Neo4j - Custom converter for field of type List - neo4j

I am trying to write a custom converter for a nested object so that this object gets saved as string in Neo4j database.
I am using #Convert annotation on my field and passing ImageConverter.class which is my AttributeConverter class.
Everything works fine as expected and I am able to save string representation of Image class in Neo4j db.
However, now instead of single image I want to have List<Image> as my nested field. In this case, putting #Convert(ImageConverter.class) doesn't work.
I see that there is a class called ConverterBasedCollectionConverter which gets used when I have a field of type List<LocalDateTime.
However, I couldn't find any exammples on how to use this class in case of custom converters.
Please can anyone help me with this or if there is any other approach to use custom converter on field of type List.
I am using Neo4j (version 3.4.1) and Spring-data-neo4j (5.0.10.RELEASE) in my application. I am also using OGM.
PS: I am aware that it is advised to store nested objects as separate node establishing a relationship with parent object. However, my use case demands that the object be stored as string property and not as separate node.
Regards,
V

It is not so difficult as I assumed it would be.
Given a class (snippet)
#NodeEntity
public class Actor {
#Id #GeneratedValue
private Long id;
#Convert(MyImageListConverter.class)
public List<MyImage> images = new ArrayList<>();
// ....
}
with MyImage as simple as can be
public class MyImage {
public String blob;
public MyImage(String blob) {
this.blob = blob;
}
public static MyImage of(String value) {
return new MyImage(value);
}
}
and a converter
public class MyImageListConverter implements AttributeConverter<List<MyImage>, String[]> {
#Override
public String[] toGraphProperty(List<MyImage> value) {
if (value == null) {
return null;
}
String[] values = new String[(value.size())];
int i = 0;
for (MyImage image : value) {
values[i++] = image.blob;
}
return values;
}
#Override
public List<MyImage> toEntityAttribute(String[] values) {
List<MyImage> images = new ArrayList<>(values.length);
for (String value : values) {
images.add(MyImage.of(value));
}
return images;
}
}
will print following debug output on save that I think is what you want:
UNWIND {rows} as row CREATE (n:Actor) SET n=row.props RETURN row.nodeRef as ref, ID(n) as id, {type} as type with params {type=node, rows=[{nodeRef=-1, props={images=[blobb], name=Jeff}}]}
especially the images part.
Test method for this looks like
#Test
public void test() {
Actor jeff = new Actor("Jeff");
String blobValue = "blobb";
jeff.images.add(new MyImage(blobValue));
session.save(jeff);
session.clear();
Actor loadedActor = session.load(Actor.class, jeff.getId());
assertThat(loadedActor.images.get(0).blob).isEqualTo(blobValue);
}

I am came up with a solution to my problem. So, in case you want another solution along with the solution provided by #meistermeier, you can use the below code.
public class ListImageConverter extends ConverterBasedCollectionConverter<Image, String>{
public ListImageConverter() {
super(List.class, new ImageConverter());
}
#Override
public String[] toGraphProperty(Collection<Image> values) {
Object[] graphProperties = super.toGraphProperty(values);
String[] stringArray = Arrays.stream(graphProperties).toArray(String[]::new);
return stringArray;
}
#Override
public Collection<Image> toEntityAttribute(String[] values) {
return super.toEntityAttribute(values);
}
}
ImageConverter class just implements AttributeConverter<Image, String> where I serialize and deserialize my Image object to/from json.
I chose to go with this approach because I had Image field in one object and List<Image> in another object. So just by changing #Convert(ListImageConverter.class) to #Convert(ImageConverter.class) I was able to save list as well as single object in Neo4j database.
Note: You can skip overriding toEntityAttribute method if you want. It doesn't add much value.
However you have to override toGraphProperty as within Neo4j code it checks for presence of declared method with name toGraphProperty.
Hope this helps someone!
Regards,
V

