SchemaExport vs HiLo algorithm - nhibernate

I'm using schemaExport to create an in-memory database for my automatic tests. I have several (5) classes mapping their HiLo identity to the same database table, using one column per class.
This gives me a table, hibernate_unique_key, w. 5 columns. When generating the database using scemaexport, however, the table only gets a single column (IPoolActivation), thereby making my querys fail, since the mappings in the model are now invalid. I've tried manually querying the in memory database to create the table, but I would rather know how to make schemaexport do it right.
A snippet from one of my mapping files;
<id name="Id" column="Id" type="Int32">
<generator class="hilo">
<param name="column">IENPool</param>
</generator>
</id>
What is the proper way to do this?

I believe there is a bug (or it is by design) in SchemaExport, and it only looks at the hibernate_unique_key table once when it sees the first entity with HiLo. As a result all HiLo entities must use the same column.
However, I recently needed a custom IdGenerator which I based off the same base class in NHibernate that HiLo inherits from. For this database I needed one column per table (it was for a legacy database that was setup with a HiLo style generator that way).
You can see what I did here if it helps:
Implementing a custom id Generator for nHibernate
Generate custom DDL for a custom id Generator
Entire post on nhibernate.info

Related

Nhibernate list mapping with not continuous index column

I am developping a ASP.NET with VB using NHIBERNATE to map the tables of a pre-existing database (SQL Server 2005). I have a many-to-many relationship between to entities, that I map like this:
<list name="PropName" table="TableHoldingRelation" lazy="false" >
<key column="idEntity1"></key>
<index column ="orderingColumn" ></index>
<many-to-many class="Entity2" column="idEntity2"></many-to-many>
</list>
The mapping works perfectly and the list(of Entity2) its ordered by the selected column.
The problem is that this column is not continuous, as there might be some values missing (ie: 0,1,3,8). NHibernate is leaving those spaces as null/nothing elements. I would want to have the list "compacted", only containing existing elements, ordered by that column.
Can I achieve this without having to update the database? (updating is not a good solution as it probably will happen in the future that some elements get removed)
Thanks in advance for your help.
EDIT: A bit more info in the problem.
The tables/entities in this case refer to Menus and MenuItems. The application that I am working is is a very complex website, with lots of diferent roles. Each role has his unique menus congfiguration, with their unique items. There are even single users with unique settings. My task is to rewrite the .NET clases and mappings, as they are really messy and other things not relevant for this question. So the database design I am mapping is (for this question, obviously there are other tables):
One table holding menus and their attributes(like wich role do they correspond to)
One table holding menuitems and their attributes (like a link they point to)
One table holding de relation menu-menuitem and a "position"/order column inside that menu.
Just in case more insight on the problem was needed.
As you can read here NHibernate Mapping - <list/> - by Ayende, this behaviour is by design. An extract from comments (close to your question):
...Because in general, having NH doing something like that for you can be
bad. There is a meaning to null values.
But broadly, it is because it is not the responsibility of NHibernate
to do so. If you want something like that, you don't need a list, you
need an ordered set...
With this we can try to change your mapping (see the order-by attribute):
<bag name="PropName" table="TableHoldingRelation"
order-by="orderingColumn"
lazy="false" >
<key column="idEntity1"></key>
<many-to-many class="Entity2" column="idEntity2"></many-to-many>
</bag>
But, this mapping won't allow you to insert into that column orderingColumn.
This documentation 24. Best Practices says:
Don't use exotic association mappings.
Good usecases for a real many-to-many associations are rare. Most of
the time you need additional information stored in the "link table".
In this case, it is much better to use two one-to-many associations to
an intermediate link class. In fact, we think that most associations
are one-to-many and many-to-one, you should be careful when using any
other association style and ask yourself if it is really neccessary.
So, maybe introduce the man-in-the middle pairing object, put the management of the OrderBy property there, and use the sorted list..

NHibernate HiLo generation and SQL 2005/8 Schemas

I have an issue on my hands that I've spent several days searching for an answer to no avail...
We're using HiLo Id generation, and everything seems to be working fine, as long as the entity table is in the same schema as the hibernate_unique_key table.
The table structure is pretty simple. I have my hi value table in the db as dbo.hibernate_unique_key. Several entity table are also in the dbo schema, and they work without issue. Then we have tables under the "Contact" schema (such as Contact.Person and Contact.Address).
In the Person Mapping file:
<class name="Person" table="Person" schema="Contact">
<id name="Id" unsaved-value="0">
<generator class="hilo">
<param name="max_lo">100</param>
</generator>
</id>
...
When I try to insert a Person entity, I get an error of "Invalid object name 'Contact.hibernate_unique_key'. That error is certainly clear enough. So I add:
<param name="schema">dbo</param>
to my mapping file/generator element. Now, when the SessionFactory is built, I get a "An item with the same key has already been added." error. So now I'm a bit stuck. I can't leave the HiLo generator without a schema, because it picks up the schema from the Class, and I can't specify the schema because it's already been added (presumably because it's my "default_schema" as identified in my XML cfg file).
Am I completely hosed here? Must I either
A) Keep all my tables in the dbo schema or
B) Create a separate HiLo Key table for each unique schema in the DB?
Neither of those scenarios is particularly palatable for my application, so I'm hoping that I can "fix" my mapping files to address this issue.
Only one such table per database should exist. Such data table should imply the following columns (let's call this table Parameters):
HiLoId
TableName
ParamName
HiLoAssigned
In addition to be used as a HiLo assignment data table, this could be used as a parameter table. As such, the ParamName field is required. This could contain data such as:
HiLoId | TableName | ParamName | HiLoAssigned
---------------------------------------------
1 | Parameters| HiLoId | 3
2 | Customers | CustomerId| 9425
3 | Invoices | InvoiceId | 134978
And when you need some other parameters, such as a parameter for a job that would prune your tables for history, then an age parameter for record could be inserted into it.
Well, I'm a little further in the subject than what you actually asked. Just sharing some additional thoughts in database design/architecture.
Take an eye out this question, and see my answer there. This might answer your question as well, and bring further information to this answer.
Have you tried specifying the schema with the table name on all generators (including the ones already in the dbo schema? I.e.
<param name="table">dbo.hibernate_unique_key</param>
The hilo generator looks for a '.' in the table name, and qualifies it (with schema) only if one isn't there.
I don't think there's anything wrong with solution B. Behavior will be pretty much the same.

