I am using NHibernate in a web application I'm building. The user can subscript to zero or more mailing-lists (there are a total of 8). This is represented on the screen with a checkbox for each mailing-list.
I would like to use NHibernate to update these in one go. A very simple sql query would be:
update mail_subscriptions set subscribed = true where mailing_list_id in (21,14,15,19) and user_id = 'me'
What is the cleanest way to perform this update via NHibernate so that I can make a single round trip to the database?
Thanks in advance
JP
NHibernate might not be able to update the mail_subscriptions in the way you have shown above but it can do it in a single round trip to the DB using batched queries.
This example considers Subscriptions mapped as a HasMany using Component although roughly the same technique can be used if the mapping was just a plain HasMany. I am also assuming that each user already has rows in the mail_subscriptions table for each mailing list set to false for subscribed.
public class User{
public virtual string Id {get; set;}
public virtual IList<MailSubscription> Subscriptions {get; set;}
}
public class MailSubscription{
public virtual int ListId {get; set;}
public virtual bool Subscribed {get; set;}
}
public void UpdateSubscriptions(string userid, int[] mailingListIds){
var user = session.Get<User>(userid);
foreach(var sub in
user.Subscriptions.Where(x=> mailingListIds.Contains(x.ListId))){
sub.Subscribed=true;
}
session.Update(user);
}
Now when the unit of work completes you should see SQL like this produced sent as a single round trip to the DB.
update mail_subscriptions set subscribed=true where user_id='me' and listid=21
update mail_subscriptions set subscribed=true where user_id='me' and listid=14
update mail_subscriptions set subscribed=true where user_id='me' and listid=15
update mail_subscriptions set subscribed=true where user_id='me' and listid=19
I think the NHibernate feature you seek is known as Executable DML.
Ayende has a blog post giving an example at http://ayende.com/blog/4037/nhibernate-executable-dml .
Depending on your names of your entities and their properties, and assuming you have an ISession instance variable called session, you would need to execute an HQL query something like:
session.CreateQuery("update MailSubscriptions set Subscribed = true where MailingList.Id in (21,14,15,19) and User.Id = 'me'")
.ExecuteUpdate();
Now, having said that, I think in the use case you describe (updating a handful of entries within a collection on a single aggregate root), there is no need to use Executable DML. Mark Perry has the right idea - you should simply modify the booleans on the appropriate entities and flush the session in the usual way. If ADO.NET batching is configured appropriately, then the child entries will cause multiple update statements to be sent to the RDBMS in a single database call.
Related
In LINQ I have written a simple query where I am searching for an animal using the ID property. However, I am also including the Farm the animal belongs using the Include property.
I want to write the same LINQ query in SQL where I can include Farm. How can I include Farm using SQL. I have an incomplete SQL syntax below. Can anyone help me out.
LINQ
await _dbContext.Animals.Where(x => x.id == 1)
.Include(x => x.Farm)
.ToListAsync();
SQL
select * from Animals where id = 1;
Apparently your database has a table with Animals and a table with Farms. There seems to be a one-to-many relation between Animals and Farms: on every Farm live zero or more Animals; every Animal lives on exactly one Farm, namely the Farm that the foreign key refers to.
I think you will have classes similar to the following:
class Farm
{
public int Id {get; set;}
public string Name {get; set;}
... // etc
// Every Farm has zero or more Animals (one-to-many)
public virtual ICollection<Animal> {get; set;}
}
class Animal
{
public int Id {get; set;}
public string Name {get; set;}
... // etc
// Every Animal lives on exactly one Farm, using foreign key
public int FarmId {get; set;}
public virtual Farm Farm {get; set;}
}
I want to write the same LINQ query in SQL where I can include Farm.
A small trick: if you want to know the SQL code generated by Entity Framework, use property DbContext.Database.Log.
using (var dbContext = new DbContext())
{
// Log generated SQL to debug window:
dbContext.Database.Log = System.Diagnostics.Debug.Write;
// execute your LINQ:
var fetchedAnimals = _dbContext.Animals.Where(x => x.id == 1)
.Include(x => x.Farm)
.ToList();
}
Write your own SQL
You'll have to join Animals with Farms, and keep only the Animal with ID = 1:
See SQL Join
// Select only the properties of Animals and Farms that you actually plan to use
SELECT Animals.Id, Animals.Name, ...,
Farms.Id, Farms.Name, ...
