|
Have you had any luck on tracking this discrepancy down? I continue to have this problem, despite a brief period where I seem to have gotten it working.
________________________________________________________________________
Dave
Y10K bug! Let's not get caught with our pants down **AGAIN**! (DC 02002)
|
|
|
|
|
I am going through the exercises in a textbook to learn about exception handling. This is my code:
<br />
Dim n1, n2 As Integer<br />
Dim n3 As Single<br />
n1 = CInt(txtIn1.Text)<br />
n2 = CInt(txtIn2.Text)<br />
Try<br />
lblOut.Text = n1 / n2<br />
Catch ex As Exception<br />
lblWarn.Text = "Warning!"<br />
End Try<br />
1] The book says Catch e As Exception, but Visual Studio defaults to 'ex' instead of 'e'. If I deliberately put in 'e', I get a build error: "Variable 'e' hides a variable in an enclosing block." What is wrong here?
2] It seems to me that, if a zero is put into txtIn2, it should throw an error. It doesn't. It handles it gracefully, but unexpectedly, putting the text 'Infinity' into lblOut.
How am I suppsoed to learn exception handling if it won't act ... exceptional!
________________________________________________________________________
Dave
Y10K bug! Let's not get caught with our pants down **AGAIN**! (DC 02002)
|
|
|
|
|
DaveC426913 wrote:
n1 = CInt(txtIn1.Text)
n2 = CInt(txtIn2.Text)
I don't think you're supposed to use CInt, you should use Convert.ToInt32. VB.NET contains a lot of VB6 crap, because the VB community complained when Microsoft initially ( and with good reason ) removed it.
DaveC426913 wrote:
The book says Catch e As Exception, but Visual Studio defaults to 'ex' instead of 'e'. If I deliberately put in 'e', I get a build error: "Variable 'e' hides a variable in an enclosing block." What is wrong here?
1/ it doesn't matter what you call your variable
2/ you have another variable called e that's visible to this one, so you can't do this, because one overrides the other
3/ you should catch DivideByZeroException, not just Exception. This will catch ANY error, including ones you have not anticipated and/or can't recover from.
DaveC426913 wrote:
2] It seems to me that, if a zero is put into txtIn2, it should throw an error. It doesn't. It handles it gracefully, but unexpectedly, putting the text 'Infinity' into lblOut.
Good question. VB sucks. Use C#. I tried this now, and even if I set n3 to be n1/n2, it still does the same. C# throws an exception, like all real programming languages.
I find it hilarious that your book doesn't know that this doesn't throw an exception, I wonder if the author bothered to try it.
*edit* I apologise if the above seemed harsh, but I have to tell you, the whole office is laughing about this VB 'feature' right now. I'm not wanting to give you a hard time, just VB.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
Christian Graus wrote:
I don't think you're supposed to use CInt, you should use Convert.ToInt32.
Actually, you can still use CInt without any worries of it being on the chopping block in 2005. CInt is actually faster than using Convert.ToInt32 because CInt's conversion code is compiled in-line with the expression. There's no method call to setup and jump to.
His code is converting the DivideByZero exception to a string before it blows up. This results in the string "Infinity" being returned.
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
Isn't CInt one of the legacy methods that was going to be removed ? I wasn't sure on this.
Dave Kreskowiak wrote:
His code is converting the DivideByZero exception to a string before it blows up. This results in the string "Infinity" being returned.
Wrong. I converted into code which didn't use strings at all, and the Single variable gets the VALUE of positive Infinity. That's where the string value comes from, the Single, which does not blow up. Then I changed the code to set to an int, and it DOES blow up then, but it doesn't throw a DivideByZero exception, it throws an OverflowException. C# will blow up on a divide by zero, no matter what type it's being set to, and it throws the right exception.
Do you have an explanation for this that doesn't involve the VB.NET design team spending most of their time drunk ? Because we spent a Friday laughing about it here ( where we never use VB.NET ).
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
Christian Graus wrote:
Wrong. I converted into code which didn't use strings at all, and the Single variable gets the VALUE of positive Infinity. That's where the string value comes from, the Single, which does not blow up.
Sorry. I suffered a momentary rectal/cranial inversion.
It would be interesting to see what IL is generated for the same code.
I'll try and take a look at that when I get to work.
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
Dave Kreskowiak wrote:
I suffered a momentary rectal/cranial inversion
*grin* I'd certainly forgive you for assuming this would not be the case, it's pretty screwed up.
Dave Kreskowiak wrote:
It would be interesting to see what IL is generated for the same code.
I'll try and take a look at that when I get to work.
