Develop Reference

Looking for a [maybe] design pattern

Here's my scenario. User selects a document in my software and my software extracts some key data out of the document. The software handles two kinds of formats; PDF and DOCX. For each of these types, there are several templates and the uploaded document is supposed to belong to one of these templates. I don't know if this is a well-known problem and if there exists an established design pattern to solve this scenario (that's why I'm on SO). Here's what I have designed so far:
Since each template has specific structure/contents, I'm thinking of creating separate classes for each template. There will be a top-level interface called IExtractor, then there will be two top-level classes called PdfExtractor and DocxExtractor, each implementing the IExtractor interface. Any functionality common to all PDF (or DOCX) templates will go into these parent classes.
Below these two parent classes, there will be several template-classes, one for each template. For example a class called Template571_PdfExtractor that inherits from PdfExtractor, has methods specific to Template 571, but provides results in the same form as any other extractor.
I'm using C# 4.0 if that matters. Here's the skeleton:
The interface:
interface IExtractor
{
void ExtractDocument(System.IO.FileInfo document, dsExtract dsToFill);
}
The two parent classes:
public class DocxExtractor : IExtractor
{
public virtual void ExtractDocument(System.IO.FileInfo document, dsExtract dsToFill)
{
}
}
public class PdfExtractor : IExtractor
{
public virtual void ExtractDocument(System.IO.FileInfo document, dsExtract dsToFill)
{
}
}
One of the concrete classes:
public class Template571_PdfExtractor : PdfExtractor
{
public virtual void ExtractDocument(System.IO.FileInfo document, dsExtract dsToFill)
{
}
}
Now there are a few key questions I'm not sure about. All of them revolve around the problem that I don't know how and where to instantiate the concrete (template) class's object. I can use file extension to decide whether I need to go down the PdfExtractor tree node or DocxExtractor node. After that, it is the file's contents that tells me the template to which user's document belongs. So where do I put this "decision" code? My idea was to put it in the PdfExtractor class (or DocxExtractor for that matter). Is that the correct way?
Sorry I got a bit long, but I didn't know how to fully describe my situation. Thanks for your ideas.
Shujaat

Once you dig deeper into design patterns and such you'll surely find out that most of the time there is no one correct way to implement something...
One possible way would be to create so-called factory classes: One for PdfExtractors, and another one for DocXExtractors. Each factory class would probably have a single static method like
public final class PdfExtractorFactory {
public static PdfExtractor getExtractor(String filename) { ... }
... // constructor, or singleton getter here
}
The logic to decide upon the concrete subclass of the PdfExtractor instance to return (i.e., which template to use) would than reside in the factory method. This way, neither the abstract base class PdfExtractor nor its subclasses would be cluttered with this decision logic. Only the factory classes would need to know about the subclasses of PdfExtractor (resp. DocXExtractor), and the rest of your code would be totally unaware of the concrete subclasses since the factories pass on instances of the superclasses.
Since you're likely to need only a single instance of PdfExtractorFactory and DocXExtractorFactory, you might choose to implement these factory classes as singletons.
Update: Of course you can use either a static factory method or the Singleton pattern and a non-static factory method (but you don't need both).

Looking for pattern to combine partial classes from different assembly's

I am having the following problem. I have a main project, and some extra projects that have similar functionality.
For example: I have an MVC website, then a class library project "A" with a "SettingsHelper". This just defines static wrappers for configuration settings so they can be used as propertys.
Then I have another class library project "B", which also contains a "SettingsHelper class".
How can I merge these SettingsHelpers in my main project, so I can use: SettingsHelper.Property from both modular extra projects.
I would like to be able to plug extra class libraries into one project.

Sounds pretty much like Dependency Injection. Normally you would expose SettingsHelper as an interface (your contract), and program against that. Then a DI container, such as Ninject, StructureMap, or Windsor would plug an implementation of that interface into the relevant parts of your code based on configuration.
This would allow you to code against a known contract and provide different libraries depending on the circumstances, the DI framework could then use that library to get the concrete implementation of the interface.
Would you need both instances at the same time?
Note that you cannot utilise the partial keyword across different assemblies, only within an assembly.
Update: based on your comment it sounds like you want to do something like Composition. Have a class that takes both classes from either library and combines them into one class that can be used by your application. Whether you then configure it to do something special or load the types when the libraries are present, it can all be encapsulated in this new class.
Update 2: alternatively, look into MEF:
http://msdn.microsoft.com/en-us/library/dd460648.aspx

That won't work. Partial classes cannot be divided over assemblies -- they don't exist in the CLR, only in the editor and the compiler. So they are compiled together into a single CLR class.
What you can do, is inherit one from the other. However, helpers tend to be static classes, so that won't work either.
The other alternative is not to write helper classes, but extension methods. You can extend classes in one assembly with methods defined in another assembly (or multiple other assemblies). See also http://msdn.microsoft.com/en-us/library/bb383977.aspx.