Related

Entity Framework Core 7 apply custom ValueConverter to Inserts and Updates

I am struggling to understand how to apply a custom ValueConverter to all the properties of all the models before every insert and update.
According to this resource it can be done:
https://learn.microsoft.com/en-us/ef/core/modeling/value-conversions?tabs=data-annotations
"Value converters allow property values to be converted when reading from or **writing ** to the database"
I was able to accomplish this, but only when reading from the database, overriding ConfigureConventions in the Context:
protected override void ConfigureConventions(ModelConfigurationBuilder configurationBuilder)
{
configurationBuilder.Properties<string>().HaveConversion<ToUpperConverter>();
}
Converter
public class ToUpperConverter : ValueConverter<string, string>
{
public ToUpperConverter() : base(v => v, v => v.Trim().ToUpper())
{
}
}
Tried also to override the method SaveChangesAsync in the Context, this works but obviously I don't want to go manually for every property and for every model considering they can change at every time.
Using reflection here can be an option, but I am not really a big fan of it.
public override async Task<int> SaveChangesAsync(CancellationToken cancellationToken = new CancellationToken())
{
foreach (var entry in ChangeTracker.Entries<Player>())
{
if (entry.State == EntityState.Modified || entry.State == EntityState.Added)
{
entry.Entity.Name = entry.Entity.Name.ToUpper();
entry.Entity.MiddleName = entry.Entity.MiddleName?.ToUpper();
entry.Entity.Surname = entry.Entity.Surname.ToUpper();
}
}
return await base.SaveChangesAsync(cancellationToken);
}
Solved, after spending on this an entire day..
The first argument of the converter is the conversion happening to the database, the second argument is the conversion happening from the database.
public class ToUpperConverter : ValueConverter<string, string>
{
public ToUpperConverter() : base(
// writing to the database
v => v.Trim().ToUpper(),
// reading from the database
v => v.Trim().ToUpper())
{
}
}
Source: https://learn.microsoft.com/en-us/dotnet/api/microsoft.entityframeworkcore.metadata.builders.propertiesconfigurationbuilder.haveconversion?view=efcore-6.0
"The type to convert to and from or a type that inherits from ValueConverter."

How to do group by key on custom logic in cloud data flow

I am trying to achieve the Groupby key based on custom object in cloud data flow pipe line.
public static void main(String[] args) {
Pipeline pipeline = Pipeline.create(PipelineOptionsFactory.create());
List<KV<Student,StudentValues>> studentList = new ArrayList<>();
studentList.add(KV.of(new Student("pawan", 10,"govt"),
new StudentValues("V1", 123,"govt")));
studentList.add(KV.of(new Student("pawan", 13223,"word"),
new StudentValues("V2", 456,"govt")));
PCollection<KV<Student,StudentValues>> pc =
pipeline.apply(Create.of(studentList));
PCollection<KV<Student, Iterable<StudentValues>>> groupedWords =
pc.apply(GroupByKey.<Student,StudentValues>create());
}
I just wanted to groupBy both the PCollection record based on the Student object.
#DefaultCoder(AvroCoder.class)
static class Student /*implements Serializable*/{
public Student(){}
public Student(String n, Integer i, String sc){
name = n;
id = i;
school = sc;
}
public String name;
public Integer id;
public String school;
#Override
public boolean equals(Object obj) {
System.out.println("obj = "+obj);
System.out.println("this = "+this);
Student stObj= (Student)obj;
if (stObj.Name== this.Name){
return true;
} else{
return false;
}
}
}
I have overridden the equals method of my custom class, but each time i am getting same instance of Student object to compare inside equals method.
Ideally it sholud compare first student key with second one.
Whats wrong i am doing here.
Why do you think you are doing anything wrong? The keys of each element are serialized (using the AvroCoder you specified) and the GroupByKey can group all of the elements with the same serialized representation together. After that it doesn't need to compare the students to make sure that the values with the same key have been grouped together.

Hibernate - how to generate id manually?

