I was wondering about the justification for a certain limitation/restriction in Hibernate tools. In the hibernate.reveng.xml file, the documentation states that in the <foreign-key> the constraint-name is required. Firstly, why does the dtd not reflect this requirement? Secondly, why is this a required element in the first place? The OverrideBinder.bindForeignKeys method calls Table.createForeignKey, which will generate a unique(?) FK name if the constraintName is not specified. Subsequent to this call, prior to processing the directives embedded in the <foreign-key> element, it checks to see if the constraint-name was specified before processing the embedded directives. So, basically, if you do not specify the constraint-name, it generates the FK anyways and then promptly ignores the embedded directives (e.g. <set exclude="true"/>).

I came across this problem in my reverse engineering when adding foreign keys in the hibernate.reveng.xml file, where I did not realize that the constraint-name was required and was baffled as to why it was ignoring my foreign-key directives. Unfortunately, the DBAs left out a bunch of foreign keys for some unknown reason, and I have 400 tables to map, adding about 1000 foreign keys. Most of these FKs are not mapped bidirectionally, so I wanted them to ignore the <set> side of things in most circumstances. But I really do not want to have to put in 1000 made up, do-nothing foreign key constraint names (FK0001, FK0002, FK0003, etc., ad infinitum, ad nauseum) when Table.createForeignKey will do it for me.

Simple fix/enhancement: Change documentation to state that if the constraint-name is not specified, a generated constraint-name will be supplied. Remove/enhance the StringHelper.isNotEmpty(constraintName) check in the OverrideBinder.bindForeignKeys method.

the constraint name is used for naming stuff if there is duplicate names.

...put any irregularities into our jira.

The only irregularity that exists is that the foreign-key directives are silently ignored when the constraint-name is not specified. Since the JDBCReader already throws a MappingException when duplicate foreign key constraint names are specified, I just did not see the purpose of requiring an essentially arbirtrary constraint-name when Table.createForeignKey will create a functional constraint name for you if you do not specify one.

I can JIRA the request for enhancement if you so desire. At least this forced me to dig down into the JDBCReader and OverrideBinder to see what happens in there. Nice stuff! :D

report in jira otherwise i forget...

one reason for requiring them is to be able to detect wether it is a duplicate or not...anyhow ..into jira