Re: [Exist-development] thoughts about next steps with RDBMS integration

[James F. <jam...@ex...>]

On 31 March 2010 15:03, Adam Retter <ad...@ex...> wrote:

> On 31 March 2010 06:45, James Fuller <jam...@ex...> wrote:
> > whilst doing xproc development, I have had great fun integrating some of
> > eXist existing extension functionality as extension steps ... a thought
> > occurred to me about the current state of integration with RDBMS.
> > Currently we have the nice SQL extension module (written by Adam Retter)
> > which uses JDBC to query a relational database ... this means we can
> easily
> > query an RDBMS through xquery using SQL statements. I think this module
> > enables a lot of scenarios but misses off the primary scenario for the
> > masses of people out there with data currently existing in RDBMS but want
> to
> > use XML database.
> > I would like to
> > install eXist and provide in some registry a connection string which
> > effectively 'mounts' the RDBMS as a collection ... from which here on out
> I
> > can use pure xquery to work on. For now a read only view of the data is
> all
> > I need but someday I would like to be able to do updates as well on the
> > RDBMS data. Lastly I should be able to configure cache characteristics if
> > the database connection is intermittent'
>
> Phew!!! For a moment then I thought you were going to ask for a SQL
> layer atop eXist-db and a JDBC driver for eXist-db.
>
> This would actually be pretty easy to achieve :-)
>
> Imagine the collection namespace root starting at /sql as opposed to /db.
>
> Configuration could go into
> /sql/system/sql/{database}/{schemaName}/{tableName}
>
> And the database configuration could again be defined in conf.xml
> files in the /sql/system/sql collections. Caching and connection
> property characteristics could also be placed in conf.xml files in the
> appropriate sql configuration collection.
>
> Almost all of the existing infrastructure could be reused. You would
> simply need to extend the Collection function so that in the case of a
> "/sql" collection it defered the document set building to code that is
> very similar to that of the sql module. You would also need to come up
> with a collection xml serialization scheme for displaying the
> available sql schemas and tables as sub-collections but thus should
> not be very hard and should be as close to possible (if not the same)
> as the existing ones for the db collections.
>
> This is quite exciting - I like :-)
>
> Can I help?
>
>
thx Adam for teasing out the detail ... this is a cool way to approach it
... I would appreciate a chat to discuss; my initial thoughts were to create
an XQuery app (using sql extension module) to achieve the above ... its a
bit high level for my tastes for this kind of thing, but perhaps the best
first step.

J
J