(For background I'm using Postgres 12.4)
I'm unclear why deletes work when there are circular FKs between two tables and both FKs are set to ON DELETE CASCADE.
CREATE TABLE a (id bigint PRIMARY KEY);
CREATE TABLE b (id bigint PRIMARY KEY, aid bigint references a(id) on delete cascade);
ALTER TABLE a ADD COLUMN bid int REFERENCES b(id) ON DELETE CASCADE ;
insert into a(id) values (5);
insert into b(id, aid) values (10,5);
update a set bid = 10 where id=5;
DELETE from a where id=5;
The way that I am thinking about this, when you delete the row in table 'a' with PK id = 5, postgres looks at tables that have a referential constraint referencing a(id), it finds b, it tries to delete the row in table b with id = 10, but then it has to look at tables referencing b(id), so it goes back to a, and then it should just end up in an infinite loop.
But this does not seem to be the case. The delete completes without error.
It's also not the case, as some sources say online, that you cannot create the circular constraint. The constraints are created successfully, and neither of them is deferrable.
So my question is - why does postgres complete this circular cascade even when neither constraint is set to deferrable, and if it's able to do so, then what is the point of even having a DEFERRABLE option?
Foreign key constraints are implemented as system triggers.
For ON DELETE CASCADE, this trigger will run a query like:
/* ----------
* The query string built is
* DELETE FROM [ONLY] <fktable> WHERE $1 = fkatt1 [AND ...]
* The type id's for the $ parameters are those of the
* corresponding PK attributes.
* ----------
*/
The query runs is a new database snapshot, so it cannot see rows deleted by previous RI triggers:
/*
* In READ COMMITTED mode, we just need to use an up-to-date regular
* snapshot, and we will see all rows that could be interesting. But in
* transaction-snapshot mode, we can't change the transaction snapshot. If
* the caller passes detectNewRows == false then it's okay to do the query
* with the transaction snapshot; otherwise we use a current snapshot, and
* tell the executor to error out if it finds any rows under the current
* snapshot that wouldn't be visible per the transaction snapshot. Note
* that SPI_execute_snapshot will register the snapshots, so we don't need
* to bother here.
*/
This makes sure that no RI trigger will try to delete the same row a second time, and thus circularity is broken.
(All quotations taken from src/backend/utils/adt/ri_triggers.c.)
Related
I'm adding a new job category to a database. There are something like 20 tables that use jobCategoryID as a foreign key. Is there a way to create a function that would go through those tables and set the jobCategoryID to NULL if the category is ever deleted in the parent table? Inserting the line isn't the issue. It's just for a backout script if the product owners decide at a later date that they don't want to keep the new job category on.
You need some action. First of all update the dirty records to NULL. For each table use:
Update parent_table
Set jobCategoryID = NULL
WHERE jobCategoryID NOT IN (select jobCategoryID FROM Reerenced_tabble)
Then set delete rule of foreign keys to SET NULL.
If you care about performance issue, follow the below instruction too.
When you have foreign key but dirty records it means, that these constraints are not trusted. It means that SQL Optimizer can not use them for creating best plans. So run these code to see which of them are untrusted to optimizer:
Select * from sys.foreign_keys Where is_not_trusted = 1
For each constraint that become in result of above code edit below code to solve this issue:
ALTER TABLE Table_Name WITH CHECK CHECK CONSTRAINT FK_Name
I have a problem when I try to add constraints to my tables. I get the error:
Introducing FOREIGN KEY constraint 'FK74988DB24B3C886' on table 'Employee' may cause cycles or multiple cascade paths. Specify ON DELETE NO ACTION or ON UPDATE NO ACTION, or modify other FOREIGN KEY constraints.
My constraint is between a Code table and an employee table. The Code table contains Id, Name, FriendlyName, Type and a Value. The employee has a number of fields that reference codes, so that there can be a reference for each type of code.
I need for the fields to be set to null if the code that is referenced is deleted.
Any ideas how I can do this?