I would say that move both Helper classes in 3rd project and add reference of that project to both of your projects. So this new library will become shared datastructures and functionalities library.
Regards.

The specific pattern you are after is called a Facade Pattern. Unfortunately you are not going to get any help from the compiler getting this right. Essentially:
Create a new CombinedSettingsHelper class in your local assembly.
If the two SettingsHelper types are in the same namespace you will need to set up aliases for them (check the reference properties in the solution explorer, and MSDN documentation for this).
Implement the object so that it can access both SettingsHelper objects.
To clean up your facade you might try having a abstract method along the lines of abstract object GetSettingValue(string name);. Your facade could then inherit from the same base class and call these on its contained children. For example:
public abstract class SettingsHelperBase { public object GetSettingValue(string settingName); }
// Assembly1
public class SettingsHelper : SettingsHelperBase { }
// Assembly2
public class SettingsHelper : SettingsHelperBase { }
public class SettingsHelper : SettingsHelperBase
{
private List<SettingsHelperBase> _backends = new List<SettingsHelperBase>();
public readonly PropertiesImpl Properties;
public class PropertiesImpl
{
private SettingsHelper _settingsHelper;
public string Name
{
get
{
return (string)_settingsHelper.GetSettingValue("Name");
}
}
internal PropertiesImpl(SettingsHelper helper)
{
_settingsHelper = helper;
}
}
public SettingsHelper()
{
_backends.Add(asm1::MyNs.SettingsHelper);
_backends.Add(asm2::MyNs.SettingsHelper);
Properties = new PropertiesImpl(this);
}
protected override object GetSettingValue(string settingName)
{
foreach (var item in _backends)
{
var val = item.GetSettingValue(settingName);
if (val != null)
return val;
}
return null;
}
}

There is a way; Visual Studio allows the same code file to be included in more than one project.
When you do “Add”/”Existing Item” to can select a file that is in the different folder.
This is what some of the silver light support does so as to allow a “common class” that has some method that are only on the server and one methods that are only on the client.
(As to the question of “good design” you will have to decide that yourself, a lot of people don’t like having the same class compiled in different ways in different projects. Think if the mess you could get in with #if XXX, when XXX is only defined in one of the projects)

What are reasons why one would want to use nested classes? [duplicate]

This question already has answers here:
Why/when should you use nested classes in .net? Or shouldn't you?
(14 answers)
Closed 10 years ago.
In this stackoverflow answer a commenter mentioned that "private nested classes" can be quite useful so I was reading about them in articles such as this one which tend to explain how nested classes function technically, but not why you would use them.
I suppose I would use private nested classes for little helper classes that belong to a larger class, but often I will need a helper class from another class and so I would just have to take the extra effort to (1) make the nested class non-nested or (2) make it public and then access it with the outer-class prefix on it, which both seems to be extra work without any added-value for having the nested class in the first place. Hence in general I really don't see a use case for nested classes, other than perhaps to keep classes a bit more organized into groups, but I that also goes against the one-class-per-file clarity that I have come to enjoy.
In what ways do you use nested classes to make your code more manageable, readable, efficient?

