Question

I wanted to implement a function which needs the values from a pymongo collection in the reverse order of inserting.

I can think of a couple of ways:

cursor = collection.find(skip=collection.count()-LIMIT_VAL,limit=LIMIT_VAL)
rows = list(cursor).reverse()

This way would be a little bad since LIMIT_VAL is a variable and I do not want to put the entire thing into memory. I do not want to supply -ve value to the skip parameter.

OR

rows = collection.find(sort={'$natural':-1}, limit=LIMIT_VAL)

I read here that you can not really rely on natural sort order. I do not know how that can be unreliable

OR

I insert a value called order_of_insert into each record and do a sort on that ensuring that I get how I have inserted.

Before I implement this I needed to make sure which of the ways would be the best thing to do.

Answer 1

Mongodb ObjectIDs store time-stamp information accurate up to the second. $natural sorts it by this _id. If your app were to insert a document within the same second, the insertion order would be ambiguous. $natural would correspond to insertion order only if your collection were capped.

1. If your application can tolerate mostly accurate insertion order and you need data to persist forever, use $natural without capping the collection.
2. If you need accurate insertion order and the data set is small/semi-persistent, cap the collection and use $natural sorting
3. Otherwise it looks like you’ll need to sort on your own time-stamp or some other field to keep track of order.