|
You could make your life a HELL of a lot simpler by not attempting to reinvent the wheel - there are loads of DI containers around (Castle Windsor, Unity et al) - take a look and you will probably find it easier (and will definately find it quicker) to use...
C# has already designed away most of the tedium of C++.
|
|
|
|
|
Hi,
I have been trying to refactor the legacy code and make it loosly coupled.
I came across a class which resembles as shown below.
<pre lang="c#">
public Class MyClass
{
private int _a;
private ushort _b;
private int _c;
public MyClass(int a, ushort b, int c)
{
_a = a;
_b = b;
_c = c;
}
public static MyClass FromBytes(byte[] params)
{
int a = BitConverter.ToInt32(params, 0);
ushort b = BitConverter.ToUInt16(params, 1);
int c = BitConverter.ToInt32(params, 2);
return new MyClass(a, b, c);
}
public int A
{
get
{
return _a;
}
set
{
_a = value;
}
}
}
public Class MyConsumer
{
public void MyMethod()
{
// 1st way to create an instance
MyClass myClass1 = new MyClass(2, 3, 4);
int a = myClass1.A;
// 2nd way of creating an instance
byte[] bytes = {8, 7, 6};
MyClass myClass2 = MyClass.FromBytes(bytes);
int a1 = myClass2.A;
}
}
</pre>
'MyClass' has been created in two ways.
- creating an instance using a 'new' operator.
- Through a static method 'FromBytes'.
As seen above MyClass is tightly coupled to MyConsumer class.
I need to provide an abstraction for MyClass.
I thought of creating an Inteface 'IMyClass' which has a property 'A'.
<pre lang="c#">
interface IMyClass
{
int A { get; set;}
}
</pre>
MyClass shall have 2 overloaded constructor as shown below.
<pre lang="c#">
public Class MyClass : IMyClass
{
private int _a;
private ushort _b;
private int _c;
public MyClass(int a, ushort b, int c)
{
_a = a;
_b = b;
_c = c;
}
public MyClass(byte[] params)
{
_a = BitConverter.ToInt32(params, 0);
_b = BitConverter.ToUInt16(params, 1);
_c = BitConverter.ToInt32(params, 2);
}
public int A
{
get
{
return _a;
}
set
{
_a = value;
}
}
}
</pre>
I wanted to create an instance of MyClass in MyConsumer class through
some creational pattern such as Factory.
I don't know how to create an instance of a class which has its constructor
overloaded.
Questions:
- How to provide abstraction to static methods?
- How to create an instance of a class which has its constructor overloaded?
I found this link useful, however it does not server my purpose of creating an object through a factory.
Kindly share your view on this and let me know if you need more information.
Thanks in advance!
|
|
|
|
|
Constructors are a specialised kind of static methods. I often prefer a "normal" static method over a constructor when it has a specialised purpose, e.g. public static MyClass FromFile(string filename).
When it comes to the factory method with overloaded constructors of the classes, there are some possibilities:
- use an overloaded factory method, corresponding to the overloads of the constructors
- use an empty constructor and then set the properties
- create an extra constructor accepting an object[] args parameter
Personally I do not like any of those options, I doubt that the factory method is a good option in such a case.
|
|
|
|
|
Thanks Bernhard for the your time.
If we go with the static methods, I believe it would be tightly coupled and factory seems reasonable to me...
|
|
|
|
|
Use a factory.
The factory is derived from an interface.
The real factory uses the methods on the class that already exists.
The consumer object uses the real factory unless a static method (on option) is called to set a different factory.
|
|
|
|
|
Thanks jschell for your time.
As per my understanding based on your inputs, we need to create a factory.
So, in the sample code, it would be something like this.
interface IMyClassFactory
{
IMyClass Create(int a, ushort b, int c);
IMyClass Create(byte[] bytes);
}
and its implementation would be something like this.
public class MyClassFactory : IMyClassFactory
{
public IMyClass Create(int a, ushort b, int c)
{
return new MyClass(a, b, c);
}
public IMyClass Create(byte[] bytes)
{
int a = BitConverter.ToInt32(params, 0);
ushort b = BitConverter.ToUInt16(params, 1);
int c = BitConverter.ToInt32(params, 2);
return new MyClass(a, b, c);
}
}
Consumer:
public Class MyConsumer
{
public void MyMethod()
{
IMyClassFactory myClassFactory = new MyClassFactory(); // we can inject this
// 1st way to create an instance
IMyClass myClass1 = myClassFactory.Create(2, 3, 4);
int a = myClass1.A;
// 2nd way of creating an instance
byte[] bytes = {8, 7, 6};
MyClass myClass2 = myClassFactory.Create(bytes);
int a1 = myClass2.A;
}
}
Please correct me if I have misunderstood something here. Appreciate, if you can explain with a small example.
