The docs say:

> * - 'queue name': The name of the queue to use to queue this task. Must
* contain a valid queue name, declared by hook_cron_queue_info().
* If queue name is given, worker callback will be ignored.

AFAICT this is wrong on two counts.

1.

/**
 * Implements hook_cron_queue_info().
 *
 * Provide queue worker information for jobs declared in
 * hook_cron_job_scheduler_info().
 */
function job_scheduler_cron_queue_info() {
  $queue = array();
  foreach (job_scheduler_info() as $name => $info) {
    if (!empty($info['jobs']) && !empty($info['queue name'])) {
      $queue[$info['queue name']] = array(
        'worker callback' => 'job_scheduler_cron_queue_worker',
        'time' => 60, // Some reasonable default as we don't know
      );
    }
  }
  return $queue;
}

This hook implementation seems to be taking care of declaring queues to Drupal core, based on hook_cron_job_scheduler_info().

This contradicts this statement:

> Must contain a valid queue name, declared by hook_cron_queue_info().

There is the matter that hook_cron_queue_info() skips a scheduler if it has no 'jobs' property. However, AFAICT, if a scheduler info has no jobs, then job_scheduler_rebuild_scheduler() doesn't write anything to the database anyway.

2.

> If queue name is given, worker callback will be ignored

If a queue name is given, then when the scheduler runs, it doesn't execute the worker callback. Instead, it queues a new item in the Drupal queue.

This uses the queue callback job_scheduler_cron_queue_worker(). job_scheduler_cron_queue_worker() executes the job, and executing the job calls the job's worker callback!

So that bit of the docs is wrong too.

Comments

bmelvin1’s picture

bmelvin1 CreditAttribution: bmelvin1 commented

It looks like there are three patterns supported by this module, even though only two are documented in README.

Pattern 1: The Client module declares one or more job schedulers, each having a set of jobs and a callback function. In this pattern, jobs are executed (callback function is called) during JobScheduler’s cron_hook. The system queue is not used. To employ this pattern, Client's implementation of hook_cron_job_scheduler_info should declare 'worker callback' and 'jobs' (but not ‘queue name’) for each scheduler.

Pattern 2: The Client module declares and controls its own queues. JobScheduler enqueues jobs instead of executing them. To employ this pattern, Client should declare queues in hook_cron_queue_info. Client's implementation of hook_cron_job_scheduler_info should provide ‘queue name’ for each queue/scheduler. Pattern 2 is used by the Feeds module.

Pattern 3: (Not documented in README) JobScheduler enqueues jobs instead of executing them but, unlike Pattern 2, queues are declared by JobScheduler. To employ this pattern, Client module’s implementation of hook_cron_job_scheduler_info should declare 'worker callback', 'jobs' and ‘queue_name’ for each job scheduler.

Given the three distinct patterns, it would be difficult to write a clear set of comments for hook_cron_job_scheduler_info! Maybe it would be less misleading to have a comment that says "look at the README file"? And add Pattern 3 to the README.