SQL Server does simple counting of cascade paths and, rather than trying to work out whether any cycles actually exist, it assumes the worst and refuses to create the referential actions (CASCADE): you can and should still create the constraints without the referential actions. If you can't alter your design (or doing so would compromise things) then you should consider using triggers as a last resort.
FWIW resolving cascade paths is a complex problem. Other SQL products will simply ignore the problem and allow you to create cycles, in which case it will be a race to see which will overwrite the value last, probably to the ignorance of the designer (e.g. ACE/Jet does this). I understand some SQL products will attempt to resolve simple cases. Fact remains, SQL Server doesn't even try, plays it ultra safe by disallowing more than one path and at least it tells you so.
Microsoft themselves advises the use of triggers instead of FK constraints.
A typical situation with multiple cascasing paths will be this:
A master table with two details, let's say "Master" and "Detail1" and "Detail2". Both details are cascade delete. So far no problems. But what if both details have a one-to-many-relation with some other table (say "SomeOtherTable"). SomeOtherTable has a Detail1ID-column AND a Detail2ID-column.
Master { ID, masterfields }
Detail1 { ID, MasterID, detail1fields }
Detail2 { ID, MasterID, detail2fields }
SomeOtherTable {ID, Detail1ID, Detail2ID, someothertablefields }
In other words: some of the records in SomeOtherTable are linked with Detail1-records and some of the records in SomeOtherTable are linked with Detail2 records. Even if it is guaranteed that SomeOtherTable-records never belong to both Details, it is now impossible to make SomeOhterTable's records cascade delete for both details, because there are multiple cascading paths from Master to SomeOtherTable (one via Detail1 and one via Detail2).
Now you may already have understood this. Here is a possible solution:
Master { ID, masterfields }
DetailMain { ID, MasterID }
Detail1 { DetailMainID, detail1fields }
Detail2 { DetailMainID, detail2fields }
SomeOtherTable {ID, DetailMainID, someothertablefields }
All ID fields are key-fields and auto-increment. The crux lies in the DetailMainId fields of the Detail tables. These fields are both key and referential contraint. It is now possible to cascade delete everything by only deleting master-records. The downside is that for each detail1-record AND for each detail2 record, there must also be a DetailMain-record (which is actually created first to get the correct and unique id).
I would point out that (functionally) there's a BIG difference between cycles and/or multiple paths in the SCHEMA and the DATA. While cycles and perhaps multipaths in the DATA could certainly complicated processing and cause performance problems (cost of "properly" handling), the cost of these characteristics in the schema should be close to zero.
Since most apparent cycles in RDBs occur in hierarchical structures (org chart, part, subpart, etc.) it is unfortunate that SQL Server assumes the worst; i.e., schema cycle == data cycle. In fact, if you're using RI constraints you can't actually build a cycle in the data!
I suspect the multipath problem is similar; i.e., multiple paths in the schema don't necessarily imply multiple paths in the data, but I have less experience with the multipath problem.
Of course if SQL Server did allow cycles it'd still be subject to a depth of 32, but that's probably adequate for most cases. (Too bad that's not a database setting however!)
"Instead of Delete" triggers don't work either. The second time a table is visited, the trigger is ignored. So, if you really want to simulate a cascade you'll have to use stored procedures in the presence of cycles. The Instead-of-Delete-Trigger would work for multipath cases however.
Celko suggests a "better" way to represent hierarchies that doesn't introduce cycles, but there are tradeoffs.
There is an article available in which explains how to perform multiple deletion paths using triggers. Maybe this is useful for complex scenarios.
http://www.mssqltips.com/sqlservertip/2733/solving-the-sql-server-multiple-cascade-path-issue-with-a-trigger/
By the sounds of it you have an OnDelete/OnUpdate action on one of your existing Foreign Keys, that will modify your codes table.
So by creating this Foreign Key, you'd be creating a cyclic problem,
E.g. Updating Employees, causes Codes to changed by an On Update Action, causes Employees to be changed by an On Update Action... etc...
If you post your Table Definitions for both tables, & your Foreign Key/constraint definitions we should be able to tell you where the problem is...
