Wilson ORMapper Forums : Feature Requests

It's been a Year

Wed, 02 May 2007 02:50:09 GMT

Paul,

Did you ever consider using a secure, shared source repository for subscriber-only access? This would also help in getting people to contribute their own patches and such...

Brian
http://www.sidesofmarch.com/

Wilson ORMapper for: WAMPS? (Windows Server 2003 Apache MySQL Php or SQL Server ?

Tue, 01 May 2007 23:17:57 GMT

Hi Adam:

The short answer is that I'm a .NET consultant, and that's not going to change. Its not that I have anything against other platforms, but they don't pay my bills. Have you looked for O/RMs for LAMP? Java has Hibernate, and Ruby on Rails is very O/RM-like, so I would have expected something for LAMP too.

Good luck, Paul Wilson

Wilson ORMapper for: WAMPS? (Windows Server 2003 Apache MySQL Php or SQL Server ?

Tue, 01 May 2007 23:14:24 GMT

Hi!

By way of introduction: I am a registered user. I am grateful that I stumbled on your site when I was going through a period not long ago where I really wanted to start to understand OR mappers and DB abstraction layers, at least at a high level, as I had decided to learn how to do OO development, and not just BS about it, after being an SQL/VBScript web-enable-the-DB guy in monolithic IT shops (Digital, CPQ, HP) for many years - some other cool things here there too like stints @ two VC backed start-ups, one still going strong http://www.ide.com , where I learned a lot about process and dev teams through pain!

Part of this journey is that I have made the change form the Windows IT Stack to LAMP. I took a very modest position at a small, but filled with talent, consultancy in Cambridge MA http://www.isovera.com , that is entirely focused on app dev using Open Source tools & platforms specifically for non-profits. I am being allowed to work on becoming a competent OOD and LAMP-skilled person, and until then, they throw all the windows/SQL Server stuff at me and no one seems to complain.

My Question: I want to become a competent developer with OO data tiers & abstraction layers and OO frameworks, and to be able to recognize what make one OR Mapper implementation / or DB abstraction layer superior to another. I understand this is a long-term process. I also understand and have decided that this will require a non-billable project that will command a long-term committment.

I would like to know if you have plans to "port" your work to the LAMP platform, or make it "interoperable" in some way, and/or if you would give consideration to guiding me in the right direction to do the work that would be required to leverage your ORMapper and your ideas to function in a LAMP environment, the back-end DB can stay SQL Server, I like SQL Server!

I do apologize for the length of this email, but i wanted you to know where iwas coming from with my questons. Regards,

Adam

It's been a Year

Tue, 01 May 2007 23:14:03 GMT

I would like to do that -- I've made too many hopeful dates that didn't pan though, so all I can say is I will when I can.

Thanks, Paul Wilson

It's been a Year

Tue, 01 May 2007 23:12:57 GMT

Hi Paul,

It has been over a year since the last major update. I think there have been a few posts outlining some fixes. I would like the transaction rollback feature as well.

Do you think you can release a new version sometime soon?

Thanks,

Nick

SQL Server Compact Edition

Tue, 27 Mar 2007 18:44:33 GMT

http://ormapper.net/Forums/Default.aspx?part=74&action=thread&id=830&key=M2gBbeyCOJ5Neo8TgUi9ug%3d%3d

Never tried it myself but you would think it would work.

SQL Server Compact Edition

Mon, 26 Mar 2007 19:23:28 GMT

Yea, I realize there's not a hard-coded provider, but there is a CustomProvider option that is used for MySql and many other databases. So I'm wondering if you've tried that and had any luck or problems. That should work for most cases pretty well, at least that would be my guess, and I guess I'm hoping to avoid another hard-coded provider if it isn't necessary.

Thanks, Paul Wilson

SQL Server Compact Edition

Mon, 26 Mar 2007 15:30:21 GMT

hi Paul,

No, I haven't tried the SQL Server Compact Edition. There's no driver/provider provided (excuse the pun).