FROM Animals INNER JOIN Farms
ON Animals.FarmId = Farm.Id
WHERE Animals.Id = 1
You should not use "" to fetch everything. If Farm [10] Has 5000 Chickens, then every Chicken will have a foreign key with a value 10. If you use "" you will transfer this value 10 more than 5000 times, while you already know the value of the foreign key.
There's room for improvement
When using entity framework to fetch data, always use Select, and select only the properties that you plan to use, even if you Select all properties. Only omit Select and / or use Include if you plan to change / update the fetched data.
The reason is, that fetching data without using Select is not very efficient.
If you fetch data without using Select, entity framework will put the fetched item in the DbContext.ChangeTracker, together with a copy of the fetched item. You get a reference to the copy. Whenever you change properties of the fetched item, you change the copy in the ChangeTracker. When you call DbContext.SaveChanges, the original is compared with the copy, property per property to see which properties are changed, and thus need to be updated in the database.
So if you don't plan to change the fetched data, it would be a waste of processing power to put this data AND a copy in the ChangeTracker. Hence: always use Select, unless you plan to update the fetched data.
I have a domain model that includes something like this:
public class Customer : EntityBase<Customer>, IAggregateRoot
{
public IList<Comment> Comments { get; set; }
}
public class Comment : EntityBase<Comment>
{
public User CreatedBy { get; set; }
public bool Private { get; set; }
}
I have a service layer through which I retrieve these entities, and among the arguments passed to that service layer is who the requesting user is.
What I'd like to do is be able to construct a DetachedCriteria in the service layer that would limit the Comment items returned for a given customer so the user isn't shown any comments that don't belong to them and are marked private.
I tried doing something like this:
criteria.CreateCriteria("Comments")
.Add(Restrictions.Or(Restrictions.Eq("Private", false),
Restrictions.And(Restrictions.Eq("Private", true),
Restrictions.Eq("CreatedBy.Id", requestingUser.Id))));
But this doesn't flow through to the lazy-loaded comments.
I'd prefer not to use a filter because that would require either interacting with the session (which isn't currently exposed to the service layer) or forcing my repository to know about user context (which seems like too much logic in what should be a dumb layer). The filter is a dirty solution for other reasons, too -- the logic that determines what is visible and what isn't is more detailed than just a private flag.
I don't want to use LINQ in the service layer to filter the collection because doing so would blow the whole lazy loading benefit in a really bad way. Lists of customers where the comments aren't relevant would cause a storm of database calls that would be very slow. I'd rather not use LINQ in my presentation layer (an MVC app) because it seems like the wrong place for it.
Any ideas whether this is possible using the DetachedCriteria? Any other ways to accomplish this?
Having the entity itself expose a different set of values for a collection property based on some external value does not seem correct to me.
This would be better handled, either as a call to your repository service directly, or via the entity itself, by creating a method to do this specifically.
To fit in best with your current model though, I would have the call that you currently make to get the the entities return a viewmodel rather than just the entities;
public class PostForUser
{
public Post Post {get; set;}
public User User {get; set;}
public IList<Comment> Comments}
}
And then in your service method (I am making some guesses here)
public PostForUser GetPost(int postId, User requestingUser){
...
}
You would then create and populate the PostForUser view model in the most efficient way, perhaps by the detached criteria, or by a single query and a DistinctRootEntity Transformer (you can leave the actual comments property to lazy load, as you probably won't use it)
With my only ORM knowledge being L2S/EF, I was surprised when the following code inserted a row into the database before I called repo.Save:
var repo = new UserRepository();
var user = new User { Name = "test" }
repo.Add(user);
//repo.Save();
Repo looks like this:
public void Add(T entity)
{
session.Save(entity);
}
public void Save()
{
session.Flush();
}
After some digging, it seems NHibernate needs to make the insert happen right away in order to get the ID of the new entity (since it's using an auto increment ID). But L2S/EF doesn't work like this; I can add many entities and save them all at the end.
