Delete one by one or USE KEYS

[icy44]

A job needs to delete old documents. each batch it deletes about 200 documents. we have the document ids.
which way will be more efficient:

• delete by query:
  delete from bucket use keys [… ids …]
  bucket->query(query)
• one by one:
  loop through all ids and
  bucket->remove(id)

thanks in advance.

[vsr1]

If you know document keys use bucket->remove(id)
You can do asynchronous and parallel too.

[icy44]

Thank you @vsr1. I assume even with Use keys it still needs to be processed by query analyzer, but with remove(), it is just a key-value operation?

[vsr1]

As long has you dont have any WHERE clause it is simple remove in loop.
If you have where clause it will do need to do fetch and remove.
NOTE: bucket.query() first need to send to to n1ql node and which need to remove from data nodes It is 2 hops operations.