I'd be really interested to hear your thoughts. Overall, I thought that VB.NET and C# did roughly the same thing, and this sort of difference is frankly astounding to me.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
After doing a little code and IL research and going back to the VB.NET docs on mathematical operators, all has become crystal clear. There are TWO division operators in VB.NET. Click here[^] for the docs, or just read on...
The normal one, that we're all used to using in any language, is "/". Well, in VB.NET, this operator is only used to divide two floating point numbers. The operator is defined for both the Single and Double types and also the Decimal type.
The Single and Double types are evaluated using standard IEEE 754 rules. Nothing new here, it's what we've been using for decades in C/C++ on floating point values and where our "interpretation" problem comes in while evaluating the original posters n1/n2 expression.
The Decimal type is a little different. If the right-hand operand is 0, the a System.DivideByZeroException is thrown. If the resulting value is too large to hold in the Decimal type a System.Overflow exception is thrown. If the resulting value is too small, then the Decimal type result will be 0.
The second division operator in VB.NET is "\" and is reserved for integer division using the VB.NET Byte, Short, Integer, and Long types. This is the division that the equivilent C# code is doing when comparing VISUALLY identical VB.NET and C# code using integer operands!
Now, according to MSDN:
"According to normal operator resolution rules, regular division purely between operands of types such as Byte, Short, Integer, and Long would cause both operands to be converted to type Decimal. However, when doing operator resolution on the division operator when neither type is Decimal, Double is considered narrower than Decimal. This convention is followed because Double division is more efficient than Decimal division."
Keeping that statement in mind, here's the proof of what's going on with the OP's code:
This is the test function code for VB.NET:
Dim n1, n2 As Integer
n1 = Convert.ToInt32("32")
n2 = Convert.ToInt32("0")
Console.WriteLine(n1 / n2)
The output is the string "Infinity". Something C# and C/C++ coders wouldn't expect to see.
The test function in C# is:
int n1;
int n2;
n1 = Convert.ToInt32(@"32");
n2 = Convert.ToInt32(@"0");
Console.WriteLine(n1 / n2);
This throws a DivideByZeroException on the Console.WriteLine call. No surprise there.
The difference (and proof) is in the IL.
First, both functions call the same conversion method to convert a string to an IL int32 type. The IL code appears to be the same between the two. Both generate:
.maxstack 2
.locals init ([0] int32 n1,
[1] int32 n2)
IL_0000: nop
IL_0001: ldstr "32"
IL_0006: call int32 [mscorlib]System.Convert::ToInt32(string)
IL_000b: stloc.0
IL_000c: ldstr "0"
IL_0011: call int32 [mscorlib]System.Convert::ToInt32(string)
IL_0016: stloc.1
Nothing surprising here. Both functions do exactly the same thing with exactly the same data types.
Both calls to convert are followed up by storing the resulting int32's into their appropriate storage locations.
The problem is in the division expression. The C# IL is:
IL_0017: ldloc.0
IL_0018: ldloc.1
IL_0019: div
IL_001a: call void [mscorlib]System.Console::WriteLine(int32)
IL_001f: nop
IL_0020: ret
Simple enough. Load the int32 in location 0 onto the eval stack, then load the int32 in location 1 onto the eval stack. Do a divide operation on those two integers (int32 / int32). Call Console.WriteLine with the result.
The VB.NET IL is VERY different using the exact same character for the operator, but it has a different function. Using "/", the expression isn't evaluated as an int32 / int32. Both operands are first converted to IL float64's (or System.Double type) before the division operation. We all know the results of storing and operating on floating-point numbers in IEEE 754 format... Anyway, this is the VB.NET IL:
IL_0017: ldloc.0
IL_0018: conv.r8 <- Converts the value at the top of the eval stack to float64!!
IL_0019: ldloc.1
IL_001a: conv.r8 <- Converts the value at the top of the eval stack to float64!!
IL_001b: div <- IEEE 754 rules are different than integer division rules!
IL_001c: call void [mscorlib]System.Console::WriteLine(float64)
IL_0021: nop
IL_0022: nop
IL_0023: ret
Now, change the VB.NET code to:
Dim n1, n2 As Integer
n1 = Convert.ToInt32("32")
n2 = Convert.ToInt32("0")
Console.WriteLine(n1 \ n2)
Of course, the resulting division IL code now falls in line with the expected C# IL code:
IL_0017: ldloc.0
IL_0018: ldloc.1
IL_0019: div
IL_001a: call void [mscorlib]System.Console::WriteLine(int32)
IL_001f: nop
IL_0020: ret
If the OP changes the operator in his expression from "/" to "\", it'll do exactly the same thing as the equivilent c# code.
