Re: Schema version control

Bill Moran wrote on 11.02.2011 00:37:
> Anyway ... based on nothing more than a quick scan of their quickstart
> page, here are the differences I see:
> * Liquibase is dependent on you creating "changesets".  I'm sure this
>    works, but we took a different approach with dbsteward.  dbsteward
>    expects you to maintain XML files that represent the entire database,
>    then dbsteward does the work of figuring out what changed.  Our
>    opinion was that svn already does the work of tracking changes, why
>    reinvent the wheel.

That sounds like a very nice feature.

> * Looks like liquibase requires you to talk to the database to push
>    the changes?  dbsteward outputs a DDL/DML file that you can push
>    in whatever way is best.  This is important to us because we use
>    Slony, and DDL changes have to be submitted through EXECUTE SCRIPT()

No, Liquibase can also emit the SQL that it would execute.

> * dbsteward has built-in Slony support (i.e. it will make slony configs
>    as well as slony upgrade scripts in addition to DDL/DML)


> * liquibase has a lot more supported platforms at this time.  dbsteward
>    only supports PostgreSQL and MSSQL (because that's all that we needed)
>    but I expect that other support will come quickly once we release it.


> * Does liquibase support things like multi-column indexes and multi-
>    column primary keys?  dbsteward does.

Yes without problems (including of course the necessary foreing keys)


> Anyway ... sorry for the teaser on this, but we're trying to get through
> all the hoops the company is requiring us to do to release it, and we
> think we're on track to be ready by PGCon, so there'll be a website up
> as soon as we can get it.

Thanks for the feedback, I would really like to see it.

The approach that you do not record the changes but simply let the software find them seems like a very nifty feature.
I wonder how you detect renaming a table or a column?

On which programming language is dbstewart based?


Regards
Thomas