Some more thoughts...
If you're dynamically changing the DB schema and entity code, where is
your business logic in this scenario and how are you not having to
redesign the UI layer after every minor change?
Seems that the controlled approach for updating the app would be to:
1) Setup a Dev/Staging environment with current app/DB
2) Generate updated schema XML and DDL scripts
3) Alter the test db
4) Regenerate the Java entities from the XML schema (and reapply any JPA
or Bean Validation annotations)
5) Update the business logic and presentation layer
6) Package up a new version of your app war/ear
7) Deploy db updates to production
8) Deploy new app to production
That way, you have a controlled roll-out of changes and can rollback to
a previous version if needed....
-Donald
C N Davies wrote:
Hmm thanks Donald I'll take a look at these projects..so much I don't know
:)
-----Original Message-----
From: Donald Woods [mailto:[email protected]]
Sent: Friday, 24 July 2009 2:31 AM
To: [email protected]
Subject: Re: Dynamic entity definition creation - is it possible?
Just taking a stab at this, as I've never tried it....
1) Use DdlUtils to dump the existing DB schema to XML, update the schema
with the new columns and then alter the DB -
http://db.apache.org/ddlutils/
2) Use Torque to generate Java classes from the schema XML -
http://db.apache.org/torque/index.html
3) Use the new JPA2 Metamodel and Criteria API features in OpenJPA trunk
- use the metamodel generator in scripts/mmg.bat to create the
*_.java classes
- use the criteria api and metamodel updated classes for your
persistence code
I may be mistaken about #3, but Pinaki can correct me and provide more
details on if those new JPA2 features apply here.
-Donald
C N Davies wrote:
I might be thinking in the wrong direction so feel free to berate me if I
am.
My app imports data from various sources including excel (uses Apache
POI),
CSV, SQL result set etc. Prior to starting to use OpenJPA, if my import
processes noted any changes to the incoming data set (ie additional
fields)
I would dynamically generate SQL to modify the data tables to accommodate
any changes to the incoming data set. (ie a vendor adds a new column to a
spreadsheet I imports regularly so I add the required columns) and then
commit the new data so that any changes to the structure of the incoming
data can be dynamically managed at runtime.
Since JPA requires class files to define the structure of an entity, if I
have an entity class that expects certain fields, but I receive an
additional column in my incoming dataset, then my entity will just miss
the
new data.
I was thinking that maybe there is a solution that would function at
runtime
like this:
1. Examine the incoming data from a CSV file for example (easy to do
and already done)
2. Create the appropriate entity definition at runtime (possible
???)
3. Create an instance of this new definition (assuming 2 can be
done
this should be as per usual)
4. Populate the new instance (assuming 2 can be done this should be
as
per usual)
5. Call OpenJPA persist/merge (What about JPA enhancement of the
modified/generated entity classes?)
I do understand the implications of doing this kind of dynamic
restructuring
in a production system, however I'm painting a generic picture of my
thought
process here in order to convey a concept rather than babble on about the
specifics of the system.
Have I lost the plot or can this be done using OpenJPA? I've read the
whole
doc a dozen times, search the mail lists but can't see any way to do it.
Thanks!
Chris
