|
Just doing my own comparisons using:
private const int COUNT = 100000000;
public class Base { }
public class Derived : Base { }
static void Main(string[] args)
{
object o = new Derived();
Derived result;
bool b;
Stopwatch sw;
sw = Stopwatch.StartNew();
for (int i = 0; i < COUNT; )
{
if ((result = (o as Derived)) != null)
{
++i;
}
}
sw.Stop();
Console.WriteLine("'As' Elapsed: {0}", sw.ElapsedMilliseconds);
sw = Stopwatch.StartNew();
for (int i = 0; i < COUNT; )
{
if (o is Derived)
++i;
}
sw.Stop();
Console.WriteLine("'Is' Elapsed: {0}", sw.ElapsedMilliseconds);
Console.ReadLine();
}
I get the following results:
'As' Elapsed: 460
'Is' Elapsed: 414
So, basically, it boils down to a 10% savings in time (ignoring the two operations are not on the same footing - 'as' requires an additional command to store the value in result). But because you spend additional time retyping using the 'as', you end up always spending more time. In essence, you might get into the correct branch to know the type, but converting the type blows away all the savings you made. The only case where you end up being faster is when you don't go through any branches and just bail out of the method entirely, so it depends on your usage.
At the very least, you could remove the additional type cast into Message that is not needed.
Ultimately, it is your code. Good read. 
|
|
|
|
|
Ok, you inspired me to do some benchmarking of my own...
Didn't think your test was quite accurate, as the declarations would have to be inside the loop, not outside... Didn't make much of a difference, if any.
Then I tried adding two more classes (Derived2, Derived3), declaring a variable of each, testing each in turn, and doing a single "as" in the second part (After it determines the type)... Basically matching the real situation perfectly.
Test object is of type 'Derived':
'As' Elapsed: 453
'Is' Elapsed: 577
Test object is of type 'Derived2' (Second test):
'As' Elapsed: 982
'Is' Elapsed: 1030
Test object is of type 'Derived3':
'As' Elapsed: 1575
'Is' Elapsed: 1529
So there is a SLIGHT gain for using the "is" strategy when testing three or more times, but since the tests would be arranged such that the more common ones are near the top, the numbers to look at would be the middle/average, in this case 'Derived2'... And your method comes out ahead
Gotta send a note to myself to swap it over when I'm back in the office on Monday.
(Granted, each of these messages is coming across a network connection, so 50ms per 100 million messages is completely insignificant, but it's the principle of the thing )
|
|
|
|
|
Ian Shlasko wrote: Granted, each of these messages is coming across a network connection, so 50ms per 100 million messages is completely insignificant, but it's the principle of the thing
I completely disagree, benchmarking and optimising such a case is a complete waste of time, instead look for places where your application could really do with optimisation, work on other features, etc.
While I generally think that Hoare's classic "Premature Optimisation is the Root of All Evil" is too often used to excuse developers from good design (not all optimisation is premature), this may be a classic case where the maxim really applies.
|
|
|
|
|
Hey, I didn't spend much time on it in actual development... But it's an interesting discussion point, in case next time a similar situation pops up, it actually IS important.
I have another routine in a related class that actually handles the routing of messages to/from the different clients and server-side workers... That one sometimes processes 200-300 messages a second (Update notifications and logs, mostly). If that one was doing any sort of casting, something like this might be significant.
So sure, shaving a couple cycles in this particular block of code is insignificant, but it's something to know for the future.
|
|
|
|
|
All calls to reflection are slow in general. In such case, I prefer adding an additional, "constant", enumerated property "type". It is much faster to do a single integer comparison than using the reflection to determine an object's type (which is done in both "is" and "as" operators).
enum MessageTypes { Command, Reply, Subscription }
class Message {
public abstract MessageTypes MessageType { get; };
(...)
}
if (msg.MessageType == MessageTypes.Command) {
}
else if (...)
Greetings - Jacek
|
|
|
|
|
With some more analysis, would it be possible to use virtual functions to take care of this?
Chris Meech
I am Canadian. [heard in a local bar]
In theory there is no difference between theory and practice. In practice there is. [Yogi Berra]
|
|
|
|
|
Nah, my Message classes are DataContract objects... Don't want anything but basic accessors and mutators in there. And besides, all of those OnXMessage functions ARE virtual... This is the base class for different types of automated clients... The individual jobs just override the handler for the messages they want to react to.
|
|
|
|
|
Some years ago, I worked at the German office of a big international hospital information system manufacturer. The server was written in MUMPS (that's really a disease...), and lots of the code were developed at the US head quarters.