I am working on data migration from one db to a new one in Java.
One of my issues are that I must keep the same id values of the entities.
I know it's possible to change the id generator, but I need to set the id manually.
Each entity I create using Map from old object to a new one and I just set values.
can't I just set manually something like:
Map<String, Object> newEntity = new HashMap<String, Object>()
newEntity.setId(oldEntity.get("id"))
map the values using mapobject.keyset() meathod in map and then process the values accordingly
HashMap newmap = new HashMap();
// populate hash map here you can assign your map values
newmap.put(1, "tutorials");
newmap.put(2, "point");
newmap.put(3, "is best");
// get keyset value from map
Set keyset=newmap.keySet();
// check key set values
System.out.println("Key set values are: " + keyset);
You can set it manually the way you described in your code.
But you need to make sure the new entity knows this.
If you only add the #Id annotation on your primary key (no #GeneratedValue, ...) and you make sure the database is also not auto generating an id it should work.
I succeeded in doing it.
In my entity I set the following:
#Id
private String id
private Date created
private Date modified
static mapping = {
id column: 'id', generator: 'assigned'
}
public String getId() {
return id;
}
public void setId(String id) {
this.id = id;
}
public Date getCreated() {
return created;
}
public void setCreated(Date created) {
this.created = created;
}
public Date getModified() {
return modified;
}
public void setModified(Date modified) {
this.modified = modified;
}
and it worked. Thanks!