BTW: This testing was done on C# 2005 and VB.NET 2005, Beta 2.
This has been a public service announcement. We now return you to your regularly scheduled program...
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
Thanks - I had pieced part of this together with some help already ( the two operators thing ), but this makes everything quite clear. The core issue then is the convert to a float64. I wonder if these conversions behind the scenes are a reason for the argument that VB.NET is slower than C# ?
Does this mean that C# can't divide using IEEE 754 rules ? Or is it just that VB.NET, being loosely typed, does conversions behind the scenes that C# does not. Certainly I could see where the overflow exception came from, that all makes sense now.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
Christian Graus wrote:
The core issue then is the convert to a float64. I wonder if these conversions behind the scenes are a reason for the argument that VB.NET is slower than C# ?
It'll definitely contribute to the speed difference because of the conversion process and the 64-bit floating division. But I think the problem lies in the fact that the people who want to bash VB.NET so bad just write some ad-hoc test code for both languages, like in my post, don't realize that they're not comparing apples to apples. When the difference comes up, wow!, they just jump all over the language.
Christian Graus wrote:
Does this mean that C# can't divide using IEEE 754 rules ? Or is it just that VB.NET, being loosely typed, does conversions behind the scenes that C# does not.
No. C#, being based on the same Framework, using the same types and generating the same IL, uses the exact same math rules VB.NET does. C# just has a syntax advantage in the compiler where is will recognize the division situation, since it only has a single division operator, and generate the correct code automatically, (for example when dividing an int32 by an int32, assume the result required is an int32 and generate code for integer division), where VB.NET puts more of that decision process on the coder. The coder is responsible for explicitly telling VB.NET which type of division to use, integer (\) or floating-point (/).
I guess you could make a case for that being a simpler advantage for VB.NET, I don't know. In C#, if you wanted to force floating-point division on two int32's, you'd have to cast the operands to some floating-point type in the expression. IMHO, I'd say it might be slightly simpler to control in VB.NET, but more readable and comprehendable in C#.
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
Dave Kreskowiak wrote:
IMHO, I'd say it might be slightly simpler to control in VB.NET, but more readable and comprehendable in C#.
Fair enough - thanks for taking the time to look into this, I've found it quite fascinating.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
|
Thanks for the links.
Mathematically, you CAN divide by zero, but what you get is "infinity".
Actually, this is not true. Google some maths sites if you don't believe me. Either way, as the article states, this leaves the way open for some serious problems in business apps that don't work hard to avoid this happening.
I don't understand the integer divide ( \ ) thing. Does that stop VB from implicitly converting ints to floats ? I guess C#, being more strongly typed, wouldn't need that, and that would go *some* way to explaining why C# and VB.NET behave so differently from one another in this instance.
Thanks for the info.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
|
Thanks
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
If you cast n1 / n2 to an integer...
lblOut.Text = CInt(n1 / n2)
It will throw a divide by zero exception.
Not sure why that is, but setting a labels text returns 'Infinity'.
Dont forget, despite the huge, widespread use of VB, its not a real programming language.
You cant write halfway decent programs with VB.net because it sucks so much.
(Sarcasm intended)
|
|
|
|
|
Daniel1324 wrote:
If you cast n1 / n2 to an integer...
lblOut.Text = CInt(n1 / n2)
It will throw a divide by zero exception
So the problem is with the Single object ? C# behaves differently in this case though, with the same object. It still threw an exception, as it should.
Daniel1324 wrote:
Not sure why that is, but setting a labels text returns 'Infinity'.
Because in VB.NET, an int divided by 0 = infinity when passed to a Single. I don't know why, or why this conversion is obviously implicit. It's the stupidest thing I've ever seen. Like I said, everyone here is still laughing about it.
Here's more fun. If you make n3 an Integer instead of a Single, this code will throw what exception ?
Dim n1, n2 As Integer
Dim n3 As Integer
Dim s As String
n1 = CInt("5")
n2 = CInt("0")
Try
n3 = n1 / n2
s = n3
Catch ex As Exception
s = "Warning!"
End Try
The answer is NOT the 'DivideByZeroException', but an 'OverflowException'. Oh yeah, VB.NET is a real programming language...
Daniel1324 wrote:
Dont forget, despite the huge, widespread use of VB, its not a real programming language.
VB is widely used because it was designed for people who can't program to be able to knock something out. For that, it works. What percentage of applications you bought and run on your desktop were written in VB ?
Daniel1324 wrote:
You cant write halfway decent programs with VB.net because it sucks so much.
