JdbcRepository doesn't override getJob, jobExists, getJobNames

Currently, getJob, jobExists and getJobNames don't override the use of the ConcurrentMap jobs in AbstractRepository. That means that in a multi-instance configuration, where multiple applications share the same database, only those that have at least one job started locally will be able to see that jobName.

From my POV:

getJobNames should do a SELECT DISTINCT on the job_instances table
jobExists should do a SELECT ... WHERE JOBNAME = ?
getJob should ideally trigger ArchiveXmlLoader.loadJobXml with the JobXmlResolver if the jobName is unknown but I guess that isn't available by default.

FWIW: my current workaround is to load all jobs from META-INF/batch up front via ArchiveXmlLoader.loadJobXml and to call addJob on them

@Xiphoseer Thanks for raising this! I'll arrange my time to check this.

I guess this issue is effectively a duplicate of JBERET-66

I want to elaborate on the motivation a bit:

Ultimately, the case where I encountered this was with JobOperator#getJobNames as defined by JSR-352 (and by extension Jakarta Batch 2.0). In our case, this method is used by existing code to trigger multiple calls to JobOperator#getJobInstances to collect all batch jobs in the system. The spec doesn't really say anything about that method, other than the following description in the javadoc:

Returns a set of all job names known to the batch runtime.

What happens though is that JBeret proxies this method to org.jberet.repository.JobRepository#getJobNames in

jsr352/jberet-core/src/main/java/org/jberet/operations/AbstractJobOperator.java

Lines 171 to 173 in c7daf81

    
           public Set<String> getJobNames() throws JobSecurityException { 
        
               return getJobRepository().getJobNames(); 
        
           }

which ends up being implemented only by

jsr352/jberet-core/src/main/java/org/jberet/repository/AbstractRepository.java

Lines 74 to 81 in c7daf81

    
           @Override 
        
           public Set<String> getJobNames() { 
        
               final Set<String> jobNames = new HashSet<String>(); 
        
               for (final ApplicationAndJobName e : jobs.keySet()) { 
        
                   jobNames.add(e.jobName); 
        
               } 
        
               return jobNames; 
        
           }

That implementation is somewhat reasonable for a generic JobRepository, as it matches jobExists and getJob on the same interface. But it's much less obvious from the point of view of the JobOperator. There we'd like behavior that matches the parameters that work for getJobInstances.

I plan to do a major beta release of JBeret today or tomorrow to include several major jakarta API upgrades. After the release is done I'll go on working on this(and maybe include the work into Final release).

	public Set<String> getJobNames() throws JobSecurityException {
	return getJobRepository().getJobNames();
	}

	@Override
	public Set<String> getJobNames() {
	final Set<String> jobNames = new HashSet<String>();
	for (final ApplicationAndJobName e : jobs.keySet()) {
	jobNames.add(e.jobName);
	}
	return jobNames;
	}