|
I have an object hierarchy where all of my objects in the application domain inherit from a base class. I also have a persistance layer which works using generics.
Looking at the code below, you realize that my object hierarchy is non-generic. But DBStore methods are genericised. How do I pass the type/class of current type?
e.g.
class DBStore
{
public T CreateObject<T>()
{
...
}
public void DeleteObject<T> (int id)
{
.. logic to delete from DB using metadata about T (containing table name) and primary key field matching 'id' ...
.. eg.. delete from typeof(T).Name where ID = id
}
}
abstract class DomainObject
{
private int _id;
private DBStore _dbStore;
public void Delete ()
{
_dbStore.DeleteObject<<big>WHAT SHOULD GO HERE</big>>(_id)
}
}
class Boat : DomainObject {...}
class Car : DomainObject {...}
I would like to be able to perform following:
Car car = new Car();
car.Delete();
or
Boat boat = new Boat();
boat.Delete();
Only things that come to my mind are:
1) Move "Delete" method from the base class to each of the deriving class: (code redundancy ..  )
2) Ungenerecise the "DBStore" method (inconsistancy in DBStore -- all methods are generic, except for Delete object)
Thanks,
- Malhar
|
|
|
|
|
GetType() perhaps ?
Christian Graus - Microsoft MVP - C++
Metal Musings - Rex and my new metal blog
"I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
|
|
|
|
|
I wished! C# compiler expects a class, and not a variable of type 'type'.
<pre>Type t = GetType();
_dbStore.DeleteObject <t> (_id);</pre>
--> The type or namespace name 't' could not be found (are you missing a using directive or an assembly reference?)
-- modified at 21:22 Wednesday 21st February, 2007
|
|
|
|
|
OK, that really sucks.
C# generics are kind of gay, they are not much use for anything too complex. They are designed so people can't hurt themselves using them, I suspect.
Christian Graus - Microsoft MVP - C++
Metal Musings - Rex and my new metal blog
"I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
|
|
|
|
|
Why instead don't you genericize abstract class DomainObject<t>?
|
|
|
|
|
If I'm not mistaken, then you're suggesting something like:
public abstract class DomainObject<T>
This will break majority of the app where I'm simply keeping a reference to the base DomainObject as opposed to DomainObject for Car. You could help clarify my misunderstanding .. but:
Assuming I changed the class declaration as above (with generics), what should I change my dependent method signature for the classes that do NOT inherit from generics?
e.g.
01. class ListViewEx
02. {
03.
04. DomainObject _selectedObject;
05.
06. public void AddItem (DomainObject obj)
07. { ... }
08.
09. public DomainObject SelectedObject
10. { get { return _selectedObject; }
11. set { ... } }
12.}
- What should the line #4, 6 & 9 look like after genericising DomainObject?
- And one thing to note is that currently, ListViewEx.AddItem is designed to allow consumer to pass both objects of type Car AND Boat. If I genericize the ListViewEx to class ListViewEx <T> then a. I'll lose the functionality of adding items of multiple types, and b. Win form designer will no longer work.
Thanks,
- Malhar
|
|
|
|
|
I don't see enough here to say what your line 4 should be (unless maybe I'm supposed to study the previous code or something). Sorry, not enough time for that.
But yes, you are correct in understanding what I am suggesting. In my haste to make the previous post, I neglected to write my < and > with code.
As to what you need to do with your line 4, I think it's a larger matter than that. If you declare your generic to descend from a base class B, then keep a reference to B. I don't see why you would have a problem with using that reference to refer to your domain object then.