You've answered your own question. Use nested classes when you need a helper class that is meaningless outside the class; particularly when the nested class can make use of private implementation details of the outer class.
Your argument that nested classes are useless is also an argument that private methods are useless: a private method might be useful outside of the class, and therefore you'd have to make it internal. An internal method might be useful outside of the assembly, and therefore you'd make it public. Therefore all methods should be public. If you think that's a bad argument, then what is different about you making the same argument for classes instead of methods?
I make nested classes all the time because I am frequently in the position of needed to encapsulate functionality in a helper that makes no sense outside of the class, and can use private implementation details of the outer class. For example, I write compilers. I recently wrote a class SemanticAnalyzer that does semantic analysis of parse trees. One of its nested classes is LocalScopeBuilder. Under what circumstances would I need to build a local scope when I am not analyzing the semantics of a parse tree? Never. That class is entirely an implementation detail of the semantic analyzer. I plan to add more nested classes with names like NullableArithmeticAnalyzer and OverloadResolutionAnalyzer that are also not useful outside of the class, but I want to encapsulate rules of the language in those specific classes.
People also use nested classes to build things like iterators, or comparators - things that make no sense outside of the class and are exposed via a well-known interface.
A pattern I use quite frequently is to have private nested classes that extend their outer class:
abstract public class BankAccount
{
private BankAccount() { }
// Now no one else can extend BankAccount because a derived class
// must be able to call a constructor, but all the constructors are
// private!
private sealed class ChequingAccount : BankAccount { ... }
public static BankAccount MakeChequingAccount() { return new ChequingAccount(); }
private sealed class SavingsAccount : BankAccount { ... }
and so on. Nested classes work very well with the factory pattern. Here BankAccount is a factory for various types of bank account, all of which can use the private implementation details of BankAccount. But no third party can make their own type EvilBankAccount that extends BankAccount.

Returning an interface to the caller whose implementation you want to hide.
public class Outer
{
private class Inner : IEnumerable<Foo>
{
/* Presumably this class contains some functionality which Outer needs
* to access, but which shouldn't be visible to callers
*/
}
public IEnumerable<Foo> GetFoos()
{
return new Inner();
}
}

Private helper classes is a good example.
For instance, state objects for background threads. There is no compelling reason to expose those types. Defining them as private nested types seems a quite clean way to handle the case.

I use them when two bound values (like in a hash table) are not enough internally, but are enough externally. Then i create a nested class with the properties i need to store, and expose only a few of them through methods.
I think this makes sense, because if no one else is going to use it, why create an external class for it? It just doesn't make sense to.
As for one class per file, you can create partial classes with the partial keyword, which is what I usually do.

One compelling example I've run into recently is the Node class of many data structures. A Quadtree, for example, needs to know how it stores the data in its nodes, but no other part of your code should care.

I've found a few cases where they've been quite handy:
Management of complex private state, such as an InterpolationTriangle used by an Interpolator class. The user of the Interpolator doesn't need to know that it's implemented using Delauney triangulation and certainly doesn't need to know about the triangles, so the data structure is a private nested class.
As others have mentioned, you can expose data used by the class with an interface without revealing the full implementation of a class. Nested classes can also access private state of the outer class, which allows you to write tightly coupled code without exposing that tight coupling publicly (or even internally to the rest of the assembly).
I've run into a few cases where a framework expects a class to derive from some base class (such as DependencyObject in WPF), but you want your class to inherit from a different base. It's possible to inter-operate with the framework by using a private nested class that descends from the framework base class. Because the nested class can access private state (you just pass it the parent's 'this' when you create it), you can basically use this to implement a poor man's multiple inheritance via composition.

I think others have covered the use cases for public and private nested classes well.
One point I haven't seen made was an answer your concern about one-class-per-file. You can solve this by making the outer class partial, and move the inner class definition to a separate file.
OuterClass.cs:
namespace MyNameSpace
{
public partial class OuterClass
{
// main class members here
// can use inner class
}
}
OuterClass.Inner.cs:
namespace MyNameSpace
{
public partial class OuterClass
{
private class Inner
{
// inner class members here
}
}
}
You could even make use of Visual Studio's item nesting to make OuterClass.Inner.cs a 'child' of OuterClass.cs, to avoid cluttering your solution explorer.

One very common pattern where this technique is used is in scenarios where a class returns an interface or base class type from one of its properties or methods, but the concrete type is a private nested class. Consider the following example.
public class MyCollection : IEnumerable
{
public IEnumerator GetEnumerator()
{
return new MyEnumerator();
}
private class MyEnumerator
{
}
}