Want to save selected (i.e., more than 1) enums as string with NHibernate

I cannot for the life of me get this to work with my existing code, but I am trying to save my enum selections as strings in NHibernate. Basically, I have a UI check box and if the user selects multiple check boxes I want to store those selections. Now, I can get NHibernate to store ONE selection (e.g., from a drop down or radio button list, where the user is limited to one choice only).
This is the jist of what I have for an enum:
public enum IncomeType
{
[Display(Name = "Full-Time Employment")]
FullTime,
[Display(Name = "Part-Time Employment")]
PartTime,
[Display(Name = "Self-Employment")]
SelfEmployed,
[Display(Name = "Rental")]
Rental,
[Display(Name = "Social Security Payments")]
SocialSecurity,
[Display(Name = "Retirement / Pension Payments")]
Retirement,
[Display(Name = "Child Support Payments")]
ChildSupport,
[Display(Name = "Spousal Maintenance")]
Maintenance,
[Display(Name = "Other")]
Other
}
I use a method to "select" whether a checkbox list is shown (if my BulkItemThreshold equals the number of options, a checkbox list is displayed). Here is that method:
public static IEnumerable<SelectListItem> GetItemsFromEnumString<T>
(T selectedValue = default(T)) where T : struct
{
return from name in Enum.GetNames(typeof(T))
let enumValue = Convert.ToString((T)Enum.Parse(typeof(T), name, true))
select new SelectListItem
{
Text = GetEnumDescription(name, typeof(T)),
Value = enumValue,
Selected = enumValue.Equals(selectedValue)
};
}
(Note: some items in there are helpers, but I don't believe they are relevant; also, the selected input is displayed using a template .cshtml file - again, not sure if that's relevant)
Now, I call this thusly:
public class IncomeTypeSelectorAttribute : SelectorAttribute
{
public override IEnumerable<SelectListItem> GetItems()
{
return Selector.GetItemsFromEnumString<IncomeType>();
}
}
And finally, we get to the virtual property (using a proxy) but this is where NHibernate throws a wrench (Note: this was working fine for me before NHibernate, and now I am trying to get many lines of code working with it WITHOUT having to re-do everything; if I re-do everything I will probably triple the code I already have to get it to work):
Property (record):
[IncomeTypeSelector(BulkSelectionThreshold = 9)]
public virtual List<string> IndividualIncomeTypeCheckBox { get; set; }
proxy (part):
public List<string> IndividualIncomeTypeCheckBox
{
get { return Record.IndividualIncomeTypeCheckBox; }
set { Record.IndividualIncomeTypeCheckBox = value; }
}
Again, this is how I was doing things and it was working great before NHibernate. But now I have to use NHibernate. No getting around it.
I am using a service class that it tying the two together in a Create method to save in the DB with NHibernate, and for the above it would ordinarily look like this:
part.IndividualIncomeTypeCheckBox = record.IndividualIncomeTypeCheckBox;
This would work if it were just one selection.
Well, I've spent a good two (2) months trying to get this to work. It's tough because I have lots of code where the user can make only one selection (such as with a radiobutton list) and it works GREAT - even with NHibernate. Let me give you an example:
public virtual IncomeType? IndividualIncomeTypeCheckBox { get; set; }
If I do the above, it will display a drop down list, and NHibernate will store the ONE allowable option selected by the user in the DB no problem. But more than one option with List<string> does not work.
Now, I have tried everything I could find here or elsewhere and nothing works. Yes, I know it should be IList<IncomeType> or some other variant. But if I use this then NHibernate requires that IncomeType be another table in the DB. This is too much code to write for such a simple thing I believe. We are not talking a many-to-many relation in the sense that this is not a User with Multiple addresses (wherein addresses would have street, city, state, zip, etc.).
I have tried different types of proxy get and set code, but nothing works. I have tried [Flags] and other things working with string only, but to no avail. Those last solutions would "work" but ONLY to save the first item selected out of multiple (i.e., in my scenario, if the user selected "FullTime" and "Rental" as Income Types, then only "FullTime" (string) would be saved or "1" ([Flags]/int), not both items selected.
I have a situation where I re-display the choices using a ReadOnly attribute like this:
[IncomeTypeSelector]
[ReadOnly(true)]
public List<string> IndividualIncomeTypeCheckBoxPost
{
get { return IndividualIncomeTypeCheckBox; }
}
This would display on the UI, but I tried doing something like this with NHibernate and it wouldn't work.
Could anyone please show me, using the above, how I can go about getting NHibernate to store more than one enum in this checkbox list scenario?
UPDATE:
More poking around here and on the web, and I came up with the following (which still does not work).
Property (record):
[IncomeTypeSelector(BulkSelectionThreshold = 9)]
public virtual IList<IncomeTypeRecord> IndividualIncomeTypeCheckBox
{
get { return incomeType; }
set { incomeType= value; }
}
private IList<IncomeTypeRecord> incomeType =
new List<IncomeTypeRecord>();
Proxy (part):
public IList<IncomeTypeRecord> IndividualIncomeTypeCheckBox
{
get { return Record.IndividualIncomeTypeCheckBox; }
set { Record.IndividualIncomeTypeCheckBox= value; }
}
And a change to the enum:
public enum IncomeType : int // removing int & value still gives validate error
{
[Display(Name = "Full-Time Employment")]
FullTime = 1,
[Display(Name = "Part-Time Employment")]
PartTime,
....
}
And I added this class to support IncomeTypeRecord
public class IncomeTypeRecord
{
public virtual int Id { get; set; }