Regards,

Sean

SQL Server Compact Edition

Mon, 26 Mar 2007 12:38:03 GMT

So have you tried using the mapper with Sql CE? If not, I would recommend that you try. If you have, what worked and what didn't work?

Thanks, Paul Wilson

SQL Server Compact Edition

Mon, 26 Mar 2007 08:36:37 GMT

Hi Paul,

I've used SQLExpress for a good while now also. However, with SQLExpress, as with MSDE before that, there's a lengthy and often complicated installation process. SQL Server Compact Ed has practically no installation, as it's an inproc DB. In fact, the DLLs can be deployed with the application installation - no cumbersome database installation necessary!

I don't know why it would not be possible to use the generic ADO.NET classes SQLconnection, SQLCommand, etc. In fact, I think the latest version of the Enterprise Library (3.0) now encorporates SQLCE data classes.

Regards,

Sean

SQL Server Compact Edition

Sun, 25 Mar 2007 22:20:21 GMT

The ORMapper works with Sql Express, which I would have thought would be sufficient for desktop applications since it is free. There's probably something obvious I've missed, but what does Sql CE bring to the desktop that causes you to choose it? I'm curious for my own interest, but regardless of your answer, I would still like to think that I could support Sql CE possibly too. I haven't worked with Sql CE in a long time though, and never much, so anything you can tell me about what doesn't work with the ORMapper would be helpful.

Thanks, Paul Wilson

SQL Server Compact Edition

Sun, 25 Mar 2007 14:41:44 GMT

Is there any plans to make and/or test WORM against the SQL Server Compact Edition. I'm currently using the Compact Edition for a desktop application, and would benefit greatly from being able to use the ORM therein.

Many thanks,

Sean

Doing a read in a transaction

Thu, 21 Dec 2006 16:18:39 GMT

Suberb! That's great news. This is the only thing that bugs me during development with WORM so a fix will make your product pretty much perfect for me!

Thanks Paul!

Stephen.

Doing a read in a transaction

Sat, 25 Nov 2006 02:04:16 GMT

Looking forward to this feature.

It will nake cleaning up after a NUnit test alot easier. I follow the patterns suggested ny NewKirk and Vorontsov in their book Test Driven Development in Microsoft .NET

Doing a read in a transaction

Fri, 17 Nov 2006 14:28:23 GMT

Light at the end of the tunnel is always good :-)
Thanks for the reply, Dirk

Doing a read in a transaction

Thu, 16 Nov 2006 21:38:25 GMT

Hi Dirk:

I've recently changed jobs, partly to get back some of my free time, but I'm also still in that transitional mode that you always have when you start new jobs. But the good news is that I can see the light at the end of the tunnel now, so I think its very reasonable to expect an SP with that before the end of the year, and hopefully the next official release not too far after that.

Thanks, Paul Wilson

Doing a read in a transaction

Thu, 16 Nov 2006 13:44:23 GMT

Hi Paul,

Any estimate on a SP date with the Transaction.GetObjectXXX ?
No pressure, just informing whether I can wait or should get started on a work-around.

Thanks,
Dirk

Query Features

Tue, 19 Sep 2006 18:56:57 GMT

Hi,

I think this video has very good ideas (Query facilities) to put in this excellent ORMapper tool:
http://www.wekeroad.com/actionpackintro.html

Cheers

Add Composite Key Support to Generic ObjectSet

Wed, 30 Aug 2006 20:00:59 GMT

Paul,

Do you know when this issue might be cleared up? I ran into it via a Composite Key mapping via a Relation.

Thanks,

Chris Trout

EntityMap & FormatException

Fri, 14 Jul 2006 22:02:41 GMT

I'll consider this -- it look good. By the way, this code should only be executing on the initialization, which is one reason its not been optimized much.

