Switch branches/tags
Nothing to show
Find file History
Kuassim Removed TransactionEnd.java
Dup of TransactionCompletion.java
Latest commit f1ebede Nov 14, 2018



ADBA is Asynchronous Database Access, a non-blocking database access api that Oracle is proposing as a Java standard. ADBA was announced at JavaOne 2016 and presented again at JavaOne 2017. The ADBA source is available for download from the OpenJDK sandbox as part of the OpenJDK project and the JavaDoc is available here. You can get involved in the ADBA specification effort by following the JDBC Expert Group mailing list.

Reading a bunch of JavaDoc and interfaces can be interesting, but it is not nearly as engaging as having actual running code to play with. To that end, we have uploaded the beginnings of an implementation of ADBA running over standard JDBC, AoJ. AoJ is available for download from GitHub under the Apache license. It should run with any reasonably standard compliant JDBC driver.

AoJ implements only a small part of ADBA, but it is enough to write interesting code. It provides partial implementations of DataSourceFactory, DataSource, Session, OperationGroup, RowOperation, CountOperation, Transaction and others. These implementations are not complete but there is enough there to write interesting database programs. The code that is there is untested, but it does work to some extent. The saving grace is that you can download the source and improve it: add new features, fix bugs, try out alternate implementations.

Oracle is not proposing AoJ as an open source project. However, because AoJ is released under the Apache license, the Java community can fork the code and create a true open source project with this upload as a base. Oracle developers may contribute when we have time, but this would have to be a Java community effort.

We could have held this code back and worked on it longer. Instead we thought it better to get it to the community as soon as we could. We hope that you agree.

Building AoJ

AoJ and ADBA require JDK 9 or later. Download ADBA from the OpenJDK sandbox. It does not have any dependencies outside of Java SE 9. Download AoJ from GitHub. Both are modularized so be sure to include the module-info.java files. AoJ depends on ADBA. The AoJ sample file depends on JUnit which is included with most IDEs but is also available here.

To run the sample file you will need a SQL database and corresponding JDBC driver. AoJ has been run with Oracle Database 12c and Oracle Database 12c JDBC, but it should work with any reasonably standard compliant SQL database and JDBC driver. The sample file uses the scott/tiger schema available here.

Start the database and load scott.sql. Edit com.oracle.adbaoverjdbc.test.FirstLight.java and set the constant URL to an appropriate value. AoJ will pass this value to java.sql.DriverManager.getSession. If you are using a database other than Oracle you should change the value of the constant TRIVIAL to some very trivial SELECT query.

Sample Code

The following test case should give you some idea of what AoJ can do. It should run with any JDBC driver connecting to a database with the scott schema. This is the last test in com.oracle.adbaoverjdbc.test.FirstLight.java. For an introduction to ADBA see the JavaOne 2017 presentation.

   public void readme(String url, String user, String password) {
   // get the AoJ DataSourceFactory
   DataSourceFactory factory = DataSourceFactory.newFactory("com.oracle.adbaoverjdbc.DataSourceFactory");
   // get a DataSource and a Session
   try (DataSource ds = factory.builder()
           Session conn = ds.getSession(t -> System.out.println("ERROR: " + t.getMessage()))) {
     // get a TransactionCompletion
     TransactionCompletion trans = conn.transactionCompletion();
     // select the EMPNO of CLARK
     CompletionStage<Integer> idF = conn.<Integer>rowOperation("select empno, ename from emp where ename = ? for update")
             .set("1", "CLARK", AdbaType.VARCHAR)
                     () -> new int[1], 
                     (a, r) -> {a[0] = r.at("empno").get(Integer.class); },
                     (l, r) -> null,
                     a -> a[0])
     // update CLARK to work in department 50
     conn.<Long>rowCountOperation("update emp set deptno = ? where empno = ?")
             .set("1", 50, AdbaType.INTEGER)
             .set("2", idF, AdbaType.INTEGER)
             .apply(c -> { 
               if (c.getCount() != 1L) {
                 throw new SqlException("updated wrong number of rows", null, null, -1, null, -1);
               return c.getCount();
             .onError(t -> t.printStackTrace())
     conn.catchErrors();  // resume normal execution if there were any errors
     conn.commitMaybeRollback(trans); // commit (or rollback) the transaction
   // wait for the async tasks to complete before exiting  
   ForkJoinPool.commonPool().awaitQuiescence(1, TimeUnit.MINUTES);

AoJ Design Spec in 100 words or less

The methods called by the user thread create a network (DAG) of CompletableFutures. These CompleteableFutures asynchronously execute the synchronous JDBC calls and the result processing code provided by the user code. By default AoJ uses ForkJoinPool.commonPool() to execute CompletableFutures but the user code can provide another Executor. When the Session is submitted the root of the CompleteableFuture network is completed triggering execution of the rest of the network.