Some day, we had problems in the communication between computers. Eventually we found out that our American colleagues stored IP addresses in two floating point numbers and then concatenated them with a ".", something like address=float1+"."+float2
It worked fine in the US, but in Germany we use a decimal comma instead of a decimal point, and "192,168.109,47" is not a correctly formed IP address...
|
|
|
|
|
Easy to fix. You can use the code we use at our company to convert floats between european/american decimals notation:
Function CAMBIADECIMAL(ByVal NUMERO As String, ByVal TIPO As Integer) As String
Dim I As Integer
Dim J As Integer
Dim ss As String
Dim ss1 As String
Dim PARTE1 As String
Dim PARTE2 As String
Select Case TIPO
Case 1
ss = NUMERO
I = ss.IndexOf(".")
If I > 0 Then
J = ss.Length - I - 1
PARTE1 = ss.Substring(0, I)
PARTE2 = ss.Substring(I + 1, J)
ss1 = PARTE1 + "," + PARTE2
Return (ss1)
End If
Return ss
Case 2
ss = NUMERO
I = ss.IndexOf(",")
If I > 0 Then
J = ss.Length - I - 1
PARTE1 = ss.Substring(0, I)
PARTE2 = ss.Substring(I + 1, J)
ss1 = PARTE1 + "." + PARTE2
Return (ss1)
End If
Return ss
End Select
End Function
|
|
|
|
|
Here in Argentina we also use comma as floating point separator, and the correct code that always works is:
a) to retrieve the current decimal separator:
System.Globalization.CultureInfo.CurrentCulture.NumberFormat.CurrencyDecimalSeparator
b) to output ANY float using "." as its decimal separator
[a float].ToString(System.Globalization.CultureInfo.InvariantCulture.NumberFormat);
c) to parse using an invariant culture, with "." as decimal separator.
System.Convert.ToSingle("3.1415926", System.Globalization.CultureInfo.InvariantCulture.NumberFormat);
|
|
|
|
|
Hey, take it easy! I was just posting another horror 
|
|
|
|
|
Is there any requirement that any particular octet of an IP address always be non-zero? I would think that if e.g. using a subnet mask of 255.255.254.0, an IP address of 192.168.1.0 or 192.168.0.3 would be perfectly valid.
|
|
|
|
|
public class DateTimeExtended
{
protected static readonly Logger Log = NLog.LogManager.GetCurrentClassLogger();
public enum Days { Sun=0, Mon, Tue, Wed, Thu, Fri, Sat };
public static string[] FerialDays = {"0101",
"0501",
"0508",
"0714",
"0815",
"1101",
"1111",
"1225",
};
public static DateTime GetDate()
{
return DateTime.Now;
}
public static string GetDate(string format)
{
try
{
if(String.IsNullOrEmpty(format))
return GetDate().ToString();
else
return GetDate().ToString(format);
}
catch
{
return DateTime.Now.ToString();
}
}
public static string GetHour()
{
return DateTime.Now.Hour.ToString("00") + ":" + DateTime.Now.Minute.ToString("00") + ":" + DateTime.Now.Second.ToString("00");
}
public static int GetDay()
{
return (int) DateTime.Now.Day;
}
public static int GetMonth()
{
return (int) DateTime.Now.Month;
}
public static int GetYear()
{
return (int) DateTime.Now.Year;
}
public static int GetDayOfWeek()
{
return (int) DateTime.Now.DayOfWeek;
}
public static string GetNextMonthStart(DateTime date, int NbMois)
{
DateTime _date = date.AddMonths(NbMois);
return _date.Year.ToString("0000") + "-" + _date.Month.ToString("00") + "-" + "01";
}
public static bool IsDate(DateTime date)
{
try
{
int day = date.Day;
int month = date.Month;
int year = date.Year;
if(month == 2)
{
if(day == 30 || day == 31)
return false;
if(day == 29)
return ((year%4 == 0 && year%100 != 0) || year%400 == 0);
}
if (day == 31)
return !(month == 4 || month == 6 || month == 9 || month == 11);
return true;
}
catch { return false; }
}
public static string GetGuid()
{
return Guid.NewGuid().ToString();
}
public static DateTime GetDateValue(string sVal)
{
DateTime dtRet;
sVal = sVal.Replace("-", "").Replace("/", "");
try
{
int nAdd = 0;
if (Convert.ToInt32(sVal.Substring(0, 2)) < 80)
{
nAdd = 2000;
}
else
{
nAdd = 1980;
}
if (sVal.Length == 6)
{
dtRet = new DateTime(nAdd + Convert.ToInt32(sVal.Substring(0, 2)), Convert.ToInt32(sVal.Substring(2, 2)), Convert.ToInt32(sVal.Substring(4, 2)), 0, 0, 0);
return dtRet;
}
if (sVal.Length == 8)
{
if (sVal.IndexOf("-") > 0)
{
dtRet = new DateTime(nAdd + Convert.ToInt32(sVal.Substring(0, 2)), Convert.ToInt32(sVal.Substring(3, 2)), Convert.ToInt32(sVal.Substring(6, 2)), 0, 0, 0);
return dtRet;
}
dtRet = new DateTime(Convert.ToInt32(sVal.Substring(0, 4)), Convert.ToInt32(sVal.Substring(4, 2)), Convert.ToInt32(sVal.Substring(6, 2)), 0, 0, 0);
return dtRet;
}
if (sVal.Length == 10)
{
dtRet = new DateTime(Convert.ToInt32(sVal.Substring(0, 4)), Convert.ToInt32(sVal.Substring(5, 2)), Convert.ToInt32(sVal.Substring(8, 2)), 0, 0, 0);
return dtRet;
}
return Convert.ToDateTime(sVal);
}
catch (Exception)
{
}
dtRet = new DateTime(DateTime.Now.Year, DateTime.Now.Month, DateTime.Now.Day, 0, 0, 0);
return dtRet;
}
}
This code has been in use for years, even if it's totaly uselless.