Thanks, Paul Wilson

EntityMap & FormatException

Fri, 14 Jul 2006 22:01:28 GMT

The constructors for EntityMap contain the following code block:

try {
if (typeValue == null) {
this.typeValue = null;
}
else {
this.typeValue = int.Parse(typeValue);
}
}
catch {
this.typeValue = typeValue.Trim('\'');
}

If you happen to turn on the debugging feature to stop at every thrown exception, this code block becomes kind of annoying, because it gets hit fairly frequently in our app. I rewrote it as follows:

int tempInt;
if( typeValue == null )
    this.typeValue = null;
else if( int.TryParse( typeValue, out tempInt ) )
    this.typeValue = tempInt;
else
    this.typeValue = typeValue.Trim( '\'' );

This also avoids using exceptions as a flow-of-control construct in the code.

ObjectSpan / LoadSpan

Wed, 21 Jun 2006 02:40:43 GMT

Have you made any plans/decisions on whether or not supporting a LoadSpan/Fetch option is something you'll be adding to the mapper? As you know this is high on my list of features that I'd like to see incorporated into your mapper.

Optionally initialize relation

Tue, 20 Jun 2006 14:49:09 GMT

Hi Paul,

Trying my request again but logged in this time.

Currently the relations are always initialized in your object Instance, in its internal constructor if the related field represents a parent. I need this initialization to be optional, so the parent can be null (in my situation that's the default).

Ideally the OR mapper deduces this from whether the reference id field is nullable. This feature would really help me out, because the alternative would be quite ugly. After doing a GetObject, or StartTracking I would have to go over my whole object structure and set properties to null again.

If you agree to such a feature I'd be happy to implement it myself and send you the code.

Thanks,

Wout

e-mail: wout at woutware dot com

Doing a read in a transaction

Mon, 05 Jun 2006 16:06:46 GMT

Look forward to getting the new version with this in it.

Serializable ObjectSpace

Sat, 03 Jun 2006 00:07:40 GMT

Hey Doug:

Well I wouldn't ever say your design is unacceptable, its just different. I'll consider it a little more, as I do see your point. We can debate forever whether its better to be loosely coupled or more context aware, but in the end there are pros and cons both ways, so I'm inclined to provide you the option -- except that I'm pretty sure I won't be testing that option and I'll have to make that clear, since I'm not sure what impact it may have on performance or other things.

Thanks, Paul Wilson

Serializable ObjectSpace

Sat, 03 Jun 2006 00:03:47 GMT

Hi Paul,

I don’t see how this design is any better positioned for SOA than the one I suggested. One way or another, the Insert-Update-Unchanged state has to be tracked. Your design is better in a chatty environment with lots of persisted data, but requires the application to track that state. Mine allows reduced code and virtually eliminates coupling between the application and the ORM… it lets the ObjectSpace do what it is meant to do.

I had considered your design, but moved away from it because of a few unsavory implications:

- After fetching a complex graph, we would have to traverse the graph to mark each class as having come from the database. This state would have to be added to the each object in the graph, forcing the objects to change where (I would argue) they should not have to.

- Each object would be responsible for keeping track of when it changes, so it can distinguish between Unchanged or Updated. I get your point on always update, but that would not work for our auditing system. Another problem I have with this that it is prone to error, and would force us to change entity class code where (again I would argue) they should not need to change.

- Our remoting design calls for remoted classes to transparently run locally if configured so. Assuming we would want to be able to reuse the object space if possible, the changes required in the entity classes (discussed above) would have to be made aware of whether they are working in a remoted environment or not, and StartTracking should not be called if running locally.

I’m not convinced that our design is unacceptable. I am aware of the performance concerns, but I think we are OK for our environment. This will be validated in time. I got the idea from reading about how the NHibernate ISession is sometimes put onto CallContext.

We are prepared to maintain this change ourselves, so I’m not trying to push you to make a change you are not comfortable with. I just thought it might be a nice design option for your clients.

- Doug

Doing a read in a transaction

Tue, 30 May 2006 08:59:59 GMT

Its coming in the next release (this particular part is already mostly done, but I'm running very short on time getting the rest done lately, even to get a subscriber preview out).

Thanks, Paul Wilson

Doing a read in a transaction

Tue, 30 May 2006 08:57:09 GMT

I have come across two instances in my work where a read during a transaction to update is required. The ORMapper doesn't do this and it would be useful to allow this occur.