NHibernate: inserting additional calculated column

What would be your approach with NHibernate to do the following:
When Comment is inserted into DB, and its property Text is inserted into column Text (due to a mapping), insert into column TextHash a hash of this property value.
It seems trivial if I map TextHash, but how should I do this without mapping it?
I do not need it in the domain model, its a DB-level optimization (I would use computed HashBytes if not for the length limit).
There are some similar questions, such as this one:
Unmapped Columns in NHibernate?
However, IInterceptor seems like an overkill for a change in a single entity insert. And EventListeners are less than perfectly documented and also somewhat too complex for a single column.
So I have decided on the solution that I see both as most reusable and most local:
<property name="Text" type="StringWithMD5HashUserType">
<column name="Text" length="20000" not-null="true" />
<column name="TextHash" length="32" not-null="true" />
</property>
Where StringWithMD5HashUserType is ICompositeUserType that reads Text from first column, but writes both Text and its Hash (I do not add the code of StringWithMD5HashUserType because it is way too long, but essentially very simple).
Well, if you don't want to use the property, declare it private and calculate it when you set the text property. It's not exactly "without mapping it", but you won't see it when you're using the class (disclaimer: advice based on Hibernate, as opposed to NHibernate).
Other than that, you could always write a trigger, although I'd personally much rather add a private field and be done with it.
Here's a suggestion:
As far as I know you can't access a column in your DB using NHibernate if it there no mapping defined for it.
To prevent other parts of your application to see this field of your class you can define its access as field in your mapping so that it can be private an no one knows that it exists but NHibernate:
In your class:
private string _textHash
In your mapping:
<property name='_textHash' column='TextHash' access='field' />

NHibernate 2.1.0.4000 doesn't seem to like batch insert

I am using NHibernate 2.1.0.4000 in one of the projects. I have set adonet.batch_size to 100 in the cfg file however I still see that insert statement is treated as single statement. Update seems to work fine. What's going on?
Updated:
Is it because I've chosen identity as the primary key generator?
<id name="Id" column="Id" unsaved-value="0" type="Int32">
<generator class ="identity"></generator>
</id>
I don't know of any issues with that particular NHibernate version.
Are your using native as the ID generator for your entities? Because this will force every insert to happen alone, selecting back the generated ID. This is because the database needs to generate every ID. This would also explain why batching seems to work on updates.
If possible, you should switch to e.g. the hilo strategy, or even guid if you don't care about (easily) readable IDs.
Fabio has han interesting post here regarding this topic.

Should NHibernate assign id to entities or should it be handled by application?

I'm writing an application and started to test my domain model entities. If I Create an instance of entity Company like this var company = new Company("my company"); I should get a valid entity, meaning the company should at this moment have an Id correct?
So the problem is that at the moment I have the Id generation made in the DB defined in an hbm file like this:
<id name="ObjectIdentity" column="CompanyId" type="System.Guid" unsaved-value="00000000-0000-0000-0000-000000000000">
<generator class="guid.comb"/>
</id>
This causes a problem when writing unittests as I do not have en entity with an Id as it dosen't touch the db in the test, aka I have an invalid entity.
Now should I assign Id in the application and not let nhibernate be in charge of this or is this the wrong way to do it?
In most cases you should be letting NHibernate do its job which is to handle persistence. This is important since it allows you to easily change things (we went from identity to hilo mid-project).
I would question why you care that a newly created object has an id or not? From a business point of view a persistence ID is not relevant and shouldn't be checked with via unit tests. This as mentioned is the domain of integration tests. You should be careful in how you are using an objects persistence Id throughout the rest of your application. Remember this should NOT be treated as the objects business id/key.
You need to call Session.Save on the entity before a guid is generated. You can call Session.Save to generate the entity without actually saving it to the database. This article does a pretty decent job of explaining it
Neither NHibernate nor the Application should be handling identifiers. Leave it to the database since this is the only concrete store of your data, and it is the only part of your application that knows what IDs have already been assigned and what ones are available.
Make an identity primary key column on your database table:
CREATE TABLE dbo.sample (
id int primary key identity(1,1),
...
...
...)
map your entities like this:
<id name="ID" column="id">
<generator class="identity" />
</id>
The Primary key will be generated automatically by the database when you save a new entity for the first time. IDENTITY(1,1) means "give new rows an Identity start at '1' then each subsequent row gets incremented by 1": so 1,2,3,4,5,6,7
You need to save the entity to get an id if you're using db generated identity,
if you wish to create your own identities - it is fine.
do what suits you, just remember what you have chosen when you test.
I usually decide on each class which is more suitable for me - either DB generated or my own.
Nhibernate should generate Id's for you. Id property of the entity must be protected. The best way is to use hilo