This is because Emplyee might have Collection of other entity say Qualifications and Qualification might have some other collection Universities
e.g.
public class Employee{
public virtual ICollection<Qualification> Qualifications {get;set;}
}
public class Qualification{
public Employee Employee {get;set;}
public virtual ICollection<University> Universities {get;set;}
}
public class University{
public Qualification Qualification {get;set;}
}
On DataContext it could be like below
protected override void OnModelCreating(DbModelBuilder modelBuilder){
modelBuilder.Entity<Qualification>().HasRequired(x=> x.Employee).WithMany(e => e.Qualifications);
modelBuilder.Entity<University>.HasRequired(x => x.Qualification).WithMany(e => e.Universities);
}
in this case there is chain from Employee to Qualification and From Qualification to Universities. So it was throwing same exception to me.
It worked for me when I changed
modelBuilder.Entity<Qualification>().**HasRequired**(x=> x.Employee).WithMany(e => e.Qualifications);
To
modelBuilder.Entity<Qualification>().**HasOptional**(x=> x.Employee).WithMany(e => e.Qualifications);
Trigger is solution for this problem:
IF OBJECT_ID('dbo.fktest2', 'U') IS NOT NULL
drop table fktest2
IF OBJECT_ID('dbo.fktest1', 'U') IS NOT NULL
drop table fktest1
IF EXISTS (SELECT name FROM sysobjects WHERE name = 'fkTest1Trigger' AND type = 'TR')
DROP TRIGGER dbo.fkTest1Trigger
go
create table fktest1 (id int primary key, anQId int identity)
go
create table fktest2 (id1 int, id2 int, anQId int identity,
FOREIGN KEY (id1) REFERENCES fktest1 (id)
ON DELETE CASCADE
ON UPDATE CASCADE/*,
FOREIGN KEY (id2) REFERENCES fktest1 (id) this causes compile error so we have to use triggers
ON DELETE CASCADE
ON UPDATE CASCADE*/
)
go
CREATE TRIGGER fkTest1Trigger
ON fkTest1
AFTER INSERT, UPDATE, DELETE
AS
if ##ROWCOUNT = 0
return
set nocount on
-- This code is replacement for foreign key cascade (auto update of field in destination table when its referenced primary key in source table changes.
-- Compiler complains only when you use multiple cascased. It throws this compile error:
-- Rrigger Introducing FOREIGN KEY constraint on table may cause cycles or multiple cascade paths. Specify ON DELETE NO ACTION or ON UPDATE NO ACTION,
-- or modify other FOREIGN KEY constraints.
IF ((UPDATE (id) and exists(select 1 from fktest1 A join deleted B on B.anqid = A.anqid where B.id <> A.id)))
begin
update fktest2 set id2 = i.id
from deleted d
join fktest2 on d.id = fktest2.id2
join inserted i on i.anqid = d.anqid
end
if exists (select 1 from deleted)
DELETE one FROM fktest2 one LEFT JOIN fktest1 two ON two.id = one.id2 where two.id is null -- drop all from dest table which are not in source table
GO
insert into fktest1 (id) values (1)
insert into fktest1 (id) values (2)
insert into fktest1 (id) values (3)
insert into fktest2 (id1, id2) values (1,1)
insert into fktest2 (id1, id2) values (2,2)
insert into fktest2 (id1, id2) values (1,3)
select * from fktest1
select * from fktest2
update fktest1 set id=11 where id=1
update fktest1 set id=22 where id=2
update fktest1 set id=33 where id=3
delete from fktest1 where id > 22
select * from fktest1
select * from fktest2
This is an error of type database trigger policies. A trigger is code and can add some intelligences or conditions to a Cascade relation like Cascade Deletion. You may need to specialize the related tables options around this like Turning off CascadeOnDelete:
protected override void OnModelCreating( DbModelBuilder modelBuilder )
{
modelBuilder.Entity<TableName>().HasMany(i => i.Member).WithRequired().WillCascadeOnDelete(false);
}
Or Turn off this feature completely:
modelBuilder.Conventions.Remove<OneToManyCascadeDeleteConvention>();
Some databases, most notably SQL Server, have limitations on the cascade behaviors that form cycles.
