Preventing StackOverflowException while serializing Entity Framework object graph into Json - c#-4.0

I want to serialize an Entity Framework Self-Tracking Entities full object graph (parent + children in one to many relationships) into Json.
For serializing I use ServiceStack.JsonSerializer.
This is how my database looks like (for simplicity, I dropped all irrelevant fields):
I fetch a full profile graph in this way:
public Profile GetUserProfile(Guid userID)
{
using (var db = new AcmeEntities())
{
return db.Profiles.Include("ProfileImages").Single(p => p.UserId == userId);
}
}
The problem is that attempting to serialize it:
Profile profile = GetUserProfile(userId);
ServiceStack.JsonSerializer.SerializeToString(profile);
produces a StackOverflowException.
I believe that this is because EF provides an infinite model that screws the serializer up. That is, I can techincally call: profile.ProfileImages[0].Profile.ProfileImages[0].Profile ... and so on.
How can I "flatten" my EF object graph or otherwise prevent ServiceStack.JsonSerializer from running into stack overflow situation?
Note: I don't want to project my object into an anonymous type (like these suggestions) because that would introduce a very long and hard-to-maintain fragment of code).

You have conflicting concerns, the EF model is optimized for storing your data model in an RDBMS, and not for serialization - which is what role having separate DTOs would play. Otherwise your clients will be binded to your Database where every change on your data model has the potential to break your existing service clients.
With that said, the right thing to do would be to maintain separate DTOs that you map to which defines the desired shape (aka wireformat) that you want the models to look like from the outside world.
ServiceStack.Common includes built-in mapping functions (i.e. TranslateTo/PopulateFrom) that simplifies mapping entities to DTOs and vice-versa. Here's an example showing this:
https://groups.google.com/d/msg/servicestack/BF-egdVm3M8/0DXLIeDoVJEJ
The alternative is to decorate the fields you want to serialize on your Data Model with [DataContract] / [DataMember] fields. Any properties not attributed with [DataMember] wont be serialized - so you would use this to hide the cyclical references which are causing the StackOverflowException.

For the sake of my fellow StackOverflowers that get into this question, I'll explain what I eventually did:
In the case I described, you have to use the standard .NET serializer (rather than ServiceStack's): System.Web.Script.Serialization.JavaScriptSerializer. The reason is that you can decorate navigation properties you don't want the serializer to handle in a [ScriptIgnore] attribute.
By the way, you can still use ServiceStack.JsonSerializer for deserializing - it's faster than .NET's and you don't have the StackOverflowException issues I asked this question about.
The other problem is how to get the Self-Tracking Entities to decorate relevant navigation properties with [ScriptIgnore].
Explanation: Without [ScriptIgnore], serializing (using .NET Javascript serializer) will also raise an exception, about circular
references (similar to the issue that raises StackOverflowException in
ServiceStack). We need to eliminate the circularity, and this is done
using [ScriptIgnore].
So I edited the .TT file that came with ADO.NET Self-Tracking Entity Generator Template and set it to contain [ScriptIgnore] in relevant places (if someone will want the code diff, write me a comment). Some say that it's a bad practice to edit these "external", not-meant-to-be-edited files, but heck - it solves the problem, and it's the only way that doesn't force me to re-architect my whole application (use POCOs instead of STEs, use DTOs for everything etc.)
#mythz: I don't absolutely agree with your argue about using DTOs - see me comments to your answer. I really appreciate your enormous efforts building ServiceStack (all of the modules!) and making it free to use and open-source. I just encourage you to either respect [ScriptIgnore] attribute in your text serializers or come up with an attribute of yours. Else, even if one actually can use DTOs, they can't add navigation properties from a child object back to a parent one because they'll get a StackOverflowException.
I do mark your answer as "accepted" because after all, it helped me finding my way in this issue.

Be sure to Detach entity from ObjectContext before Serializing it.
I also used Newton JsonSerializer.
JsonConvert.SerializeObject(EntityObject, Formatting.Indented, new JsonSerializerSettings { PreserveReferencesHandling = PreserveReferencesHandling.Objects });

Related

Usage of implementsInterface element on entities in guidewire

I would like to know why do we use implementsInterface element in entities. I know one example where they use it to make it as assignable entity. But I could not understand what other purpose and how/why it is being used in entities.
Example: Injuryincident entity has claimantsupplier and coveragesupplier interface
I like to see it from this prespective, simplified and assuming that you have some java background:
As you probably already know it, having an entity means in the end of the day, having a Java class... Well, by using the implementsInterface element in your entity, is similar to implement an interface in you java class.
Here you have a quick example...
Consider the following:
MyEntiti.eti
<?xml version="1.0"?>
<entity
xmlns="http://guidewire.com/datamodel"
entity="MyEntity"
table="myentity"
type="retireable"/>
AnInterface.gs
package mypkg
interface AnInterface {
function doSomething()
}
AnInterfaceImpl.gs
package mypkg
class AnInterfaceImpl implements AnInterface {
override function doSomething() {
print("Hello!")
}
}
Image that you need MyEntity to have the ability of "doSomething", you just need to add the implementsInterface:
<?xml version="1.0"?>
<entity
xmlns="http://guidewire.com/datamodel"
entity="MyEntity"
table="myentity"
type="retireable">
<implementsInterface
iface="mypkg.AnInterface"
impl="mypkg.AnInterfaceImpl"/>
</entity>
By doing that, the following code must work:
var myEntity = new MyEntity()
myEntity.doSomething() //this will call the method defined in the interface-implementation
And even better, you migth let you implementation to recognize the related object of MyEntity and use it as per your needs:
package mypkg
class AnInterfaceImpl implements AnInterface {
private final var _relatedEntity : MyEntity
construct(relatedTo : MyEntity) {
_relatedEntity = relatedTo
}
override function doSomething() {
var createUser = _relatedEntity.CreateUser // you can accees to whatever you need
print("Hello!, this is the related instace of MyEntity: ${_relatedEntity}")
}
}
Hope it helps, regards!
I won't be repeating the other answer describing how it works, but I would like to mention how implementing an interface on an entity is different (and serves different purposes) compared to using enhancements.
On basic level both approaches let you add extra functionality to your entity classes. In most cases what you really want to do is just create/expand an enhancement - they are easier to write, more convenient to modify and just as effective when all you want is to just add a new function or calculated property.
When you implement an interface, you're bringing in some more serious guns. While this approach takes more work and requires creation of several files (not to mention modifying the entity itself), it gives you two important advantages over the enhancement mechanism:
The same interface can be implemented by several entities (typically each having its own implementation class) as well as non-entity classes. Objects of all such classes can then be used interchangeably in contexts expecting the interface (you can create an array of entity instances of several entities and even gosu-only wrappers/temporary objects and present it comfortably in the UI).
You can leverage polymorphism. While enhancement functions can't be overridden, the interface implementations allow you full flexibility of polymorphic OOP. You can, for example, set up a default "do nothing" implementation on high level entity that you intend to use and then add more meaningful implementations for specific subtypes meant to really make use of the new functionality.
It does have some overhead and complicates things, however. As mentioned - Enhancements are typically simpler. In practice you should ask yourself whether the extra effort of creating and implementing the interface is worth it - in many cases even situations seemingly calling for polymorphism can be handled well enough by a simple switch typeof this in the enhancement to provide all the necessary type-based logic.
In personal experience I've used interfaces in quite a few situations, but Enhancements are my first choice in overwhelming majority of cases.
As a final note I'd like to mention a delegate entity. If what you want to add to some unrelated entities is not functionality but Properties with underlying database fields, creating a delegate entity and "implement" it with the desired standalone entities. A delegate entity does work a bit like an interface (you can use entity objects implementing the delegate interchangeably in situations where the delegate is expected) and you can set-up both interface implementation and enhancements on delegate level as well.

