Developing projects that support multiple Databases

add to with other 0 users
It is a common scenario in product/project development that there is a requirement to support multiple databases. This actually leads to 3 challenges - generating DDLs for the DBMS, keeping the queries in Data Access Layer portable across the DBMS platforms and optimizing the read SQLs for the specific DBMS platform. This blog entry will cover only introduce the probable solution to the first of the three challenges.

DDL usually differs from one DBMS to the other. When it comes to generating DDLs there is a issue of creating or altering a table. i.e. when the DDLs are being executed for the first time it should create the database, whereas on consecutive execution it should only execute the alter table table columns to ensure that only the changes are taking effect. There is also the issue of being able to generate the DDLs when required - both the whole DB script and the alter table script. We would also like the script to be generated and executed only on specific cycles of build process.

After searching the internet the tool that caught my eye is the DDL Utils project of the Apache DB Project. This tool fulfils the requirements described above when used in conjunction with Maven. Have a look at the following sample project to see how it can be achieved.
Sample Project
After decompressing the sample project please go to the maven-plugin folder and execute the shell script. Windows users can simply use the same command in it :). As the example uses MySQL please check the connection parameters in the /src/main/resources/com/smartitengineering/.../schema/

The project is build around maven and thus the DB operations are also attached with the Maven commands compile, install and deploy. Their outcomes are as follows:
  • mvn compile - Generates the DDL for the whole schema regardless of the current DB state
  • mvn install - Generates the DDL required to sync the current schema to the DB; that is generates the minimum possible alter table statements
  • mvn deploy - Execute the alter table ddls on the DB to sync with the latest state
The first 2 only generates the DDLs but does not execute them. You can find the DDLs in the target folder. Another thing to note is that the POM file takes the DB parameters as properties so they can lie anywhere. Currently I am reading it from a properties file in the classpath and the additional plugin is required for this purpose. Important links of the tool are -
DDL-Utils Ant Tasks
Schema Documentation
In order to get the deploy working you will need a valid repository to deploy. So that is the only thing that needs to be edited in the POM file.

All in all the tools is very useful handy when it comes to DDL generation and history maintenance. The only thing I found missing in the tool's release is support of ON UPDATE/DELETE CASCADE/RESTRICT/.. support. Though the schema supports it, the implementation does not support it; but considering the whole this is quite a useful tool. The good news is that the source code in the SVN has support for it during creation only and not in altering, so we can hope to see it soon as well


  1. nice findings, it seems like, i have to maintain a xml file.
    in ruby on rails, we use migration script. which is based on ruby and easy to use and multi versioned.

    this seems good but not yet much promising for me. perhaps maintaining multi verionsed ddl yet far better.

    best wishes,

  2. The good thing is that I dont have to do anything even though I am changing the DB platform during project life cycle. Furthermore developers have the liberty to use the DB they like of their choice during development and creation of new tables is part of the Maven build cycle. I think in Java based server side application development this is a huge plus.

  3. Not sure but, i was forced to add -DgeneratePom=true in plugin-install-command to get it work properly.

  4. Thanks for the feedback; if anyone faces any issues with the shell script they know what to do :).