Serializable ObjectSpace

Sun, 28 May 2006 00:26:22 GMT

Hi Doug:

I'm actually suggesting still another option, which doesn't require either an ObjectSpace to be maintained across calls nor a new fetch of an existing object. The trick is to realize that the StartTracking method allows you to call specify that an object is an existing object that has been already been updated, thus allowing PersistChanges to work. This basically means that you can work "disconnected" from an ObjectSpace and re-establish a "connection" whenever you need to persist your work -- just call StartTracking. Note there is a difference between InitialState Unchanged and Updated -- specifying Unchanged without making any changes after that call will result in no persistence. Anyhow, you can always call StartTracking, specifying Updated, and know that your changes will be saved when you call PersistChanges, without the overhead of the same ObjectSpace or another fetch.

Thanks, Paul Wilson

Serializable ObjectSpace

Sun, 28 May 2006 00:15:03 GMT

Paul,

Thanks for responding.

We are centralizing database and other resource activity into remoting servers. Thick client, and public facing web servers use remoting servers for, say, all database activity.

To restate the probably obvious: A typical interaction is: Client requests widget from remoting server -> Remoting server fetches widget using WORM and serializes back to client -> Client allows user to play with widget and save changes back to server -> Server uses WORM to save changes with the context developed during the fetch.

I am shooting for a design where I can support a simple RemotingServer.FetchWidget() and RemotingServer.SaveWiget(wiget).

I’m guessing you are recommending that the remoting server’s Saves(wiget) re-fetches the widget using a new ObjectSpace and overlays the changes that were passed in. Did I guess right?. If so, I’d argue that the overlay functionality is an unappealing alternative to letting WORM do its work as if the same ObjectSpace created during the fetch was still available when Save was called.

I do not believe that the Fetch and Save are any less loosely coupled if the Save requires (or at least can take advantage of) the context that was developed during the fetch.

Our current design is to keep the ObjectSpace on the System.Runtime.Remoting.Messaging.CallContext class. This relieves the application from having to pass it around.

I certainly don’t want to do anything with obvious problems. I would love to hear your feedback if I am doing anything too crazy. I’m a bit new to ORM, but this design seemed like a natural way to get the full benefits of it in a remoted environment.

Thanks

- Doug

Assume insert if not tracked

Fri, 26 May 2006 18:14:44 GMT

I'm looking at a few cases similar to this, but I haven't done enough yet to promise this specific case.

Thanks, Paul Wilson

Assume insert if not tracked

Fri, 26 May 2006 18:12:46 GMT

Hi Paul,

I made a change to the Context class to help me avoid having to track wheter StartTracking had been called on each object. Basically, I think it is OK for me to assume that if StartTracking had not been called, then it is an insert. The assumption is made that the same instances hash table is used for selects and inserts. I am working a solution for this in a remoted environment... but that is another story.

I can continue to make this change in new ORMapper versions... but I thought I'd try to get a 'sanctioned' version of this change into the main code line.

For what it is worth, I believe NHibernate uses this approach.

Thanks,

- Doug

Here is the change:

public Instance this[object entityObject] {

get {

// Jeff Lanning (jefflanning@gmail.com): Made thread-safe (and improved performance a bit).

EntityKey entityKey = new EntityKey(this, entityObject, true);

Instance instance = (Instance)this.instances[entityKey];

if (instance == null) {

// HERE IS THE CHANGE: assume tracking an insert instead of throwing

this.StartTracking(entityObject, InitialState.Inserted);

entityKey = new EntityKey(this, entityObject, true); // guid keys get assigned at start tracking

instance = (Instance)this.instances[entityKey];

//END OF CHANGE

// throw new ORMapperException("ObjectSpace: Entity Object not being Tracked - " + entityKey);

}

instance.EntityObject = entityObject; // Refresh WeakReference

return instance;

}

Serializable ObjectSpace

Fri, 26 May 2006 18:12:17 GMT

Hi Doug:

I don't want to imply that its not doable, but I do think its fair to say that this would not be a typical design of a distributed app. While you can build them many ways technically, most of the time its advisable to stick to loosely coupled message-oriented systems (SOA is the current buzzword). Any thoughts? I suppose you could just argue that I can simply mark them serializable and let the user beware, but I've tried to avoid non-recommended possibilities so far. I also haven't tested what you're wanting, and since I don't currently see a reason to want to test that it puts me in a tough spot.

