/**
* @param IJobSpecification $job
* @return array
*/
protected function insertFields(IJobSpecification $job)
{
$dbw = $this->getMasterDB();
return array('job_cmd' => $job->getType(), 'job_namespace' => $job->getTitle()->getNamespace(), 'job_title' => $job->getTitle()->getDBkey(), 'job_params' => self::makeBlob($job->getParams()), 'job_id' => $dbw->nextSequenceValue('job_job_id_seq'), 'job_timestamp' => $dbw->timestamp(), 'job_sha1' => Wikimedia\base_convert(sha1(serialize($job->getDeduplicationInfo())), 16, 36, 31), 'job_random' => mt_rand(0, self::MAX_JOB_RANDOM));
}
/**
* @param IJobSpecification $spec
*
* @return Job
*/
public function jobFromSpecInternal(IJobSpecification $spec)
{
return Job::factory($spec->getType(), $spec->getTitle(), $spec->getParams());
}
/**
* Register the "root job" of a given job into the queue for de-duplication.
* This should only be called right *after* all the new jobs have been inserted.
* This is used to turn older, duplicate, job entries into no-ops. The root job
* information will remain in the registry until it simply falls out of cache.
*
* This requires that $job has two special fields in the "params" array:
* - rootJobSignature : hash (e.g. SHA1) that identifies the task
* - rootJobTimestamp : TS_MW timestamp of this instance of the task
*
* A "root job" is a conceptual job that consist of potentially many smaller jobs
* that are actually inserted into the queue. For example, "refreshLinks" jobs are
* spawned when a template is edited. One can think of the task as "update links
* of pages that use template X" and an instance of that task as a "root job".
* However, what actually goes into the queue are range and leaf job subtypes.
* Since these jobs include things like page ID ranges and DB master positions,
* and can morph into smaller jobs recursively, simple duplicate detection
* for individual jobs being identical (like that of job_sha1) is not useful.
*
* In the case of "refreshLinks", if these jobs are still in the queue when the template
* is edited again, we want all of these old refreshLinks jobs for that template to become
* no-ops. This can greatly reduce server load, since refreshLinks jobs involves parsing.
* Essentially, the new batch of jobs belong to a new "root job" and the older ones to a
* previous "root job" for the same task of "update links of pages that use template X".
*
* This does nothing for certain queue classes.
*
* @param IJobSpecification $job
* @throws MWException
* @return bool
*/
public final function deduplicateRootJob(IJobSpecification $job)
{
if ($job->getType() !== $this->type) {
throw new MWException("Got '{$job->getType()}' job; expected '{$this->type}'.");
}
$ok = $this->doDeduplicateRootJob($job);
return $ok;
}
/**
* @param IJobSpecification $job
* @return array
*/
protected function getNewJobFields(IJobSpecification $job)
{
return array('type' => $job->getType(), 'namespace' => $job->getTitle()->getNamespace(), 'title' => $job->getTitle()->getDBkey(), 'params' => $job->getParams(), 'rtimestamp' => $job->getReleaseTimestamp() ?: 0, 'uuid' => UIDGenerator::newRawUUIDv4(UIDGenerator::QUICK_RAND), 'sha1' => $job->ignoreDuplicates() ? Wikimedia\base_convert(sha1(serialize($job->getDeduplicationInfo())), 16, 36, 31) : '', 'timestamp' => time());
}
/**
* Register the "root job" of a given job into the queue for de-duplication.
* This should only be called right *after* all the new jobs have been inserted.
* This is used to turn older, duplicate, job entries into no-ops. The root job
* information will remain in the registry until it simply falls out of cache.
*
* This requires that $job has two special fields in the "params" array:
* - rootJobSignature : hash (e.g. SHA1) that identifies the task
* - rootJobTimestamp : TS_MW timestamp of this instance of the task
*
* A "root job" is a conceptual job that consist of potentially many smaller jobs
* that are actually inserted into the queue. For example, "refreshLinks" jobs are
* spawned when a template is edited. One can think of the task as "update links
* of pages that use template X" and an instance of that task as a "root job".
* However, what actually goes into the queue are range and leaf job subtypes.
* Since these jobs include things like page ID ranges and DB master positions,
* and can morph into smaller jobs recursively, simple duplicate detection
* for individual jobs being identical (like that of job_sha1) is not useful.
*
* In the case of "refreshLinks", if these jobs are still in the queue when the template
* is edited again, we want all of these old refreshLinks jobs for that template to become
* no-ops. This can greatly reduce server load, since refreshLinks jobs involves parsing.
* Essentially, the new batch of jobs belong to a new "root job" and the older ones to a
* previous "root job" for the same task of "update links of pages that use template X".
*
* This does nothing for certain queue classes.
*
* @param IJobSpecification $job
* @throws MWException
* @return bool
*/
public final function deduplicateRootJob(IJobSpecification $job)
{
$this->assertNotReadOnly();
if ($job->getType() !== $this->type) {
throw new MWException("Got '{$job->getType()}' job; expected '{$this->type}'.");
}
return $this->doDeduplicateRootJob($job);
}
