Fix finding document entries with a bad pacer_doc_id

[ttys0dev]

I think we need a fallback here in case we have entries with bad pacer_doc_id's.

[mlissner]

Thanks! One day, we'll catch all these variations, and retire to a beach, I suppose. Until then, Alberto, can you please review? Seems like we'll want a test for this as well?

[albertisfu]

Thanks @ttys0dev I've added a couple of related tests. They're currently failing due to two issues that may require to be addressed:

1. When removing the invalid pacer_doc_id from the fallback lookup, the RECAPDocument is matched but retains its previous invalid pacer_doc_id. Should we update the incorrect pacer_doc_id with the new one? Currently, the pacer_doc_id is assigned here:
   

   courtlistener/cl/recap/mergers.py

   Line 965 in 3060ac0

   rd.pacer_doc_id = rd.pacer_doc_id or docket_entry["pacer_doc_id"]

   While prioritizing the current pacer_doc_id is generally expected behavior, in this case we'd need to update it from the docket_entry if it's not blank.

2. The fallback lookup without pacer_doc_id can raise a MultipleObjectsReturned error if there is more than one main PACER Document in the docket entry. Should we apply the same logic to select the latest RECAPDocument with PDF and remove duplicates?
   

   courtlistener/cl/recap/mergers.py

   Line 948 in 3060ac0

   except RECAPDocument.MultipleObjectsReturned:

   
   We could create a method for this logic and use it in both try/except blocks.

[ttys0dev]

1. Should we update the incorrect pacer_doc_id with the new one?

I think so, added handling for this.

We could create a method for this logic and use it in both try/except blocks.

Yeah, cleaned this up a bit by making a function for removing duplicates.

[albertisfu]

@ttys0dev thank you! Changes look good.

Just a couple of remaining comments, please.

[albertisfu]

Thanks @ttys0dev for the changes. I've made a small tweak to a docstring and moved the line rds_created.append(rd) as explained in a comment.

@mlissner this looks good to be merged.

[albertisfu]

I moved this line rds_created.append(rd) to apply only in cases where the RD is created. Previous approach was also applying it to existing RDs.