public virtual IncomeType Value { get; set; }
}
HOWEVER, when I get to the UI screen and pick one or more options I get a validation error (value not valid). For example, say I pick FullTime alone, or pick FullTime and Retirement, then the UI will display the following error:
The value 'FullTime' is invalid.
The value 'FullTime,Retirement' is invalid.
(respectively)
Even if I remove the int declaration for the enum and get rid of the value I started with "1", I still get this validation error. I tried messing around with and adding different model binders (which now has me stumped as to whether my original problem still exists and now I have a different problem - but you still get bounty points :) ).
Pulling my hair out. If I could offer more bounty I would. I need a definitive solution. I appreciate any help.
UPDATE
This is what I have so far:
Record:
public virtual string IndividualIncomeTypeCheckBox{ get; set; }
Part:
//If I do IEnumberable<string> my .Select throws a cast error
public IEnumerable<IncomeType> IndividualIncomeTypeCheckBox
{
get
{
return Record
.IndividualIncomeTypeCheckBox
.Split(',')
.Select(r => (IncomeType)Enum.Parse(typeof(IncomeType), r));
}
set { Record.IndividualIncomeTypeCheckBox= value
== null ? null : String.Join(",", value); }
}
Service Class:
public SimplePart CreateSimple(SimplePartRecord record)
{
SimplePart simple = Services.ContentManager.Create<SimplePart>("Simple");
...
//How I would save a FirstName property (example Part / PartRecord below)
//public virtual string FirstName { get; set; } - PartRecord
//public string FirstName - Part
//{
// get { return Record.FirstName ; }
// set { Record.FirstName= value; }
//}
simple.FirstName = record.FristName;
...
//I obviously cannot do the following with the above IncomeType
//Getting cannot convert string to IEnumerable error
//How would I write this:
simple.IndividualIncomeTypeCheckBox = record.IndividualIncomeTypeCheckBox;
...
}
And this is how it's called in a controller (this persists to DB): (Updating Controller code)
public ActionResult Confirm(string backButton, string nextButton)
{
if (backButton != null)
return RedirectToAction("WrapUp");
else if ((nextButton != null) && ModelState.IsValid)
{
_myService.CreateSimple(myData.SimplePartRecord);
return RedirectToAction("Submitted");
}
else
return View(myData);
}
Updating with additional code (serialization and view model):
"myData" is defined in the controller (using Serialization) as:
private MyViewModel myData;
protected override void OnActionExecuting(ActionExecutingContext filterContext)
{
var serialized = Request.Form["myData"];
if (serialized != null)
{
myData = (MyViewModel)new MvcSerializer().Deserialize
(serialized, SerializationMode.Signed);
TryUpdateModel(myData);
}
else
myData = (MyViewModel)TempData["myData"] ?? new MyViewModel();
TempData.Keep();
}
protected override void OnResultExecuted(ResultExecutedContext filterContext)
{
if (filterContext.Result is RedirectToRouteResult)
TempData["myData"] = myData;
}
I use Serialization because I set up a multi-step wizard (as seen in the controller action "backButton" "nextButton) on the front-end. I am not using a driver (which can only display Admin or on the front-end but then only on .cshtml files directly under the ~/Views folder (not in a structured folder list like I am using)). No driver = no update view model type code = no mechanism to "create" the data in the DB. If I do not use some "create" type method, the form will submit but all the data will be "NULL".
When you say that the data should be persisted automatically, I am sorry but I do not see how. All the stuff I read or code I review has SOME method of updating the DB with whatever is entered in a form. If I am missing something, my apologies.
"MyViewModel" is pretty straightforward:
[Serializabel]
public class MyViewModel
{
public SimplePartRecord SimplePartRecord { get; set; }
}
And, just in case, here is the relevant portion of the migration (return 1 is a completely separate and unrelated table):
public int UpdateFrom1()
{
SchemaBuilder.CreateTable("SimplePartRecord",
table => table
.ContentPartRecord()
...
.Column("IndividualIncomeTypeCheckBox", DbType.String)
...
);
ContentDefinitionManager.AlterPartDefinition("SimplePart",
part => part
.Attachable(false));
return 2;
}
The error I am getting is
Cannot implicitly convert type 'string' to 'System.Collections.Generic.IEnumerable'"
when I do the following in the "Create" method of my service class:
simple.IndividualIncomeTypeCheckBox = record.IndividualIncomeTypeCheckBox;
One additional thought: I tried using the n-n Relation sample to handle this scenario. Aside from it being a lot of extra code for what I thought should be straightforward and simple, because of the way I am using Serialization I had a lot of object reference errors and could not figure out how to properly code my controller to handle it.
There's a lot of info to wade through here so hopefully I haven't missed the point. It appears to me that the goals are:
The business class has a collection property of IList<IncomeType> without requiring an additional table
The values in that collection should be persisted as a delimited string of the enum names
The best approach is to use a custom user type (an implementation of NHibernate.UserTypes.IUserType) to map the property. Below is a generic IUserType that will map an enum of type T from an IList<T> property to a comma delimited string in the database and back again. There's no easy way to restrict T to an enum but the code will only work with enums.
Mapping a property using the custom type is simple with Fluent NHibernate:
public class Person
{
public Person()
{
IncomeTypes = new List<IncomeType>();
}
public virtual int PersonId { get; protected set; }