|
|
|
|
|
What I meant was currently line
#4 is: DomainObject _selectedObject
#6 is: public void AddItem (DomainObject obj)
#9 is: public DomainObject SelectedObject
If DomainObject is converted to a generic class, then I can't really compile the ListViewEx class with the above code lines. I must change the declarations as well to .. :
#4 is: DomainObject<??> _selectedObject
#6 is: public void AddItem (DomainObject<??> obj)
#9 is: public DomainObject<??> SelectedObject
So the correct question is, what would I substitue "??" with? if you're implying the following, then there may be a possiblity but that'd just create unnecessary classes and cause confusion:
abstract class BaseObj
{
}
abstract class DomainObject <T> : BaseObj
{
}
class Car : DomainObject <Car>
{
}
class Boat : DomainObject <Boat>
{
}
class ListViewEx
{
void AddItem (BaseObj obj)
{
}
}
The "BaseObj" is simply an empty class and it just changes the object model where consumers can begin using 'baseobj' instead of DomainObject -- since classes such as "ListViewEx" have been changed to accept BaseObj.
- Malhar
|
|
|
|
|
Giving serious attention to this is going to swerve me way off course at the moment. Generics should always be addressed very carefully. I think you have some really sour syntax here.
When you call your DeleteObject() method, it must be able to handle an object of class/type t. You shouldn't have to tell it what class to handle at that point. Your method should just act on an index, and assumably remove a reference.
I think the compiler is possibly trying to interpret your _dbStore.DeleteObject <t> (_id); statement as a type declaration, and before it gets very far with that, it chokes on the <t> syntax.
|
|
|
|
|
mike montagne wrote: When you call your DeleteObject() method, it must be able to handle an object of class/type t.
I couldnt agree with you more on this point. But there are cases when from you non-generic class, you want to call a generic method and all you want do is to pass the "current type" and not any other type. In most cases, you will hard-code the class name, but if you're working in the base class, you would not know the type of the deriving class. The only workaround for this limiation, is to make the method from your base class abstract and let the deriving classes implement the call. That said, if I have to do this, then I'm working against the what the intention of generics was in the first place.
e.g.
class DomainObject
{
abstract bool Delete(int id)
}
class Car
{
override bool Delete(int id)
{
DBStore.SomeGenericMethod<Car>(args);
}
}
class Boat
{
override bool Delete(int id)
{
DBStore.SomeGenericMethod<Boat>(args);
}
}
Anyways, your input is much appreciated .. I'm simply being a devil's advocate here. I'm sure i'm not the only person encountering such scenarios and this would a common problem.
- Malhar
|
|
|
|
|
malharone wrote: I couldnt agree with you more on this point. But there are cases when from you non-generic class, you want to call a generic method and all you want do is to pass the "current type" and not any other type. In most cases, you will hard-code the class name, but if you're working in the base class, you would not know the type of the deriving class.
I think there's a conceptual problem here. First of all, you can overload your method if you need to process a handful of known classes, or you can simply let descendants override your method (if they need to), and whatever calls the method ends up firing, being as they belong to the instance you (or whoever) calls them on, this ends up firing the method implementation of the closest descendant to the class it's called on (which may be the instance class itself). As your class design is a matter of writing these methods *for each class* (if necessary to override them), this *is* the right method (whether overloads or overrides are necessary or not -- they may very well not be).
Now, if you need to identify the class otherwise (which isn't evident from your example), you can also pass your own identifier argument. It doesn't even have to be a real class -- it could be an enumerated type... anything, even an integer. C++ guys do this all the time to eliminate dynamic casting, and it's perfectly sound if you cover your bases... and faster than greased things unmentionable in C# forums.
Reflection will also work to determine the class (although I have to admit I'm a bit raw with this with C#, as it's something I'll be getting to in the near future). I know for instance I have substantial notes and collected material on the Activator class, which I intend to use to create instances of unknown classes, which certainly is an intersecting area.
-----
Now, I also have another thought. You have a mixture of generics and non-generics interacting. Something you wrote earlier also tells me you are flexible -- you could write this as all non-generic then?
If you can, that's a good way to work out the problem. Get it working without generics. This will tell you exactly how to build generics into it, if that is a good thing to do.
It probably is a good thing to do if you have to do any casting, because it can eliminate casting altogether, and provide wholy type safe operation with minimal code.