Is it acceptable to create an overloaded method in a factory as factory takes and parameter which defines the type of object required?
|
|
|
|
|
public static class FactorySource
{
public static IMyClassFactory Factory { get; set; }
static FactorySource()
{
Factory = new MyClassFactory();
}
}
...
public void MyMethod()
{
IMyClassFactory myClassFactory = FactorySource.Factory;
...
|
|
|
|
|
We've been asked to investigate porting some of our apps to iOS and Android and WinPhone. Currently we are using C++ and target Windows/Lunix (html only version seperate to this). We split the app into 3 layers already, UI, MainLogic and MachineAccess (database, disk, network etc). I accept that UI and MachineAccess are platform specific, and I probably need developers and resources for each platform, but how should I handle the "mainlogic" ?
This mainlogic layer is lots of lines (complex operations) and I dont really want to duplicate it and have to change it in multiple teams/codebases. Do I just have a mainlogic team that publishs the source to the platform teams? (publish and copy) This is kinda what we do for Win/unix but is there a better way? Having a single team on mainlogic is fine, I happy to break the workload along functional lines.
I've also heard that WinPhone wont allow C++/C in the future - any bright ideas how to handle that? Yes we could re-code to a different language, but I suspect sooner or later this problem will occur again.
|
|
|
|
|
Do you think that the mainlogic source will be different for the different platforms? Or only for certain portions?
If the code is (for the most part) the same:
Couldn't you use SCC and share the code over different projects?
If the code is different, could you use an automatic code converter?
We had worked out such a scenario for going from a dying language (VO) to .NET (Vulcan). During the transition period (a year) the code would be transported numerous times from VO to Vulcan until all of it would work. Parts would need a rewrite. The VO code would be adapted too if possible to make the transition possible. The VO code needed to produce valid apps all the time.
An alternative: can you make an abstract in meta language for the 'complex operations'? You can use a code-generator to create the code you need in any language.
Regards ... OttO
|
|
|
|
|
The mainlogic is identical on all platforms, it is essentially a pile of APIs and events. I had figured that SCC was the only way, but I thought there might be a better way now-a-days.
Rewriting the mainlogic away from C++ to another meta language is possible, but the question is then what is a nice future proof generic meta language?
To answer the next reply too - we do have a Web API layer already too (and HTML5 apps) - so I guess those platforms that cannot support C++ just have to use the Web API, unless there is a meta language that solves my problem.
Thanx
|
|
|
|
|
The beauty of a meta language is... you can create your own.
It can also be a pitfall.
Regards ... OttO
modified 2-Mar-12 1:07am.
|
|
|
|
|
If users will typically have internet access on their devices, the main logic could be wrapped up as a web service. No need to throw any of that away, just build the interface. If the apps don't need access to things like cameras, email, contacts etc, create a HTML5 web page(s). That way the apps are no more than a pointer to the web pages.
If this isn't an option, the only real commonality is potentially the webservice otherwise SQLite.
"You get that on the big jobs."
|
|
|
|
|
Richard Brett wrote: but how should I handle the "mainlogic" ?
1. Analyze the existing code
2. Define a limited subset of C++ that allows you implement the existing code. This involves restricting features that currently exist in C++ (such as very limited template use.)
3. If necessary re-write the logic to match the subset of 2.
4. Create translators for the other target languages. Basically cross compilers that translate from 2 to other languages.
5. Add 4 into your build process for the main logic to insure that non-supported features do not creep in.
2/4 can be complex or simple depending on what the code does and what the target languages are.
|
|
|
|
|
Hello 2 all!
I would like to run or participate in a project of compiler (or, maybe, a translator, or an interpretter). During last months I have gained some skills in this area and would like to enlarge them. Now I started writing a compiler of a C-like language right into the machine code (previously I wrote a translator of a C-like code into MASM, compiled programs worked well, so I do have some background). Though, it will be hard to make it alone, maybe by myself I can gather less clear ideas the compiled language could look like. So I would like to find a partner, or maybe just join some beginning project like mine. Any suggestions?
modified 27-Feb-12 17:49pm.
|
|
|
|
|
You might be better to post this in the Collaboration / Beta Testing forum.
"If you think it's expensive to hire a professional to do the job, wait until you hire an amateur." Red Adair.
nils illegitimus carborundum
me, me, me
|
|
|
|
|
my app uses SQLServer 2000,2005 or 2008R2 depending on the client.
At the start the app checks the table structure StoredProcedures etc.
The description has been stored in XML.
All checks are designed for 2000. Few things have been changed to support 2005 etc.
Now I want to take advantage of more recent stuff like VarChar(max) instead of TEXT.
What is the best way to do this?
Easy maintainable and still keep support for the different versions (90% of our clients still use SQLServer 2000).