Extending a JOOQ Table class

I have a 'document' table (very original) that I need to dynamically subset at runtime so that my API consumers can't see data that isn't legal to view given some temporal constraints between the application/database. JOOQ created me a nice auto-gen Document class that represents this table.
Ideally, I'd like to create an anonymous subclass of Document that actually translates to
SELECT document.* FROM document, other_table
WHERE document.id = other_table.doc_id AND other_table.foo = 'bar'
Note that bar is dynamic at runtime hence the desire to extend it anonymously. I can extend the Document class anonymously and everything looks great to my API consumers, but I can't figure out how to actually restrict the data. accept() is final and toSQL doesn't seem to have any effect.
If this isn't possible and I need to extend CustomTable, what method do I override to provide my custom SQL? The JOOQ docs say to override accept(), but that method is marked final in TableImpl, which CustomTable extends from. This is on JOOQ 3.5.3.
Thanks,
Kyle
UPDATE
I built 3.5.4 from source after removing the "final" modifier on TableImpl.accept() and was able to do exactly what I wanted. Given that the docs imply I should be able to override accept perhaps it's just a simple matter of an erroneous final declaration.
Maybe you can implement one of the interfaces
TableLike (and delegate all methods to a JOOQ implementation instance) such as TableImpl (dynamic field using a HashMap to store the Fields?)
Implement the Field interface (and make it dynamic)
Anyway you will need to remind that there are different phases while JOOQ builds the query, binds values, executes it etc. You should probably avoid changing the "foo" Field when starting to build a query.
It's been a while since I worked with JOOQ. My team ended up building a customized JOOQ. Another (dirty) trick to hook into the JOOQ library was to use the same packages, as the protected identifier makes everything visible within the same package as well as to sub classes...

Retrieving a value object without Aggreteroot

I'm developing an application with Domain Drive Design approach. in a special case I have to retrieve the list of value objects of an aggregate and present them. to do that I've created a read only repository like this:
public interface IBlogTagReadOnlyRepository : IReadOnlyRepository<BlogTag, string>
{
IEnumerable<BlogTag> GetAllBlogTagsQuery(string tagName);
}
BlogTag is a value object in Blog aggregate, now it works fine but when I think about this way of handling and the future of the project, my concerns grow! it's not a good idea to create a separate read only repository for every value object included in those cases, is it?
anybody knows a better solution?
You should not keep value objects in their own repository since only aggregate roots belong there. Instead you should review your domain model carefully.
If you need to keep track of value objects spanning multiple aggregates, then maybe they belong to another aggregate (e.g. a tag cloud) that could even serve as sort of a factory for the tags.
This doesn't mean you don't need a BlogTag value object in your Blog aggregate. A value object in one aggregate could be an entity in another or even an aggregate root by itself.
Maybe you should take a look at this question. It addresses a similar problem.
I think you just need a query service as this method serves the user interface, it's just for presentation (reporting), do something like..
public IEnumerable<BlogTagViewModel> GetDistinctListOfBlogTagsForPublishedPosts()
{
var tags = new List<BlogTagViewModel>();
// Go to database and run query
// transform to collection of BlogTagViewModel
return tags;
}
This code would be at the application layer level not the domain layer.
And notice the language I use in the method name, it makes it a bit more explicit and tells people using the query exactly what the method does (if this is your intent - I am guessing a little, but hopefully you get what I mean).
Cheers
Scott

How to access certain EStructuralFeatures of an EMF Model?