I really can't see what the problem is. I look at your question, and you say you have to pass the type, and I ask myself, is he trying to call a static method without an instance (which is a place where the method doesn't know what type it belongs to, because there is no inherent this or instance pointer). Otherwise, a method always knows what type it belongs to, and doesn't need to reference the type. So unless a method has to process a further type which hasn't been cast for processing, I have yet to see in your question why you can't get the job done one of these ways.
Actually, you say:
malharone wrote: The only workaround for this limiation, is to make the method from your base class abstract and let the deriving classes implement the call.
So we are in complete agreement. Why then can't you let that method call preciptate to the descendant, which, if it had needed to, has revised the method to handle itself however necessary? That's just general practice.
|
|
|
|
|
PS.
My own use of (or need to use) generics has never encountered any challenging hurdle that I can't just write them at will -- and any usage has always been very straightforward. I still think you have a basic conceptual class design problem, but just in case you might find deep generic study helpful, I think Richter is a truly excellent source (CLR Via C#, Second Edition). On p. 374, you will find an excellent chapter on Generic Methods. On p. 375, there is an excellent subtitle on Generic Methods and Type Inference. Much further material addresses everything you need to know about generics. I highly encourage you to use them for the same reasons Richter does -- processing is inherently much faster, because casting/boxing are eliminated (he cites many examples, and practice proves they are quite correct). Good programming will always use them where they benefit a project in this way -- generics are unavoidable necessities to good practice. But they really aren't that difficult.
|
|
|
|
|
Just for example (before I get back to work),
Here's a declaration from current work which is not really difficult to grasp. It's declaring an outer generic container of 2 levels of inner generic containers. What are the ramifications of this? Not much, really. It just allows me to do a lot of work without casting -- which in fact is very important to high speed iterative processes which have to act very often on each internal instance.
protected List<List<List<Bitmap>>> v_Maps_MASTER = new List<List<List<Bitmap>>>( );
This is just a container for further generics which are declared like this:
protected List<Bitmap> v_Maps_Button_UP = new List<Bitmap>( );
In the end however, every member is just a bitmap, and therefore every instance can be called just as "if" it is a bitmap, because *it IS* a bitmap ("generic" or not):
if ( ( enumContours.TopAndBottom == f_Contours ) || ( enumContours.TopOnly == f_Contours ) )
{
v_Maps_UP[ i_Btn_BL_BR ][ i_NF ].SetPixel( 0, 0,
Color_IncOrDecLum( v_cBaseUP_Glossed, lumDiffAccumulated_I + f_LumDif_EdgeBevels ) );
}
else
{
v_Maps_UP[ i_Btn_BL_BR ][ i_NF ].SetPixel( 0, 0,
Color_IncOrDecLum( v_cFLATFILL_UPContours, lumDiffAccumulated_I + f_LumDif_EdgeBevels ) );
}
So this is just basic practice. Generics just help us process *a class* (or classes, as they are usually processed).
|
|
|
|
|
PS.
*If one or however many of the instances descended from Bitmap, it still could legally be assigned to the instance pointer comprising each "member." If my methods act on that instance, and if it overrides a Bitmap method that I call, the proper method is automatically called.
|
|
|
|
|
malharone wrote: But there are cases when from you non-generic class, you want to call a generic method and all you want do is to pass the "current type" and not any other type.
Whenever you need to do this, this is the way to write your call:
( CurrentType )InstanceReference.Method( args )
|
|
|
|
|
malharone wrote: WHAT SHOULD GO HERE
The name of a class.
You could use the name of the base class DomainObject.
|
|
|
|
|
Yes, the name of the class would go in the place holder. But it would be the name of the concrete type as opposed to hard-coding base type.
E.g.
If I have Car c = new Car();
and when I perform c.Delete() operation, I'd like to be able to execute _dbStore.DeleteObject<Car>(_id) from the DomainObject.