I usually do it when I need a combination of SRP (Single Responsibility Principal) in certain situations.
"Well, if SRP is your goal, why not split them into different classes?" You will do this 80% of the time, but what about situations where the classes you create are useless to the outside world? You don't want classes that only you will use to clutter your assembly's API.
"Well, isn't that what internal is for?" Sure. For about 80% of these cases. But what about internal classes who must access or modify the state of public classes? For example, that class which was broken up into one or more internal classes to satisfy your SRP streak? You would have to mark all the methods and properties for use by these internal classes as internal as well.
"What's wrong with that?" Nothing. For about 80% of these cases. Of course, now you're cluttering the internal interface of your classes with methods/properties that are only of use to those classes which you created earlier. And now you have to worry about other people on your team writing internal code won't mess up your state by using those methods in ways that you hadn't expected.
Internal classes get to modify the state of any instance of the type in which they are defined. So, without adding members to the definition of your type, your internal classes can work on them as needed. Which, in about 14 cases in 100, will be your best bet to keep your types clean, your code reliable/maintainable, and your responsibilities singular.

They are really nice for, as an example, an implementation of the singleton pattern.
I have a couple of places where I am using them to "add" value, as well. I have a multi-select combobox where my internal class stores the state of the checkbox and the data item as well. no need for the world to know about/use this internal class.

Private anonymous nested classes are essential for event handlers in the GUI.
If some class is not part of the API another class exports, it must be made private. Otherwise you are exposing more than you intend. The "million dollar bug" was an example of this. Most programmers are too slack about this.
Peter

The question is tagged C# so I'm not sure this is of interest, but in COM you can use inner classes to implement interfaces when a class C++ implements multiple COM interfaces... essentially you use it for composition rather than multiple-inheritance.
Additionally in MFC and perhaps other technologies you might need your control/dialog to have a drop-target class, which makes little sense other than as a nested class.

If it is necessary for an object to return some abstract information about its state, a private nested class may be suitable. For example, if an Fnord supports "save context" and "restore context" methods, it may be useful to have the "save context" function return an object of type Fnord.SavedContext. Type access rules aren't always the most helpful; for example, it seems difficult to allow Fnord to access properties and methods of a Fnord.SavedContext without making such properties and methods visible to outsiders. On the other hand, one could have Fnord.CreateSaveContext simply create a New Fnord.SaveContext with the Fnord as a parameter (since Fnord.SaveContext can access the internals of Fnord), and Fnord.LoadContextFrom() can call Fnord.SaveContext.RestoreContextTo().

Organizing c# code into different files

I've gotten to a point where my main code file is about a thousand lines long and it's getting un-manageable; that is, I'm starting to get confused and not know where to locate some things. It's well-commented but there's just too much stuff.
I'd really like to be able to organize my code into different files, each with its own purpose. I want to get all the help VS gives me as I type when I edit these other files. A picture can say a thousand words:
alt text http://img64.imageshack.us/img64/7848/codeorganizationscreens.png
Is what I'm trying to do even possible?

Yes, but you need to be in the same namespace and declare the class just like you did in the main file, an example:
file1.cs
namespace Names
{
public partial class Hello
{
public void DoSomething() { }
}
}
file2.cs
namespace Names
{
public partial class Hello
{
public void Go() { DoSomething(); }
}
}

Although what other people say about partial classes is true. I'd also suggest you to analyze refactoring opportunities on your class.
If you're having problems to manage it, you could try to split your single class in several classes with less responsibilities.
IMHO partial classes may not help very much. Do you have your class separated in regions? Regions improve the readability of your code.

Yes you can split any partial class across as many files as you like.

Strip out each decent size class into at least one file. Wrap each class in the same namespace.
For large classes use either:
a. Region blocks eg
#region // Members
int my_int;
// other members...
#endregion
b. partial keyword to break a single class accross several files.

C# share code between classes

In Visual Studio 2008 using C#, what is the best way to share code across multiple classes and source files?
Inheritance is not the solution as the classes already have a meaningful hierarchy.
Is there some neat feature that's like a C include file that let's you insert code anywhere you want in another class?
EDIT:
ok, i guess we need a concrete example...
There are several hundred classes in the domain with a well thought out class heirarchy. Now, many of these classes need to print. There is a utility printer class that handles the printing. Let's say there are 3 different print methods that are dependent on the class that is being printed. The code that calls the print method (6 lines) is what I'm trying to avoid copying and pasting across all the different client class pages.
It'd be nice if people wouldn't assume they knew more about the domain that the op - especially when they specifically mention techniques that don't fit...