Question is: is there a way to achieve the same thing with NHibernate, while still using auto increment IDs, and out of interest does anyone know why it works like this?
Fabio Maulo already blogged about the usage of identity generator a few times. The answer is: use hilo, guid.comb or something like this.
NHibernate needs the identity because every entity in the session (they are called "persistent entities") needs to be identified. The identity is also normally used to determine if the record already exists in the database (unsaved value).
session.Save actually only makes a transient entity persistent. When the database is generating the id, it needs to be stored to get the id. If NH can create the id itself (eg using hilo), it could be stored next time when the session gets flushed.
I am working on an NHibernate project and have a question regarding updating transient entities.
Basically the workflow is as follows:
Create a DTO (projection) and send over the wire to client. This has a small subset of properties from the entity.
Client sends back the changed DTO
Map the DTO properties back onto the appropriate enitity so an UPDATE statement can be generated and executed by NH.
Save the entity
Point 4 is where I have the issue. Currently I can achieve this update using the session.Merge() method, however it must first load the entity from the db (assume no 2LC) before updating. So, both a select and an update statement are fired.
What I would like to do is create a transient instance of the entity, map the new values from the DTO, then have NH generate a SQL statement using only the properties I have changed. The additional select should be unnecessary as I already have the entity ID and the values required for the SET clause. Is this possible in NH?
Currently using session.Update(), all properties will be included in the update statement and an exception is raised due to the uninitialized properties that are not part of the DTO.
Essentially I need a way to specify which entity properties are dirty so only these are included in the update.
== EDIT ==
For example...
public class Person
{
public virtual int PersonId { get; set; }
public virtual string Firstname { get; set; }
public virtual string Nickname { get; set; }
public virtual string Surname { get; set; }
public virtual DateTime BirthDate { get; set; }
}
And the test case.
// Create the transient entity
Person p = new Person()
p.id = 1;
using (ISession session = factory.OpenSession())
{
session.Update(p);
// Update the entity – now attached to session
p.Firstname = “Bob”;
session.Flush();
}
I was hoping to generate a SQL statement similar to ‘UPDATE Persons SET Firstname = ‘Bob’ WHERE PersonID = 1’. Instead I get a DateTime out of range exception due to BirthDate not being initialised. It shouldn’t need BirthDate as it is not required for the SQL statement. Maybe this isn’t possible?
== /EDIT ==
Thanks in advance,
John
Dynamic-update is what you're looking for. In your mapping file (hbm.xml):
<class name="Foo" dynamic-update="true">
<!-- remainder of your class map -->
Be aware of the potential problems that this may cause. Let's say you have some domain logic that says either FirstName or Nickname must not be null. (Completely making this up.) Two people update Jon "Jonboy" Jonson at the same time. One removes his FirstName. Because dynamic-update is true, the update statement just nulls out Jon and the record is now "Jonboy" Jonson. The other simultaneous update removes his Nickname. The intent is Jon Jonboy. But only the null-out of the Nickname gets sent to the database. You now have a record with no FirstName or Nickname. If dynamic-update had been false, the second update would have set it to Jon Jonboy. Maybe this isn't an issue in your situation, but setting dynamic-update="true" has consequences and you should think through the implications.
UPDATE: Thanks for the code. That helped. The basic problem is NHibernate not having enough information. When you say session.Update(p), NHibernate has to associated a disconnected entity with the current session. It has a non-default PK. So NHibernate knows that it's an update and not an insert. When you say session.Update(p), NHibernate sees the whole entity as dirty and sends it to the database. (If you use session.Merge(obj), NHibernate selects the entity from the database and merges obj with it.) This is not what you really mean. You want to associate your object with the current session, but mark it as clean. The API is somewhat non-intuitive. You use session.Lock(obj, LockMode.None) as below.
using(var session = sessionFactory.OpenSession())
using(var tx = session.BeginTransaction()) {
var p = new Person {PersonId = 1};
session.Lock(p, LockMode.None); // <-- This is the secret sauce!
p.Firstname = "Bob";
// No need to call session.Update(p) since p is already associated with the session.
tx.Commit();
}
(N.B. dynamic-update="true" is specified in my mapping.)