If I have Boat b = new Boat();
and when I perform b.Delete() operation, I'd like to be able to execute _dbStore.DeleteObject<Boat>(_id) from the DomainObject.
I just dont know how I could dynamically "pass" the runtime class name to the generic method in DBStore.
- Malhar
|
|
|
|
|
<br />
class DBStore<T> where T : class, new()<br />
{<br />
public T CreateObject()<br />
{<br />
return new T();<br />
}<br />
public void DeleteObject<T>(int id)<br />
{<br />
Type t = typeof(T);<br />
}<br />
public DBStore()<br />
{<br />
}<br />
}<br />
abstract class DomainObject<T> where T : class, new()<br />
{<br />
protected int _id;<br />
protected DBStore<T> _dbStore;<br />
public void Delete ()<br />
{<br />
_dbStore.DeleteObject<T>(_id);<br />
}<br />
}<br />
class Boat : DomainObject<Boat><br />
{<br />
public Boat()<br />
{<br />
}<br />
public Boat(int id)<br />
{<br />
this._id = id;<br />
this._dbStore = new DBStore<Boat>();<br />
}<br />
}<br />
class Car : DomainObject<Car><br />
{<br />
public Car()<br />
{<br />
}<br />
public Car(int id)<br />
{<br />
this._id = id;<br />
this._dbStore = new DBStore<Car>();<br />
}<br />
}<br />
<br />
class TestClass<br />
{<br />
public void Run()<br />
{<br />
Boat boat = new Boat(1234);<br />
boat.Delete();<br />
<br />
Car car = new Car(5678);<br />
car.Delete();<br />
}<br />
}<br />
|
|
|
|
|
I think the forum software treated the <s & >s as HTML tags. Do you mean the following instead?
abstract class DomainObject <T> where T : class, new() and
class Boat:DomainObject<Boat> and
class Car:DomainObject<Car>
I have two nags now (nohing major):
a) change method signature from ListViewItem GetItem (DomainObject o) to ListViewItem GetItem (DomainObject <DomainObject> o) in non-generic classes (such as list views)
b) redundancy in class declaration: class Car:DomainObject<Car>
Any opinions?
I think you also meant to genericise the DBStore class, but the lts & gts were converted to HTML. Anyways, in my case, I am planning on keeping the DBStore a non-generic singleton. So, only the methods in it will have generic declaration.
Nonethelss, thanks for your previous suggestion.. I didnt know about the class keyword. i'll look into it!
- Malhar
|
|
|
|
|
If I'm not mistaken, then you're suggesting something like:
public abstract class DomainObject<T>
This will break majority of the app where I'm simply keeping a reference to the base DomainObject as opposed to DomainObject for Car. You could help clarify my misunderstanding .. but:
Assuming I changed the class declaration as above (with generics), what should I change my dependent method signature for the classes that do NOT inherit from generics?
e.g.
01. class ListViewEx
02. {
03.
04. DomainObject _selectedObject;
05.
06. public void AddItem (DomainObject obj)
07. { ... }
08.
09. public DomainObject SelectedObject
10. { get { return _selectedObject; }
11. set { ... } }
12.}
- What should the line #4, 6 & 9 look like after genericising DomainObject?
- And one thing to note is that currently, ListViewEx.AddItem is designed to allow consumer to pass both objects of type Car AND Boat. If I genericize the ListViewEx to class ListViewEx <T> then a. I'll lose the functionality of adding items of multiple types, and b. Win form designer will no longer work.
Thanks,
- Malhar
|
|
|
|
|
I am having to strip some info from some text files to dump into a data table. The problem, well, not a problem, but a nuance, is that the information needed from these text files are at the very end of the file, which average about 400K.
Here is the code I am using:
objRead = new StreamReader(m_strFileName);
while (objRead.Peek() > -1)
{
strRecord = objRead.ReadLine();
if (strRecord.Length > 20)
{
if (strRecord.Substring(44,3) == "NEW")
{
if (strRecord.Substring(87,11) != " ")
{
decDepr[0] = Decimal.Parse(strRecord.Substring(87,11));
}
else{decDepr[0] = 0;}
}
You get the idea...