If you have functionality that you use frequently in classes that represent very different things, in my experience that should fall into just a few categories:
Utilities (e.g. string formatting, parsing, ...)
Cross-cutting concerns (logging, security enforcement, ...)
For utility-type functionality you should consider creating separate classes, and referencing the utility classes where needed in the business class.
public class Validator
{
public bool IsValidName(string name);
}
class Patient
{
private Validator validator = new Validator();
public string FirstName
{
set
{
if (validator.IsValidName(value)) ... else ...
}
}
}
For cross-cutting concerns such as logging or security, I suggest you investigate Aspect-Oriented Programming.
Regarding the PrintA vs. PrintB example discussed in other comments, it sounds like an excellent case for the Factory Pattern. You define an interface e.g. IPrint, classes PrintA and PrintB that both implement IPrint, and assign an instance of IPrint based on what the particular page needs.
// Simplified example to explain:
public interface IPrint
{
public void Print(string);
}
public class PrintA : IPrint
{
public void Print(string input)
{ ... format as desired for A ... }
}
public class PrintB : IPrint
{
public void Print(string input)
{ ... format as desired for B ... }
}
class MyPage
{
IPrint printer;
public class MyPage(bool usePrintA)
{
if (usePrintA) printer = new PrintA(); else printer = new PrintB();
}
public PrintThePage()
{
printer.Print(thePageText);
}
}

You can't just load in code that you'd like to have added into a class in C# via a preprocessor directive like you would in C.
You could, however, define an interface and declare extension methods for that interface. The interface could then be implemented by your classes, and you can call the extension methods on those classes. E.g.
public interface IShareFunctionality { }
public static class Extensions
{
public static bool DoSomething(this IShareFunctionality input)
{
return input == null;
}
}
public class MyClass : Object, IShareFunctionality
{
public void SomeMethod()
{
if(this.DoSomething())
throw new Exception("Impossible!");
}
}
This would allow you to reuse functionality, but you cannot access the private members of the class like you would be able to if you could, say, hash include a file.
We might need some more concrete examples of what you want to do though?

A C# utility class will work. It acts like a central registry for common code (or like the VB.NET Module construct) - it should contain code that's not specific to any class otherwise it should have been attached to the relevant class.
You don't want to start copying source code around if you don't have to because that would lead to code update problems considering the duplication.
As long as the source doesn't need to retain state, then use a static class with static method.
static public class MySharedMembers {
static public string ConvertToInvariantCase(string str) {
//...logic
}
// .... other members
}

If the classes are in the same namespace, there's no need for an include analog. Simply call the members of the class defined in the other function.
If they're not in the same namespace, add the namespace of the classes you want to use in the usings directives and it should work the same as above.
I'm confused by the question: it seems you need to work on your basic OO understanding.

Checkout extension methods: http://msdn.microsoft.com/en-us/library/bb383977.aspx

I don't know of a way to include portions of files but one thing we do frequently is to add an existing file and "link" it from its current location. For example, we have an assemblyInfo.cs file that every project refers to from a solution directory. We change it once and all the projects have the same info because they're referring to the same file.
Otherwise, suggestions about refactoring "common" routines in a common.dll are the best thing I've come up with in .Net.

I am not sure exactly what you mean by a "meaningful" structure already, but this sounds like a place where you could use base class implementation. Though not as "verbose" as C++ multiple inheritance, you might get some benefit out of using chained base class implementation to reuse common functions.
You can preserve class hierarchy, at least visually and override behavior as needed.

Pull out the repetitive code into services. The repetitive code is a clue that there might be some room for refactoring.
For example, create a "PrintingService" which contains the logic needed to print. You can then have the classes that need to print have a dependency on this service (either via the constructor or a parameter in a method which requires the service).
Another tip i have along these lines is to create interfaces for base functionality and then use the interfaces to code against. For example, i had bunch of report classes which the user could either fax, email, or print. Instead of creating methods for each, i created a service for each, had them implement an interface that had a single method of Output(). I could then pass each service to the same method depending on what kind of output the user wanted. When the customer wanted to use eFax instead of faxing through the modem, it was just a matter of writing a new service that implemented this same interface.

To be honest I can't think of anything like includes in Visual C#, nor why you would want that feature. That said, partial classes can do something like it sounds what you want, but using them maybe clashes against your "classes already have a meaningful hierarchy" requirement.

You have many options, TT, extension method, delegate, and lambda