There are two ways to handle this situation:
1.Change one or more of the relationships to not cascade delete.
2.Configure the database without one or more of these cascade deletes, then ensure all dependent entities are loaded so that EF Core can perform the cascading behavior.
please refer to this link:
Database cascade limitations
Mass database update to offset PKs: make a copy of the database instead.
Special use case: company A uses a database with the same schema as company B. Because they have merged, they want to use a single database. Hence, many tables from company B's database must have their primary keys offset to avoid collision with company A's records.
One solution could have been to define foreign keys as ON UPDATE CASCADE, and offset the primary keys having the foreign keys follow. But there are many hurdles if you do that (Msg 1785, Msg 8102, ...).
So a better idea that occurs to me is simply to make a copy of the database, DROP and re CREATE the tables that must have their PKs|FKs offset, and copy the data (and while doing so, offset the primary keys and the foreign keys).
Avoiding all the hassle.
My solution to this problem encountered using ASP.NET Core 2.0 and EF Core 2.0 was to perform the following in order:
Run update-database command in Package Management Console (PMC) to create the database (this results in the "Introducing FOREIGN KEY constraint ... may cause cycles or multiple cascade paths." error)
Run script-migration -Idempotent command in PMC to create a script that can be run regardless of the existing tables/constraints
Take the resulting script and find ON DELETE CASCADE and replace with ON DELETE NO ACTION
Execute the modified SQL against the database
Now, your migrations should be up-to-date and the cascading deletes should not occur.
Too bad I was not able to find any way to do this in Entity Framework Core 2.0.
Good luck!
I'm confident that this is possible but for the life of me I can't figure it out.
What I have created is a user history MSSQL table to hold the changes made to a user and by whom. This table contains two foreign keys which reference my other table (User) - one fkey for the affected user and the other fkey for the user making the changes.
What I need is for any changes to the (User) table to cascade and update the corresponding entries in this new table.
The fields in the new table (User_History) are as follows (Each user is identified by two fields):
Affected_User_House_Id - int
Affected_User_Id - int
Modified_By_User_House_Id - int
Modified_By_User_Id – int
Modification_Date - datetime
ModificationMade - ntext
Each field is a primary key except for ‘ModificationMade’. The field ‘Modification_Date’ is accurate down to 1 second.
The problem I am having is creating said cascade.
I have tried running the following T-SQL code:
ALTER TABLE [User_History] WITH CHECK
ADD CONSTRAINT [FK_User_History_User] FOREIGN KEY([Affected_User_House_Id], [Affected_User_Id])
REFERENCES [User] ([User_House_Id], [User_ID])
ON UPDATE CASCADE
GO
ALTER TABLE [User_History] CHECK CONSTRAINT [FK_User_History_User]
GO
ALTER TABLE [User_History] WITH CHECK
ADD CONSTRAINT [FK_User_History_User_ModifiedBy] FOREIGN KEY([Modified_By_User_House_Id], [Modified_By_User_Id])
REFERENCES [User] ([User_House_Id], [User_ID])
ON UPDATE CASCADE
GO
ALTER TABLE [User_History] CHECK CONSTRAINT [FK_User_History_User_ModifiedBy]
GO
This T-SQL gave me the following error:
*'User' table saved successfully
'User_History' table
- Unable to create relationship 'FK_User_History_User_ModifiedBy'.
Introducing FOREIGN KEY constraint 'FK_User_History_User_ModifiedBy' on table 'User_History' may cause cycles or multiple cascade paths. Specify ON DELETE NO ACTION or ON UPDATE NO ACTION, or modify other FOREIGN KEY constraints.
Could not create constraint. See previous errors.*
The code works if I remove the second “ON UPDATE CASCADE” the however that will mean the values in the fields “Modified_By_User_House_Id” and “Modified_By_User_Id” will not be updated to match their referenced values in the User table.
I am at a lost as to how to acomplish this goal.