Yes, you can write a decent program in VB. Hell, you can write a decent program in almost any language. But that doesn't make it a good tool for the job. On the other end of the coin, you can write any program you like in C. But why would you want to ? C is at least useful, but like VB, it has a lot of hidden traps for the unwary traveller. And that would be my main point. A language for beginners that is rife with hidden gotchas, unexplainable syntax and inexcusable behavour ? Give me a break.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
What percentage of applications you bought and run on your desktop were written in VB ?
Good point.
I actually prefer C# and I am at least as good in C# as I am in VB. But I'm starting college Monday and part of the course is learning VB.net.
I have a love hate relaionship with it... On one hand, I love writing programs in it because, well... its easy. On the other hand, I hate having to tell people that its written Visual Basic.
|
|
|
|
|
Daniel1324 wrote:
Good point.
*grin*
Daniel1324 wrote:
On one hand, I love writing programs in it because, well... its easy.
I don't think C# is any harder. In fact, I think it's easier, because of the many traps in VB.
Christian Graus - Microsoft MVP - C++
|
|
|
|
|
Now that the rest of you are done, trying to thrash VB.Net
Here is the reason the lblOut Has Infinity in it.
Exception
Invalid operation Exception - An operand is invalid for the operation to be performed.
Return Value: NaN
Division by zero Exception
An attempt is made to divide a non-zero value by zero.
Returns: Infinity (1/-0 = negative infinity)
Overflow Exception
The result of an operation is too large to be represented.
Returns: Positive or negative infinity.
Underflow Exception
The result of an operation is too small to be represented.
Returns: Positive or negative zero.
Inexact Exception
The rounded result of an operation is not exact.
Returns: The calculated value.
cref: IEEE-754/IEC 60559
progload
|
|
|
|
|
Did I stumble into the wrong forum? No wait, it still says Visual "Basic / VB.NET" across the top. Are there any VB programmers in here? That's sort of what I was, you know, expecting.
So anyways. Progload, if division by zero *is* an exception thrown, why did it not get caught?
________________________________________________________________________
Dave
Y10K bug! Let's not get caught with our pants down **AGAIN**! (DC 02002)
|
|
|
|
|
Dave,
The Divide-by Function("\") caught the exception, and produced a Text Message = "Infinity"
Thus the statement "ex" would be false, No Exception.
That is by design... It's no "trick or trap".
You can test this by doing this, It will now throw the exception to "ex":
Dim n1, n2, n3 as integer
Try
n3 = n1 \ n2
Catch ex As Exception
Label2.Text = "Warning! " & ex.Message
End Try
Hope this helps dave.
Progoad
|
|
|
|
|
So, the key is making n3 an integer. If it's anything else, it won't throw the exception.
It's confusing, but it's really just an example from the book, so no big deal. I should have checked out the book more carefully before purchasing - it's 3 years old.
Anyway, now that I'm into it, I'm seeing that this book is covering basic stuff I already know about VB. (The first time I looked at VB.NET code, I was terrified by all the "Region / Friend WithEvents Button1 As System.Windows.Forms.Button" stuff - I thought I was starting at square one.)
Once the object-oriented stuff from my little bit of Java programming comes back, I'll be more comfortable.
________________________________________________________________________
Dave
Y10K bug! Let's not get caught with our pants down **AGAIN**! (DC 02002)
|
|
|
|
|
Yes,
The label control is a text control, thus if you return the text ("Infinity") into it, why would it throw an exeception (to ex) because it is text?
The author was not following proper basic syntax coding...
and created his own "bug".
By the Way, I realy enjoy VB and VB.Net, and it's Math capability.
Beyond what you see a lot of folks saying... I think it is a very powerfull language.
Have fun learning it.. I am..
progload
|
|
|
|
|
progload wrote:
thus if you return the text ("Infinity") into it
Ah, well I'd assumed it would be the / operation that would throw it.
I see what you're saying now. The / operation catches the error internally and simply returns 'infinity'. Thus, externally, it is not an error at all until I attempt to do something with it.
In fact, the error it's been coded to throw isn't arithmetical at all, it is a type error - I can't assume that the / operation will return a numerical value; it could return a string.
Small potatoes. This is actually baby steps towards developing an app that will access the Windows Media Services 9 SDK to manipulate broadcast media publishing points. And even THAT is an intermediate step to transposing it to ASP.NET so I can do it from a web app.
Can you say 'out of my league'? Why do I do this to myself?
________________________________________________________________________
Dave
Y10K bug! Let's not get caught with our pants down **AGAIN**! (DC 02002)
|
|
|
|
|