public virtual string FirstName { get; set; }
public virtual string LastName { get; set; }
public virtual IList<IncomeType> IncomeTypes { get; protected set; }
}
public class PersonMap : ClassMap<Person>
{
public PersonMap()
{
Table("Person");
Id(x => x.PersonId).GeneratedBy.Identity();
Map(x => x.FirstName);
Map(x => x.LastName);
Map(x => x.IncomeTypes).CustomType<EnumAsDelimitedStringType<IncomeType>>();
}
}
And here's the code for the user type:
public class EnumAsDelimitedStringType<T> : IUserType
{
public new bool Equals(object x, object y)
{
if (ReferenceEquals(x, y))
{
return true;
}
var xList = x as IList<T>;
var yList = y as IList<T>;
if (xList == null || yList == null)
{
return false;
}
// compare set contents
return xList.OrderBy(xValue => xValue).SequenceEqual(yList.OrderBy(yValue => yValue));
}
public int GetHashCode(object x)
{
return x.GetHashCode();
}
public object NullSafeGet(IDataReader rs, string[] names, object owner)
{
var outValue = NHibernateUtil.AnsiString.NullSafeGet(rs, names[0]) as string;
if (string.IsNullOrEmpty(outValue))
{
return new List<T>();
}
var getValueArray = outValue.Split(new[] {','}, StringSplitOptions.RemoveEmptyEntries);
return Array.ConvertAll(getValueArray, s => (T)Enum.Parse(typeof(T), s)).ToList();
}
public void NullSafeSet(IDbCommand cmd, object value, int index)
{
var inValue = value as IList<T>;
// set to string.Empty if you prefer to store that instead of null when the collection is null or empty
object setValue = null;
if (inValue != null && inValue.Any())
{
var setValueArray = Array.ConvertAll(inValue.ToArray(), v => Enum.GetName(typeof(T), v));
setValue = string.Join(",", setValueArray);
}
NHibernateUtil.AnsiString.NullSafeSet(cmd, setValue, index);
}
public object DeepCopy(object value)
{
return value;
}
public object Replace(object original, object target, object owner)
{
return original;
}
public object Assemble(object cached, object owner)
{
return cached;
}
public object Disassemble(object value)
{
return value;
}
public SqlType[] SqlTypes
{
get { return new[] {new SqlType(DbType.AnsiString)}; }
}
public Type ReturnedType
{
get { return typeof(IList<T>); }
}
public bool IsMutable
{
get { return false; }
}
}
I think you're on the right track pursuing a [Flags] enum. You may have done this, but just in case -- making an enum flags-worthy is more than adding the attribute. You also have to specify the value for the items in a binary-friendly manner. I've found the easiest way to do this is as follows:
[Flags]
public enum IncomeType : long // you'll need the room with several options
{
FullTime = 1,
PartTime = 1 << 1,
SelfEmployed = 1 << 2
// And so on
}
If you don't do this, then you'll get sequential integer values, which breaks the bitwise comparison that allows you to do multiple values in a single integer.
Your code to create the SelectList looks fine. Your options should construct form values that get posted back with the same name. If you want to use the default modelbinder, that means the associated property on your view model would need to be List<int>. If you're not using a view model (you probably should) you can pull it out of the forms collection.
Once you have this set up, then translating from your view model to your NHibernate entity is simple if a little annoying. You basically have to cycle through the values in the list and |= them onto your NHibernate entity's single enum property.
So let's assume you have a view model like this:
public class MyEditViewModel
{
public string Name { get; set; }
public List<int> IncomeSelections { get; set; }
// You'll probably have this to populate the initial view rendering
public SelectList AllIncomeOptions { get; set; }
}
You'll build your view using your helpers and all that, then build the checkboxes using the SelectList but making sure the input name is IncomeSelections, then when it's posted back you will push the view model data into your NHibernate entity something like this:
var myNHEntity = new NHEntity();
// If you're editing an existing entity, then be sure to reset the enum
// value to 0 before going into the following foreach loop...
foreach (var incomeSelection in viewModel.IncomeSelections)
{
myNHEntity.IncomeSelection |= incomeSelection;
}
There's probably a more clever way to do this, and you might have to cast the int to your enum type, but you'll figure that out (I'd do it for you, but it is Friday and I already have a beer open).
NHibernate should persist it without you having to do anything funky on the NH side.
In summary...
It seems like this is more a problem of how you handle the posted data than the NHibernate side. If you implement something like this, then be sure to use Fiddler or FireBug to inspect the posted values to make sure 1) they're integers and 2) the names are the same so they'll be added to the list.
Good luck!
The problem is simply that it won't be able to map a List without building a full relationship with an intermediate association table. It is way simpler to have the record store the values as a comma-separated string (so your record property is a string, not a list of string) and your part can map back and forth between string and List.
You can find an example of something very close here:
https://bitbucket.org/bleroy/nwazet.commerce/src/d722cbebea525203b22c445905c9f28d2af7db46/Models/ProductAttributesPartRecord.cs?at=default
https://bitbucket.org/bleroy/nwazet.commerce/src/d722cbebea525203b22c445905c9f28d2af7db46/Models/ProductAttributesPart.cs?at=default
It's not using enum values, instead it's a list of ids, but that should give you a good idea about how to make this work fairly simply: parsing enums you already know how to do.
Let me know if you need more details, but I think that's what you needed to get unblocked.