I know that there are ways to access an EAttribute of an Eclipse EMF model by its featureID or by its name via different indirect approaches. For that I found the following: Eclipse EMF: How to get access EAttribute by name?
But what if I don't know the name of the attribute I want to get? Let's say, based on the design, the model has some fixed attributes by the developer, along with the features that can be set dynamically by the user.
So, for the time being I use the getEAllStructuralFeatures() and use indexes via get() to reach to the by-the-user-created attributes, since I know that the list I get will have the fixed attributes of the model as its first elements beginning with the index 0. But I find this solution unclear and inefficient. Also in some cases, that I want to work, not suitable.
E.g: IEMFEditProperty prop = EMFEditProperties.list(editingDomain, EMFMODELPackage.Literals.EMFMODEL.getEAllStructuralFeatures().get(X));
Do you know a solution or a workaround for this problem? As far as I can see, there are no direct methods to get such dynamically created features of the model.
Every help will be appreciated.
I have been working on a similar case recently where I first tried to define an EStructuralFeature to access exactly the setting/attribute of the object that I needed.
But if you look at how things work internally in ECore, you will find out, that there is no way this will ever work, since the indices are bound to the object identity of the EStructuralFeature objects created at runtime for the specific context (i.e. EClass instance).
My approach was then to either inspect the features proposed by EClass.getEAllStructuralFeatures or to iterate over the features and inspect the object returned by EObject.eGet for this very feature (where EClass eClass = eObject.eClass()).
Example: In a UML profile I have defined a UML Stereotype called "Bean" with a property called FactoryEntity. The property shall reference a UML Class with the Stereotype "Entity" that is closest to this very bean and for which a static factory method will be generated.
In the model I would then have one UML Class typed as Bean and one as Entity.
And for the Class typed as "Bean" I would then set a value for the attribute/property factoryEntity defined in the profile.
The question was then how the property value would be accessible in ECore. I ended up iterating the List of available EStructuralFeature of the EClass of the EObject and checking the type of the object returned by eGet.
final EObject eObject = (EObject) holdingClass.getValue(stereotype, stereoTypePropertyName);
final EList<EStructuralFeature> allEStructFeats = eObject.eClass().getEAllStructuralFeatures();
for(EStructuralFeature esf : allEStructFeats)
{
final Object o = eobject.eGet(esf);
if(o instanceof org.eclipse.uml2.uml.Class)
{
return (org.eclipse.uml2.uml.Class) o;
}
}
Maybe that is not the most elegant way to access structural features but it is the only one I thought was robust enough to last.
Please let me know if you have any suggestions on how to improve this.

Regarding MOSS or WSS 3.0, What parts of the API might have been implemented better?

Such as:
Sealed Methods you might have liked to extend
Exceptions thrown are more vague than is helpful
Elimination of Connected Content which was a major feature in MCMS 2002
HTML is stripped from fields when stored and returned. No easy option to work around this problem
Creating an SPWeb takes an eternity.
Nonexistant migration path from MCMC 2002
I wish that the Sharepoint object model was purely managed code. Although having .NET wrappers is convenient, having to worry about disposing the many objects that implement IDisposable is a pain. It's so easy to run into memory issues when dispose does not get called in a WSS app. And I thought the reason for moving to .NET was to free developers from having to deal with memory management...
How about refactoring Properties that result in additional database calls to methods instead, for example the Items property on SPList.
Any of the SPList API could use a complete rewrite. Trying to deal with libraries with nested folders is a complete nightmare with the list being completely flattened with no obvious hierarchical structure.
Another wonderful addition would be adding interfaces to SPWeb, SPList and other Sharepoint classes to aid testing.
Has anyone seen this method:
http://msdn.microsoft.com/en-us/library/microsoft.sharepoint.spsecurity.runwithelevatedprivileges.aspx This method shows the unbelievable nonsense that Sharepoint exposes to developers.
My personal favourite is the SPField.GetFieldValue Method. I have no idea why they designed it the way they did, but to me it does hardly make sense. To get a object out of a ListItem you have to do somethine like:
SPField field = ((SPList)list).Fields.GetField("FieldName");
object fieldValue = field.GetFieldValue(((SPListItem)item)[field.Title].ToString());
Getting an object out of a ListItem is IMO a basic operation, so this should not be that complicated.
Inconsistencies when passing field names to methods or arrays. For example:
SPFieldCollection.ContainsField(): Internal name or display name
SPFieldCollection.GetField(): Internal name or display name
SPFieldCollection.GetFieldByInternalName(): Internal name
SPFieldCollection.Item: Display name
SPListItem.Item: Internal name or display name
SPListItem.GetFormattedValue(): Internal name or display name
SPViewFieldCollection.Exists: Internal name
To put the icing on the cake, there is usually no documentation about whether a method takes internal and/or display name.

Resources