I have a database at work that has tables without primary keys (Sql Server). It uses a different schema. It has a unique (sometimes compound) key for each table. I cannot modify database in any way. It is a vendor's database.
Is it ok to trick EF into thinking the unique key is the primary key? Will there be any issues with inserting or updating records?
Thanks,
Rick
Does the EntityType.KeyMembers property contain just the primary key, or also unique keys and foreign keys?
I received an answer from the entity framework team on twitter. I am told that it contains unique keys and primary keys.
I need to map some external database, I can't modify the schema. But the tables don't have primary key but rather columns like Client_ID, Calendar_ID but they are not null and entity can map them, much worse is if these columns in few tables can be null, then Entity Framework throws an error that it can't be mapped.
My question is: can I somehow disable entity tracking and map these tables without primary key and with columns as null?
Or can I use code-first approach, does it let me to create and map class with no primary key and all columns as is null?
Entity Framework must have a key field in order to work. It doesn't have to be a true primary key in the database but it needs to be unique. If you have tables that have a nullable field and no true primary key and you can't modify the database then maybe Entity Framework isn't the best fit for this project. It might work if you never try and load the entities with the null values, but it will throw errors when you do (as you have noticed).
This question already has answers here:
Why doesn't EF 4 generate association for FK relation to column with unique index?
(2 answers)
Closed 1 year ago.
I have a table tblEvent and another table tblContact. There is a field ContactID in tblEvent that is linked by a foreign key to ContactID in tblContact. I've triple checked; the foreign key is in the database.
The Entity Framework model DOES NOT generate a navigation property and foreign key association for it as it does for the other foreign keys.
What could be different about this foreign key that EF does not generate a navigation property for it?
Thanks!
I found the solution here:
Why doesn't EF 4 generate association for FK relation to column with unique index?
The reason it was not recognizing it is was because the table had a unique non-clustered index on the foreign key field.
Apparently, a unique, non-clustered index allows for a null value which can't be mapped.
I know it is an old post, but I was faced with this same problem, I got to this post and rsteckly's comment made me check and indeed I have an index on my main table as non-clutered for my key to improve performance.
A solution I found is to add a filter to the index (WHERE tableId IS NOT NULL), in this way all the references were generated correctly.
Hope this solution will help someone
if you are are using 3.5sp1 that could be the issue.
This article touches on some of the changes and if nothing else you can always manually add the fk association in the entity designer as described in the article.
Link
I am looking at multi-tenancy database schema design for an SaaS concept. It will be ASP.NET MVC -> EF, but that isn't so important.
Below you can see an example database schema (the Tenant being the Company). The CompanyId is replicated throughout the schema and the primary key has been placed on both the natural key, plus the tenant Id.
Plugging this schema into the Entity Framework gives the following errors when I add the tables into the Entity Model file (Model1.edmx):
The relationship 'FK_Order_Customer' uses the set of foreign keys '{CustomerId, CompanyId}' that are partially contained in the set of primary keys '{OrderId, CompanyId}' of the table 'Order'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_OrderLine_Customer' uses the set of foreign keys '{CustomerId, CompanyId}' that are partially contained in the set of primary keys '{OrderLineId, CompanyId}' of the table 'OrderLine'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_OrderLine_Order' uses the set of foreign keys '{OrderId, CompanyId}' that are partially contained in the set of primary keys '{OrderLineId, CompanyId}' of the table 'OrderLine'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_Order_Customer' uses the set of foreign keys '{CustomerId, CompanyId}' that are partially contained in the set of primary keys '{OrderId, CompanyId}' of the table 'Order'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_OrderLine_Customer' uses the set of foreign keys '{CustomerId, CompanyId}' that are partially contained in the set of primary keys '{OrderLineId, CompanyId}' of the table 'OrderLine'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_OrderLine_Order' uses the set of foreign keys '{OrderId, CompanyId}' that are partially contained in the set of primary keys '{OrderLineId, CompanyId}' of the table 'OrderLine'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The relationship 'FK_OrderLine_Product' uses the set of foreign keys '{ProductId, CompanyId}' that are partially contained in the set of primary keys '{OrderLineId, CompanyId}' of the table 'OrderLine'. The set of foreign keys must be fully contained in the set of primary keys, or fully not contained in the set of primary keys to be mapped to a model.