Thanks, Paul Wilson

Serializable ObjectSpace

Fri, 26 May 2006 18:07:12 GMT

Hi Paul,

We need to make the ObjectSpace serializable. We want the IsolatedContexts to endure across remoting calls. Yeah, we are taking the performance concerns into account. It would be helpful if these classes were marked as [Serializable] in your main code line. We have loosely tested this. There does not appear to be a need for special serialize code... just the class attribute

ORMapper
CustomProvider.cs
ObjectSpace.cs

Internals
Connection.cs
Context.cs
EntityKey.cs
Instance.cs

Internals.Mappings
AttributeMap.cs
ChildMap.cs
EntityMap.cs
FiledMap.cs
LookupMap.cs
ManyMap.cs
Mappings.cs
ParentMap.cs
RelationshipMap.cs

Internals.Providers
Commands.cs
MSCommands.cs

Thanks,

- Doug

Support For Subentities

Fri, 05 May 2006 00:46:56 GMT

Hi Phil:

Its something I might support in the future, and its gradually rising higher on my list as I knock of other things. But its so far not something I view as necessary either -- even the creators of Hibernate warn people that using the other type of "database inheritance" with their mapper can be very inefficient in many cases. Instead, in most cases where you need a more complicated database model, you also need to go to greater pains to model it efficiently and optimize it -- a process that tends to not be very generic, i.e. not one-size-fits-all.

Thanks, Paul Wilson

Support For Subentities

Fri, 05 May 2006 00:39:53 GMT

Hi,

I've seen the WIKI FAQ detailing the limitation that subentities must store their properties in the same table as the base class and I've seen a number of posts asking about this feature and workarounds to it (which I'm sure are what prompted the FAQ entry).

Can you give an indication as to whether this is something you will never support, something you might support in the future or something that is high on your list of changes to make and might be released shortly?

Thanks,

Phil

Resync IList collections request.

Tue, 25 Apr 2006 19:00:47 GMT

Hi Knave:

Hmmm, I'll look into it and try to have it fixed for the next release, probably appearing in a Subscriber Preview as soon as I can fit it in.

Thanks, Paul Wilson

Resync IList collections request.

Tue, 25 Apr 2006 18:59:07 GMT

Hi Paul,

When using the powerful Generics for .NET 2.0, I realized that when attempting to resync a generic collection, it will throw an error.

e.g. Manager.Resync( myObj.ListOfChildrens ); // where ListOfChildrens is of type IList<Child>.

Some food for thought, which perhaps you can incorporate in the next version?
Thanks.

Knave.

Add Composite Key Support to Generic ObjectSet

Sun, 23 Apr 2006 20:28:02 GMT

Hey Andy:

Hmmm, I don't guess anyone ever noticed the discrepancy before. And yes, I should be able to add that now that I know I need to do so. :)

Thanks, Paul Wilson

Add Composite Key Support to Generic ObjectSet

Sun, 23 Apr 2006 18:56:15 GMT

Paul,

I noticed that there is support for composite keys in ObjectSet, but not ObjectSet_T? Currently, when I attempt to RemoveByKey I get a null refence exception since it doesn't use the ArrayKeyComparer that was contributed by Peter Goetzl. When I changed the hashtable to use the ArrayKeyComparer everything seemed to work correctly. Any plans to add this to a future version?

Andy

Reverse Mapping...

Tue, 18 Apr 2006 15:32:46 GMT

I'm not sure I agree that's what it means, but if it is that then it won't happen. I'm a firm believer that an O/R Mapper, and for that matter any other runtime library, should not modify a database in any way. I do understand that there are scenarios where you would want that functionality, but I think it should be a separate tool, or your own code, that does that. Why? Because the vast majority of users would steer clear, and rightfully so, of any tool that has the potential to alter their database on the fly -- even if its just a possibility.

