Getting rid of JobIdStealer.java

[maoo]

As documented in the class itself....

 * This is a hack to get the jobId given a {@link org.apache.manifoldcf.crawler.system.SeedingActivity}.
 * TODO: If a way to get the job id from within a connector in ManifoldCF is found, delete this.

This class is used to access the Job ID from the Alfresco Manifold Connector. We use the Job ID as identifier (primary key) of entries that we log (using CrawlLogger.java) into the connector to keep the state of the crawling (Last Transaction ID, Last ACL Changeset ID)

In order to solve this issue it's possible to:

• Find an alternative way to access the Job ID from a Manifold Connector
• Use something different from the Job ID to use as primary key that is guaranteed to be unique across different job executions

[maoo]

Apache Manifold is planning on adding a Seeding Version String, that can be used to replace the JobID and therefore deprecate JobIdStealer - https://issues.apache.org/jira/browse/CONNECTORS-971

[maoo]

I can see the following method being added to BaseRepositoryConnector.java

  @Override
  public String addSeedDocumentsWithVersion(ISeedingActivity activities, Specification spec,
    String lastSeedVersion, long seedTime, int jobMode)
    throws ManifoldCFException, ServiceInterruption

I can use lastSeedVersion to store lastTxnId and lastACLChangesetId, divided by char '|' and deprecating JobIdStealer and CrawlLogger classes:

      StringTokenizer tokenizer = new StringTokenizer(lastSeedVersion,"|");
      long lastTransactionId = 0;
      long lastAclChangesetId = 0;

      if (tokenizer.countTokens() == 2) {
         lastTransactionId = new Long(tokenizer.nextToken());
         lastAclChangesetId = new Long(tokenizer.nextToken());
      }

However, I don't know how to update the lastSeedVersion as soon as I collect the last docs processed; the old syntax is

crawlLogger.log(JobIdStealer.stealId(activities), lastTransactionId, lastAclChangesetId);

I was expecting to use a syntax like

super.setLastSeedVersion(lastTransactionId + '|' + lastAclChangesetId);

but maybe I'm misinterpreting this feature.

[maoo]

As confirmed by Manifold Committers, using

return lastTransactionId + "|" + lastAclChangesetId;

was enough to update the lastSeedVersion; tests are passing, now it needs some integration testing

[maoo]

Many thanks to @OpenPj and Karl Wright for the support on this issue!