Linq to SQL using Repository Pattern: Object has no supported translation to SQL

I have been scratching my head all morning behind this but still haven't been able to figure out what might be causing this.
I have a composite repository object that references two other repositories. I'm trying to instantiate a Model type in my LINQ query (see first code snippet).
public class SqlCommunityRepository : ICommunityRepository
{
private WebDataContext _ctx;
private IMarketRepository _marketRepository;
private IStateRepository _stateRepository;
public SqlCommunityRepository(WebDataContext ctx, IStateRepository stateRepository, IMarketRepository marketRepository)
{
_ctx = ctx;
_stateRepository = stateRepository;
_marketRepository = marketRepository;
}
public IQueryable<Model.Community> Communities
{
get
{
return (from comm in _ctx.Communities
select new Model.Community
{
CommunityId = comm.CommunityId,
CommunityName = comm.CommunityName,
City = comm.City,
PostalCode = comm.PostalCode,
Market = _marketRepository.GetMarket(comm.MarketId),
State = _stateRepository.GetState(comm.State)
}
);
}
}
}
The repository objects that I'm passing in look like this
public class SqlStateRepository : IStateRepository
{
private WebDataContext _ctx;
public SqlStateRepository(WebDataContext ctx)
{
_ctx = ctx;
}
public IQueryable<Model.State> States
{
get
{
return from state in _ctx.States
select new Model.State()
{
StateId = state.StateId,
StateName = state.StateName
};
}
}
public Model.State GetState(string stateName)
{
var s = (from state in States
where state.StateName.ToLower() == stateName
select state).FirstOrDefault();
return new Model.State()
{
StateId = s.StateId,
StateName = s.StateName
};
}
AND
public class SqlMarketRepository : IMarketRepository
{
private WebDataContext _ctx;
public SqlMarketRepository(WebDataContext ctx)
{
_ctx = ctx;
}
public IQueryable<Model.Market> Markets
{
get
{
return from market in _ctx.Markets
select new Model.Market()
{
MarketId = market.MarketId,
MarketName = market.MarketName,
StateId = market.StateId
};
}
}
public Model.Market GetMarket(int marketId)
{
return (from market in Markets
where market.MarketId == marketId
select market).FirstOrDefault();
}
}
This is how I'm wiring it all up:
WebDataContext ctx = new WebDataContext();
IMarketRepository mr = new SqlMarketRepository(ctx);
IStateRepository sr = new SqlStateRepository(ctx);
ICommunityRepository cr = new SqlCommunityRepository(ctx, sr, mr);
int commCount = cr.Communities.Count();
The last line in the above snippet is where it fails. When I debug through the instantiation (new Model.Community), it never goes into any of the other repository methods. I do not have a relationship between the underlying tables behind these three objects. Would this be the reason that LINQ to SQL is not able to build the expression tree right?
These are non-hydrated queries, not fully-hydrated collections.
The Communities query differs from the other two because it calls methods as objects are hydrated. These method calls are not translatable to SQL.
Normally this isn't a problem. For example: if you say Communities.ToList(), it will work and the methods will be called from the objects as they are hydrated.
If you modify the query such that the objects aren't hydrated, for example: when you say Communities.Count(), linq to sql attempts to send the method calls into the database and throws since it cannot. It does this even though those method calls ultimately would not affect the resulting count.
The simplest fix (if you truly expect fully hydrated collections) is to add ToList to the community query, hydrating it.
Try adding another repository method that looks like this:
public int CommunitiesCount()
{
get { return _ctx.Communities.Count(); }
}
This will allow you to return a count without exposing the entire object tree to the user, which is what I think you're trying to do anyway.
As you may have already guessed, I suspect that what you are calling the anonymous types are at fault (they're not really anonymous types; they are actual objects, which you are apparently partially populating in an effort to hide some of the fields from the end user).

Resources