We've been looking at this issue, too, and were thinking of implementing
it (in the short term) by exporting to XML, transforming using XSLT,
then importing into a new repository with the new schema. Of course,
it's a pretty shoddy approach really, as it will inevitably mean long
downtimes if you've lots of data. We've also struggled a little with how
to approach this when using versionable nodes. The spec is quite vague
about this, only implying that it should be possible, but going no
further than that.
Of course, something like your DDL example would be ideal, perhaps with
options for how to handle versions ("Truncate history", "Migrate prior
versions", etc.)
Miro
-----Original Message-----
From: Jukka Zitting [mailto:[EMAIL PROTECTED]
Sent: 20 March 2006 23:31
To: [email protected]
Subject: Re: migrating repository schemas
Hi,
On 3/21/06, Brian Moseley <[EMAIL PROTECTED]> wrote:
> to do in-place migrations, jackrabbit would have to support alteration
> of existing node types (adding and removing property types and child
> node types). having looked through the javadoc for
> o.a.j.core.nodetype, i don't see any such api. am i looking in the
> wrong place, or does it truly not exist? seems like a pretty
> complicated feature to add, if it's not already there.
It doesn't exist. There currently isn't even support for removing
unused node types. There's a Jira issue JCR-322 for adding these
features, but nobody is working on it at the moment.
I think the only way to remove or modify an *unused* node type at the
moment is to shutdown the repository and manually edit the
custom_nodetypes.xml file. Just remember to keep good backups...
BR,
Jukka Zitting
--
Yukatan - http://yukatan.fi/ - [EMAIL PROTECTED]
Software craftsmanship, JCR consulting, and Java development