The question is can I start reading from the end of the file or is there a smarter way to do this? I am having to repeat this process 22 times for 22 different reports, which takes some time.
Thanx in advance!
Jude
|
|
|
|
|
from this code, it seems that whether the data is in the end or start it will take the same time, since you are continuing the loop even after getting the data. so i do not see any difference in time even if you start reading from bottom.
Regards
Shajeel
|
|
|
|
|
Well, if I do start reading from the bottom I will get the information I need then be able to close the stream read and go on to the next file. The data that needs to be read is at the very bottom of the file. I just don't know how to start reading from the end of the file.
Thanx for the reply!
Jude
|
|
|
|
|
Hi,
have a look at FileStream.Seek() or FileStream.Position which will allow you to read
into a byte[] the last N bytes of a file.
|
|
|
|
|
I have some problems with microsoft´s example under void GetObjectData(SerializationInfo info, StreamingContext context) in MSDN VS5. I have executed the code and it´s running, but when I remove formatter.Serialize(fs, a1) and try to Deserialize the binary file, it fails.
Have I misunderstod the example or What is wrong? Can I fix the code I need to serialize a singleton.
Best regards, Gywox
Modified Example:
<br />
using System;<br />
using System.Web;<br />
using System.IO;<br />
using System.Collections;<br />
using System.Runtime.Serialization.Formatters.Binary;<br />
using System.Runtime.Serialization;<br />
using System.Security.Permissions;<br />
<br />
[Serializable]<br />
public sealed class Singleton : ISerializable<br />
{<br />
private static readonly Singleton theOneObject = new Singleton();<br />
private string someString_value;<br />
private Int32 someNumber_value;<br />
<br />
public string SomeString<br />
{<br />
get<br />
{<br />
return someString_value;<br />
}<br />
set<br />
{<br />
someString_value = value;<br />
}<br />
}<br />
<br />
public Int32 SomeNumber<br />
{<br />
get<br />
{<br />
return someNumber_value;<br />
}<br />
set<br />
{<br />
someNumber_value = value;<br />
}<br />
}<br />
<br />
private Singleton()<br />
{<br />
}<br />
<br />
public static Singleton GetSingleton()<br />
{<br />
return theOneObject;<br />
}<br />
<br />
[SecurityPermissionAttribute(SecurityAction.LinkDemand, Flags = SecurityPermissionFlag.SerializationFormatter)]<br />
void ISerializable.GetObjectData(SerializationInfo info, StreamingContext context)<br />
{<br />
info.SetType(typeof(SingletonSerializationHelper));<br />
}<br />
}<br />
<br />
<br />
[Serializable]<br />
internal sealed class SingletonSerializationHelper : IObjectReference<br />
{<br />
public Object GetRealObject(StreamingContext context)<br />
{<br />
return Singleton.GetSingleton();<br />
}<br />
}<br />
<br />
class App<br />
{<br />
[STAThread]<br />
static void Main()<br />
{<br />
FileStream fs = new FileStream("DataFile.dat", FileMode.OpenOrCreate);<br />
<br />
try<br />
{<br />
BinaryFormatter formatter = new BinaryFormatter();<br />
<br />
Singleton[] a2 = (Singleton[])formatter.Deserialize(fs);<br />
Console.WriteLine("Do both array elements refer to the same object? " + (a2[0] == a2[1]));<br />
Console.WriteLine(a2[0].SomeString + " " + a2[0].SomeNumber);<br />
Console.WriteLine(a2[1].SomeString + " " + a2[1].SomeNumber);<br />
}<br />
catch (SerializationException e)<br />
{<br />
Console.WriteLine("Failed to serialize. Reason: " + e.Message);<br />
throw;<br />
}<br />
finally<br />
{<br />
fs.Close();<br />
}<br />
}<br />
}
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
