[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3532?page=c...
]
Dan Allen commented on HHH-3532:
--------------------------------
I tested the change and it works. It would be a good idea to have a test for this, but
I'm not familiar enough with the Hibernate test suite to do this in a reasonable
amount of time.
Here's what needs to be done.
Create two tables, one with a foreign key to another. Use a named foreign key which would
guarantee under the old behavior where the UpdateSchema task creates a duplicate foreign
key. Run the UpdateSchema task. Verify that an extra foreign key is not created.
Here is a sample schema (hsqldb)
create table Person (
username varchar(10) not null,
name varchar(100) not null,
primary key (username),
unique (username)
)
create table Vehicle (
state varchar(2) not null,
registration varchar(8) not null,
year integer not null,
make varchar(50) not null,
model varchar(50) not null,
ownerUsername varchar(10),
primary key (state, registration)
)
alter table Vehicle
add constraint owner
foreign key (ownerUsername)
references Person
schema update task should look for foreign key signature
--------------------------------------------------------
Key: HHH-3532
URL:
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3532
Project: Hibernate Core
Issue Type: Improvement
Components: metamodel
Affects Versions: 3.3.1
Reporter: Dan Allen
Priority: Critical
Fix For: 3.2.x, 3.3.x
Attachments: HHH-3532-3.2.x.patch, HHH-3532.patch
Original Estimate: 1 hour
Remaining Estimate: 1 hour
When the schema update task runs, it attempts to locate a foreign key in the database
that matches the name provided in the Hibernate mapping information for that particular
association. (The name of the foreign key is specified either in the foreign-key attribute
in a mapping file or the @ForeignKey annotation). That much works fine. It's when a
foreign key name is not provided, common in the case when standard JPA mappings are used,
that problems arise.
When a foreign key name is not provided, Hibernate generates a key name as follows:
"FK" + hash code of referenced entity name + hash codes of columns
The chances of an existing foreign key in the database using this generated name is slim
to none (every DBA has their own conventions that they follow). Thus, as soon as the
scheme update task hits a database under these conditions, duplicate foreign key
definitions appear all over the place.
If a foreign key name is not provided in the mapping, Hibernate should be looking instead
to see if the signature of the foreign key matches that which is required to fulfill the
mapping. In fact, I would even argue that the name is irrelevant and should only be used
when actually *creating* the foreign key.
Here is the existing logic in the generateSchemaUpdateScript in Configuration:
ForeignKey fk = (ForeignKey) subIter.next();
boolean create = tableInfo == null || tableInfo.getForeignKeyMetadata( fk.getName() ) ==
null;
It should be like this instead:
ForeignKey fk = (ForeignKey) subIter.next();
boolean create = tableInfo == null || tableInfo.getForeignKeyMetadata( fk ) == null;
getForeignKeyMetadata should retrieve based on the fk signature rather than the fk name.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira