Something pointed out by Adam Vessey w/r/t this line here: https://github.com/Islandora/islandora_batch/blob/60435a2d500140c3e2b74515bbde01029dc1dc49/includes/ingest.batch.inc#L422
There doesn't seem to really be any benefit in replacing the serialized preprocessor object with the serialized ingested object. If the object has already been ingested, we don't really need to have a copy of it hanging around in the database; one can simply load the actual Fedora object.
Overwriting the preprocessor object also causes two problems that make both developing and running batches a pain:
- It makes it much more difficult to diagnose issues that may have arisen during the batch processing phase, as information unique to the preprocessor object is lost
- It makes it impossible to purge and re-run a partial or complete set in the case where issues did arise during the batch processing phase that the batch processor was unable to automatically diagnose (e.g., incorrect derivative configuration settings, files missing or corrupt at the source, batch running from a mounted file system that disappeared); the set must be removed and the preprocessor re-run, which will be a waste of resources and/or time if it was possible to just resolve the issue and purge/re-run.
Proposing we just remove this line and hang on to the preprocessor object in all cases