Thanks, Paul Wilson

Reverse Mapping...

Tue, 18 Apr 2006 15:27:00 GMT

I think he means to create the database tables from his classes in any supported database.

Timur Zanagar

ObjectSpan / LoadSpan

Tue, 11 Apr 2006 02:16:56 GMT

See latest news on the Subscriber Preview v4.2.2.1 for a few of these minor requests.

Sorting feature

Tue, 11 Apr 2006 02:15:14 GMT

See latest news on the Subscriber Preview v4.2.2.1.

Reverse Mapping...

Thu, 06 Apr 2006 22:22:03 GMT

What would that be? A way to get a Class.Alias name from a Table.Field name?

Thanks, Paul Wilson

Reverse Mapping...

Thu, 06 Apr 2006 22:20:26 GMT

Hi Paul,

I would like to have a Reverse Mapping in WORM. Is this a possible wish?

Thank you very much in advance.

Lazy load controlled

Fri, 31 Mar 2006 22:20:57 GMT

I understand you, but here's a good pattern that solve this problems :

public class Person
{
Guid m_PersonId;
String m_FirstName;
SaleOrderCollection m_SaleOrder;

public Guid PersonId {get{ return m_PersonId;} set {m_PersonId=value;}}
public String FirstName { get { return m_FirstName; } set { m_FirstName = value; } }
public SaleOrderCollection SaleOrder { get { return m_SaleOrder; } set { m_SaleOrder = value; } }

}

public interface IPersonFactory
{
void Load(Person objInfo, int nDepth);
void Insert(Person objInfo, int nDepth);
void Update(Person objInfo, int nDepth);
void Delete(Person objInfo, int nDepth);
void Save(PersonCollection objList, int nDepth);
void DeleteColl(PersonCollection objList, int nDepth);
}

I will try to create a 'Template' to generate the correct implementation of this.

Lazy load controlled

Fri, 31 Mar 2006 12:22:44 GMT

Hi Alexnaldo:

That of course isn't a problem with the default lazyLoad="true" case, but you are correct that when you disable lazy-loading then you lose that flexibility. This is one reason that distributed apps are much harder to design correctly -- its not necessarily a coding challenge as much as it is a design challenge. You either need to decide what you want and go with it at all times, or you need to create two different types that differ only in their relationships (not typically recommended, but listed here since it is an alternative). My advice is that if you almost always want the relationships then you just accept that you will always get them, or you go without them and get them with your own code only when needed. Which is best? That very much depends on your application and its usage scenarios -- thus the design challenge of distributed apps.

Thanks, Paul Wilson

Lazy load controlled

Fri, 31 Mar 2006 12:16:33 GMT

Hi Paul,

I have one relation Order->OrderDetails with lazyLoad=false

If I do : ObjectSpace.Save(..) I need to inform the "persistDepth", but isn't possible to do it :

ObjectSpace.GetObject(typeof(Order),PersistDepth.SingleObject)

Why I need it ?

Sometime I need all object graph, sometime not. It's very good for distributed application, because I need to controller the 'LazyLoad' for performance issue.

Thanks,
Alexnaldo Santos

Aggregate Feature?

Thu, 09 Mar 2006 00:37:05 GMT

Aggregate type attribute, something similar to a <lookup />, see psuedo example below:

<entity type="Order" table="Orders" keyMember="id">
...

This would be faster than loading the "orders" collection and calling customer.Orders.Count property I would think. Or instead of childFieldToAggregate, maybe use childMemberToAggregate or something? I realize this can be accomplished by using a View, but not all views are updateable. This member of course would be readonly.

Well aftering thinking about it more, I realize the data in said member would easily become dirty/outdated. As the OneToMany collection data changed, this aggregate member's data would not without resyncing the entity. :-(

But thought I would throw this out there anyway. It was one of those things that hit you in the shower.