Any tools, tips, tricks in SQLServer or external?
Regards .... OttO
|
|
|
|
|
Depends on your app and how you access data.
I suspect that in some situations that if you change the structure of a database then you will need to change the app to correctly access it.
If you don't need to do that then no problem. If you do then either you don't make such changes or you will need to provide a value that lets your app know how to vary depending on the structure of your database.
It might be relevant also to keep in mind that someone might upgrade their database server but not update the database structure. Also what do you think will be in the impact of incremental upgrades over time if you keep adding flags to the app?
|
|
|
|
|
The apps (it is a conglomerate of multiple exes in various versions) accesses the DB via ODBC.
The db structure is upgraded for higher versions of the app via an 'upgrade' tool. The tool is kept up to date and recognizes the current version of each apps own tables. It knows to what version the upgrade must be done per app, and knows the steps. It can change the structure, update the content etc.
The column type, size, default value, 'null allowed' is checked by the apps at startup. same goes for StoredProcedures, functions and nessecary views.
Upgrading SQLServer version hasn't been a major issue in the past. Only some minor differences have been adressed like the change in 'default value'.
Users etc are allowed to add indexes, views etc, but they arent allowed to change the apps tables by adding or changing types of columns.
Our clients can get twice a year a major build upgrade and in between sometimes a update if nessecary. No client follows the upgrades close by, only on a 'need' basis. Clients use at the moment ca 6 minor/major versions.
I get a feeling that not many applications check at startup the structure of the database they use. Is that correct?
Regards ... OttO
|
|
|
|
|
OChristiaanse wrote: I get a feeling that not many applications check at startup the structure of the
database they use
The structure itself? No.
I suspect the majority don't do any validation at all. The app assumes the structure is correct.
Myself I version the database via data to allow runtime checks. But I wouldn't check the schema.
|
|
|
|
|
You never got problems that a client updated the software (adding columns and tables) and after a crash restored a database from before the update?
How would you prevent this, other than checking the structure?
Regards ... OttO
|
|
|
|
|
OChristiaanse wrote: You never got problems that a client updated the software (adding columns and
tables)
Normally the systems I work on do not allow that. Not sure I have worked on any that did that.
OChristiaanse wrote: and after a crash restored a database from before the update?
If they added them and failed to restore them then that would be their problem.
OChristiaanse wrote: How would you prevent this, other than checking the structure?
I would start by not allowing that in the first place.
The first possible solution to allow for user defined data is to provide a meta data structure in the database, via appropriate tables, rather than allowing users to add their own.
Other than that how do you prevent them from adding something you don't know about? If you don't know about it no solution you come up with will allow you to detect if it is missing.
|
|
|
|
|
(was away for a few days)
Sorry, miscommunication: I meant to say:
The software was updated, and in the update proces columns and tables are added, and sometimes the content of some tables will be altered. The client doesnt add the columns etc, only performs the update of the software.
The client restores an older database. And as result of that, the application crashes after a while, for instance a month later. And I agree: it is the problem of the client, not ours in the first place. The problem is that the updatertool hasn't updated this backupdatabase. Because of this:
- a month work has been added in a not so healthy environment 
- corruption/inconsistencies in the db are possible.
Nevertheless, the client will come to us with questions like:
- Why didn't you prevent this from happening (thats why we added a startup dbstructure check )
- Can you fix this (yes of course we can try, but that will cost a lot of time, and you don't get any garantees).
- Can you garantee that no work has been lost (no ofcourse not... If you hadn't restored a prehistoric database, and the software isn't meant to work with that structure...)
I can't inmagine that software packages like Exact, etc don't perform somesort of db check at startup.
But maybe I'm to optimistic...
Regards ... OttO
|
|
|
|
|
I'm confused as to what the issue is here. The normal way to solve this issue is to have a decent backup solution implemented on the server. Why is this not appropriate for your problem?
|
|
|
|
|
My questions are:
Does your app check the db structure? And how?
And how do you deal with a change of database version like from SQLServer 2000 to 2008R2?
How do you take advantage of datatypes like VarChar(max) which is available in 2005 and further, but not in 2000, without forcing all clients to migrate to 2008R2?
I then got a question 'why would you check the db structure?' upon which I described a not so fictive situation WHY I would want to check the db structure. See my previous post. I've seen this kind of situations numerous times in the past, most of the time here in the house: 'I have a db, don't know the version, but it works' a few hours later: 'oh no, it doesn't work'
Backup's aren't the problem. Stupid users and ignorant developers are.
Regards ... OttO
|
|
|
|
|
But you talked about the user recovering from a catastrophic database failure by reinstalling the database. That is not the way out of the problem. Being able to seamlessly recover from a backup is the way to bypass this issue.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