You can only specify a single cascade. Here's an attempt to simulate multiple cascades with two triggers:
create table TabA (
ID1 int not null,
ID2 int not null,
_RowID int IDENTITY(1,1) not null,
constraint PK_TabA PRIMARY KEY (ID1,ID2),
constraint UQ_TabA__RowID UNIQUE (_RowID)
)
go
create table TabB (
ID1a int not null,
ID2a int not null,
ID1b int not null,
ID2b int not null,
constraint PK_TabB PRIMARY KEY (ID1a,ID2a,ID1b,ID2b)
)
They're simpler than your tables, but hopefully close enough. We need an immutable identifier in TabA, and obviously the IDs aren't it, since the whole point is to cascade changes to them. So I've added _RowID.
It would be nice to implement at least a real foreign key and just simulate the cascade behaviour on top of that, but some simple reflection will demonstrate that there's always a point where the FK would be broken. So we simulate it:
create trigger FK_TabB_TabA on TabB
after insert,update
as
set nocount on
if exists (
select
*
from
inserted i
left join
TabA a
on
i.ID1a = a.ID1 and
i.ID2a = a.ID2
left join
TabA b
on
i.ID1b = b.ID1 and
i.ID2b = b.ID2
where
a._RowID is null or
b._RowID is null)
begin
declare #Error varchar(max)
set #Error = 'The INSERT statement conflicted with the Foreign Key constraint "FK_TabB_TabA". The conflict occurred in database "'+DB_NAME()+'", table "dbo.TabB".'
RAISERROR(#Error,16,0)
rollback
end
And then the cascading update:
create trigger FK_TabB_TabA_Cascade on TabA
after update
as
set nocount on
;with Updates as (
select
d.ID1 as OldID1,
d.ID2 as OldID2,
i.ID1 as NewID1,
i.ID2 as NewID2
from
inserted i
inner join
deleted d
on
i._RowID = d._RowID
)
update b
set
ID1a = COALESCE(u1.NewID1,ID1a),
ID2a = COALESCE(u1.NewID2,ID2a),
ID1b = COALESCE(u2.NewID1,ID1b),
ID2b = COALESCE(u2.NewID2,ID2b)
from
TabB b
left join
Updates u1
on
b.ID1a = u1.OldID1 and
b.ID2a = u1.OldID2
left join
Updates u2
on
b.ID1b = u2.OldID1 and
b.ID2b = u2.OldID2
where
u1.OldID1 is not null or
u2.OldID1 is not null
go
Some simple inserts:
insert into TabA (ID1,ID2)
values (1,1),(1,2),(2,1),(2,2)
go
insert into TabB (ID1a,ID2a,ID1b,ID2b)
values (1,1,2,2)
Then the following gets an error. Not quite like a built in FK violation, but close enough:
insert into TabB (ID1a,ID2a,ID1b,ID2b)
values (1,1,2,3)
--Msg 50000, Level 16, State 0, Procedure FK_TabB_TabA, Line 28
--The INSERT statement conflicted with the Foreign Key constraint "FK_TabB_TabA". The conflict occurred in database "Flange", table "dbo.TabB".
--Msg 3609, Level 16, State 1, Line 1
--The transaction ended in the trigger. The batch has been aborted.
This is the update that we wanted to be able to perform:
update TabA set ID2 = ID2 + 1
And we query the FK table:
select * from TabB
Result:
ID1a ID2a ID1b ID2b
----------- ----------- ----------- -----------
1 2 2 3
So the update cascaded.
Why you can't use real FKs:
You want to have cascading updates. That means that the ID values in TabA are going to change to a new value that doesn't currently exist (caveat - we're excluding situations where 2n rows swap their identity values) - otherwise, the primary key constraint will be broken by this update.
As such, we know that the new key value will not yet exist. If we were to attempt cascading updates using an INSTEAD OF trigger (to update the child table before the parent) then the new values we attempt to update to in TabB do not yet exist. Alternately, if we attempt to do cascading updates using an AFTER trigger - well, we're too late. The FK constraint has already prevented the update.
