppendHeader("Content-Length", Size.ToString());
Response.WriteFile(path);

/*

System.IO.FileStream fs = new System.IO.FileStream(path, System.IO.FileMode.Open, System.IO.FileAccess.Read);
int CHUNK = 2048;
byte[] buffer = new byte[CHUNK];

while (fs.Read(buffer, 0, CHUNK) > 0)
{
      if (!Response.IsClientConnected)
               break;

     Response.OutputStream.Write(buffer, 0, buffer.Length);
     Response.Flush();
}

fs.Close();
fs = null;

*/

Update: I got it working, there was something in my base page class that was goofing the stream up somehow. I think it had to do with cookies, but I will investigate further.