|
|
|
|
|
My favs are:
public static string GetHour() - very confusing if you take a look at functions like GetDay/GetMonth...
public static bool IsDate(DateTime date) - how is this function able not to return true?
Greetings
Covean
|
|
|
|
|
Some developers don't mind multiple returns in one function, but for me, that's a wtf in itself. Not to mention returning a complete statement instead of a value.
V.
|
|
|
|
|
I agree that it is a useless class.
Usually such classes are very useful for future scaling and I do use such classes to make sure that all references are taken from these classes only. It is like creating your own custom controls derived from base controls. Any future modifications are resembled uniformly.
But even after thinking a lot, I couldn't understand what this guy was trying to achieve here. It can't be used for scaling also 
|
|
|
|
|
Thanks, that made standing up early on sunday worth it!
|
|
|
|
|
This is the best part of the class
doud wrote: public static string GetGuid()
{
return Guid.NewGuid().ToString();
}
Maybe rename the class to DateTimeExtendedGetNewGuidAsString
10 PRINT "Software is hard. - D. Knuth"
20 GOTO 10
|
|
|
|
|
We all know Math.Round() basically rounds a particular decimal value. But if we look into this method closely it rounds the value to the nearest even value.
For example if we do Math.Round(2.5,0) it will return 2. But we may expect 3 intead of 2. In that case we can use MidPointRounding parameter of
Math.Round() method.
Math.Round can also take a parameter known as MidPointRounding which helps us to select the Rounding option. That means whether we want to round towards even number or to the number away from zero.
For example, when someone do the rounding of 8.5 . MidpointRounding will help to specify whether rounding will be towards the even number (8) or it will be away from zero (9).
http://www.mindfiresolutions.com/MidPointRounding-for-MathRound-method-in-Net-Application-293.php[^]
Cheers,
Eliza
|
|
|
|
|
Wrong forum, this forum is for code that is soo ugly, its own momma vomits when she looks at it.
Your post would be better in "Tips and Tricks".
Dalek Dave: There are many words that some find offensive, Homosexuality, Alcoholism, Religion, Visual Basic, Manchester United, Butter.
Pete o'Hanlon: If it wasn't insulting tools, I'd say you were dumber than a bag of spanners.
|
|
|
|
|
|
Not really relevant to this forum, but it has caught my attention.
Are these really the only two options for rounding in .NET? What if I want (for example) to always round up (i.e. 1.5 becomes 2, but -1.5 becomes -1). That sounds a bit odd at first, but in some banking applications I have worked on that was how it worked because you rounded in the bank's favour. (Banks never miss a trick when it comes to money.)
|
|
|
|
|
Use the Math.Ceiling(1.5) = 2, Math.Ceiling(1.0001) = 2
Math.Ceiling(-1.5) = -1, Math.Ceiling(-1.99999) = -1
...
Is what you are looking for?
PS: If it is, I'll start hiding my savings under my bed 
|
|
|
|
|
Well I was thinking more in terms of being able to say something like MidPointRounding.Up, or something like that. I'm basically lazy and prefer to rely on other people to do the grunt work and write the boring little stuff.
It comes into things like currency trading for example. If the bank is calculating the rate at which they want to sell you something (dollars, for example) and they need to round it to a certain number of decimals, they will round it in such a way that is favourable to them. It may be a small thing on each trade, but it adds up.
When they are doing things like calculating interest payments, they use "traditional" rounding.
|
|
|
|
|
Please tell which banks these were for so I can bank elsewhere! 
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