I suppose you could implement an INSTEAD OF trigger that inserts the new rows as "duplicates", updates the children, then deletes the old rows. In such a circumstance, I think you could have real FKs. But I don't want to try writing that trigger to be right in all circumstances (e.g where you have three rows being updated. Two swap their ID values and the other creates a new ID)
According to this knowledge base article, this error message occurs when "a table cannot appear more than one time in a list of all the cascading referential actions that are started by either a DELETE or an UPDATE statement."
Since you have two paths coming from the same table, a possible workaround may involve creating a new key on the parent table and creating a single foreign key on the child ([Affected_User_House_Id], [Affected_User_Id], [Modified_By_User_House_Id], [Modified_By_User_Id]). This however, will likely create a lot of overhead. As a last resort, you can use triggers to enforce the relational integrity.
I have a huge pl/sql stored procedure, where we make some deletions as long as insertions.
Procedure starts with the statement
EXECUTE IMMEDIATE 'SET CONSTRAINTS ALL DEFERRED'
And at the last commit statement I receive ORA-02292: integrity constraint AAA violated.
The questions is that I don't know which statement exactly causes it, because I have both deletion from parent table (before child one) and insertions into child table before parent.
I tried to google it, but everywhere it's said that 02292 happens when I try to delete only.
Could this error happen when I try to insert value in the child table but there is no this entry in the parent?
Also, what is the difference between 02292 and 02291?
ORA-02292 indicates that the error occurred because A) the constraint has no ON DELETE clause specified, and B) you deleted a row from the master table which had matching references in the child table. Your choices are to modify the constraint so have an ON DELETE CASCADE or to ensure that all child records are deleted before deleting from the master. My preference would be to add ON DELETE CASCADE but I suppose there could be reasons not to do so. See ORA-02292.
ORA-02291 is sort of the opposite of this. ORA-02291 will be raised if you attempt to insert a row into a child table, but the key field values on your new child row as specified in the constraint do not exist in the master table. See ORA-02291.
If you want to disable the constraint from the name to solve ORA-02292.
Look for the table name bounded to that constraint
SELECT owner, table_name FROM dba_constraints WHERE constraint_name = '{CONSTRAINT_NAME}';
Disable constraint (this command should be executed by an admin user)
ALTER TABLE {TABLE_NAME} DISABLE constraint {CONSTRAINT_NAME} cascade;
Do Foreign Key constraints get checked on an SQL update statement that doesn't update the columns with the Constraint? (In MS SQL Server)
Say I have a couple of tables with the following columns:
OrderItems
- OrderItemID
- OrderItemTypeID (FK to a OrderItemTypeID column on another table called OrderItemTypes)
- ItemName
If I just update
update [dbo].[OrderItems]
set [ItemName] = 'Product 3'
where [OrderItemID] = 2508
Will the FK constraint do it's lookup/check with the update statement above? (even thought the update is not change the value of that column?)
No, the foreign key is not checked. This is pretty easy to see by examining the execution plans of two different updates.
create table a (
id int primary key
)
create table b (
id int,
fkid int
)
alter table b add foreign key (fkid) references a(id)
insert into a values (1)
insert into a values (2)
insert into b values (5,1) -- Seek on table a's PK
update b set id = 6 where id = 5 -- No seek on table a's PK
update b set fkid = 2 where id = 6 -- Seek on table a's PK
drop table b
drop table a
No. Since the SQL update isn't updating a column containing a constraint, what exactly would SQL Server be checking in this case? This is similar to asking, "does an insert trigger get fired if I only do an update?" Answer is no.
There is a case when the FK not existing will prevent updates to other columns even though the FK is not changed and that is when the FK is created WITH NOCHECK and thus not checked at the time of creation. Per Books Online:
If you do not want to verify new CHECK or FOREIGN KEY constraints
against existing data, use WITH NOCHECK. We do not recommend doing
this, except in rare cases. The new constraint will be evaluated in
all later data updates. Any constraint violations that are suppressed
by WITH NOCHECK when the constraint is added may cause future updates
to fail if they update rows with data that does not comply with the
constraint.