In Mongodb's Mapreduce, Before I think there was "db"(like db.anotherCollection.find()) object accessible inside map function. But this features has been removed(from version 1.6 or so on), which make difficult in case of join. what was the reason? why it has been removed?
1
-
Due to security concerns, plus it was never a good idea to do it anyway. Also MR via a JOIN? Please don't even trySammaye– Sammaye2014-01-09 15:29:08 +00:00Commented Jan 9, 2014 at 15:29
Add a comment
|
1 Answer
As at MongoDB 2.4 there are several reasons to disallow access to the db object with from within Map/Reduce functions including:
Deadlocks: There are potential deadlock scenarios between database and/or JavaScript locks called from within the same server-side function.
Performance: The Map/Reduce pattern calls
reduce()multiple times; each iteration is a different JavaScript context and would have to open new connections to the database and allocate additional memory for query results. Long-running JavaScript operations will block other operations.Security: Cross-database queries require appropriate authentication checks.
The above issues could be further complicated for Map/Reduce jobs reading or writing to sharded clusters. The MongoDB Map/Reduce implementation is currently only designed to work with data from a single input collection, and any historical abuses of db object within Map/Reduce functions should be considered a bug rather than a feature.
If you want to merge data with Map/Reduce, you can use an Incremental Map/Reduce. Depending on what outcome you are trying to achieve, there are other approaches that may be more straightforward such as adjusting your schema or doing joins in your application code via multiple queries.