The question is in two parts:
Is my database design incorrect? Should I refrain from these compound primary keys? I'm questioning my sanity regarding the fundamental schema design (frazzled brain syndrome). Please feel free to suggest the 'idealized' schema.
Alternatively, if the database design is correct, then is EF unable to match the keys because it perceives these foreign keys as a potential mis-configured 1:1 relationships (incorrectly)? In which case, is this an EF bug and how can I work around it?
On a quick scan of EF's error messages, it clearly doesn't like the way you're setting up compound keys, and I think it's probably nudging you in the right direction. Give some thought again to what makes your primary keys unique. Is the OrderID alone not unique, without a CompanyID? Is a ProductID not unique, without a CompanyID? An OrderLine certainly should be unique without a CompanyID, since an OrderLine should be associated only with a single Order.
If you truly need the CompanyID for all of these, which probably means that the company in question is supplying you with ProductID and OrderID, then you might want to go a different direction, and generate your own primary keys that are not intrinsic to the data. Simply set up an auto-increment column for your primary key, and let these be the internal OrderID, OrderLineID, ProductID, CompanyID, etc. At that point, the OrderLine won't need the customer's OrderID or CompanyID; the foreign key reference to the Order would be its starting point. (And the CustomerID should never be an attribute of an order line; it's an attribute of the order, not the order line.)
Compound keys are just messy. Try designing the model without them, and see if it simplifies things.
I think that the error is not in the design.
Is not in the EF.
Is in Sql Server relations.
Read the EF message:
The relationship 'FK_Order_Customer'
uses the set of foreign keys
'{CustomerId, CompanyId}' that are
partially contained in the set of
primary keys '{OrderId, CompanyId}' of
the table 'Order'. The set of foreign
keys must be fully contained in the
set of primary keys, or fully not
contained in the set of primary keys
to be mapped to a model.
ERROR
Actualy the relation betwen Order and Customer use only one field (probably you dragged with the mouse the field "CustomerId" from teh Order table to the "Id" of the Customer table)
SOLUTION
Right click on the wire that connect Order and Customer and in the relation add also the CompanyId
PS: The design is correct.
Putting the CompanyId in each table is rith solution in multi-tenant architecture because help to scale (usualy always want to select only records from the loggedIn company).
I think storing the company number in each of the tables is hurting you more than helping. I can understand why you want to do this (as the programmer/dba you can just go into any table and 'see' what data belongs to who which is comforting), but it is getting in the way of you setting up the database the way it should be.
Avoid the compound keys and your design gets a whole lot simpler.
If you have to absolutely add CompanyID column to each table, add it as a regular column and not a composite Key. Composite key is mostly used when you have to implement many to many relationship.
As someone mentioned also create a Non-clustered Index on CompanyID so joins to the Company table are benefitted.
Thanks!
First: like others said, when referencing a foreign key, use the whole primary key in the other table (ie. both fields).
Second, I cannot imagine not using a CompanyID column in most tables in a serious application. Orderdetail might perhaps be an exception in this case (also global lookup tables perhaps, unless they are tenant dependant). Thing is, you cannot do any safe sort of free form search on a table without either adding the CompanyID, or doing JOINs up until the point you reach a table which has that column. The latter one obviously costs performance. Perhaps in this case you could make an exception for orderdetail and only search in the joined version (only two tables). Then again, its not really consistent.
Also regarding making a it a compound key or not: its possible, but opens up the possibility that a bug writes information incorrectly (into non existent, or other people's administrations) for the duration of the bug. Try to fix that in production, not to mention explain it to customers why there are seeing their competitors orders in their system.