Implement "RESTRICT" foreign key constraint in Wt::Dbo module

[ekondayan]

The RESTRICT foreign key constraint is supported by all the major databases.
It is a must have feature for quite a lot of database schemes.
It's implementation in the Wt::Dbo module is very straight forward process, so there is no apparent reason no to do so.

Tested:

1. Wt compiles successfully
2. SQLite3 - SQL generation for the tables works as expected

[emweb]

Thanks for your contribution. Your changes look good, and I know it's fairly trivial, but what would be better is if we had some unit tests, so if you can manage to write some unit tests that would be ideal. If you can check it with Sqlite3, we can make sure it works with all of the other ones.

I suppose this is a case of checking Session::tableCreationSql() and seeing if Session::createTables() works?

Regards,
Roel

[ekondayan]

I've written the test, but I do not know how to include it into the framework.
Should I put into DboTest.C or should I create a new file DboTest8.C?
Could you guide me how to do it?

[emweb]

That's a good start, I can take it from there.

Regards,
Roel

[ekondayan]

There is something that bothers me a bit.
File Wt/Dbo/Session.C Lines: 912
The if statement check the variable haveSupportUpdateCascade_ which is set here

void Session::initSchema() const
{
...
haveSupportUpdateCascade_ = conn->supportUpdateCascade();
...
}

Do we need to add another variable haveSupportUpdateRestrict_ next to haveSupportUpdateCascade_

void Session::initSchema() const
{
...
haveSupportUpdateCascade_ = conn->supportUpdateCascade();
haveSupportUpdateRestrict_ = conn->supportUpdateRestrict();
...
}

and change the if statement to

std::string Session::constraintString(Impl::MappingInfo *mapping, ...
{
...
else if (field.fkConstraints() & Impl::FKOnUpdateRestrict
&& haveSupportUpdateRestrict_)
...
}

Or since all databases support the RESTRICT constraint, checking haveSupportUpdateRestrict_ is useless?
Or to leave it like it is?

[emweb]

Good observation. It appears that that check is there for Oracle. I'm still checking to see what Oracle supports. The implementation seems to suggest that ON UPDATE is just not supported in general, since it is also used to disable ON UPDATE SET NULL, so it's kind of poorly named.

In any case, I don't think it's necessary to introduce a supportUpdateRestrict().

Regards,
Roel