This results in the following SQL:
UPDATE Person
SET Firstname = 'Bob' /* #p0_0 */
WHERE PersonId = 1 /* #p1_0 */
I have 3 related objects (Entry, GamePlay, Prize) and I'm trying to find the best way to query them for what I need using NHibernate. When a request comes in, I need to query the Entries table for a matching entry and, if found, get a) the latest game play along with the first game play that has a prize attached. Prize is a child of GamePlay and each Entry object has a GamePlays property (IList).
Currently, I'm working on a method that pulls the matching Entry and eagerly loads all game plays and associated prizes, but it seems wasteful to load all game plays just to find the latest one and any that contain a prize.
Right now, my query looks like this:
var entry = session.CreateCriteria<Entry>()
.Add(Restrictions.Eq("Phone", phone))
.AddOrder(Order.Desc("Created"))
.SetFetchMode("GamePlays", FetchMode.Join)
.SetMaxResults(1).UniqueResult<Entry>();
Two problems with this:
It loads all game plays up front. With 365 days of data, this could easily balloon to 300k of data per query.
It doesn't eagerly load the Prize child property for each game. Therefore, my code that loops through the GamePlays list looking for a non-null Prize must make a call to load each Prize property I check.
I'm not an nhibernate expert, but I know there has to be a better way to do this. Ideally, I'd like to do the following (pseudocode):
entry = findEntry(phoneNumber)
lastPlay = getLatestGamePlay(Entry)
firstWinningPlay = getFirstWinningGamePlay(Entry)
The end result of course is that I have the entry details, the latest game play, and the first winning game play. The catch is that I want to do this in as few database calls as possible, otherwise I'd just execute 3 separate queries.
The object definitions look like:
public class Entry
{
public Guid Id {get;set;}
public string Phone {get;set;}
public IList<GamePlay> GamePlays {get;set;}
// ... other properties
}
public class GamePlay
{
public Guid Id {get;set;}
public Entry Entry {get;set;}
public Prize Prize {get;set;}
// ... other properties
}
public class Prize
{
public Guid Id {get;set;}
// ... other properties
}
The proper NHibernate mappings are in place, so I just need help figuring out how to set up the criteria query (not looking for HQL, don't use it).
since you are doing this in each request maybe it should be better to set up two formula-properties in your entity.
The first one should fetch the latest Gameplay-Id and the other the first Gameplay-Id with a not Null property
this could be as such in the xml mapping file of Entry
<property name="LatestGameplay" formula="select top(1)gp.Id from Gameplay gp where gp.FK_EntryId = PK_EntryId order by gp.InsertDate desc" />
this leaves you with the Gameplay Id's on the Entry entity and after you fetch it it would require another round trip to the DB to GetById-fetch the gameplay's
Alternatively you could work-around using filters.
Set the collection back to "lazy"
and create these nice filters
Gameplay latest = NHibernateSession.CreateFilter(entry.GamePlays , "order by InsertDate desc").SetMaxResults(1).SetFirstResult(1).UniqueResult<Gameplay>();
Gameplay winner = NHibernateSession.CreateFilter(entry.GamePlays , "where FK_PrizeId is not null order by InsertDate asc ").SetMaxResults(1).SetFirstResult(1).UniqueResult<Gameplay>();
And IFilters can be used in a multiquery as so have 2 db hits: one for the original Entry and one for the multiquery.
Last but not least, you could define 2 bags in the Entry entity, one IList<GamePlay> Latest and one IList<Gameplay> Winner which in the Entry mapping file would be filtered with the appropriate query (although i don't remember now if you can define TOP clauses in the filters) and set those as non-lazy. Then with a single round-trip you can have all the data you want with the following (ugly) syntax
Entry entry = findEntry(phoneNumber);
Gameplay winner = entry.Winner[0]; //check this if null first
Gameplay Latest = entry.Latest[0]; //ditto
note that of all the solutions the 3rd is the one that provides a mechanism to generate additional queries, as the bag can be used in a Criteria/HQL query