Sample JSON

gistfile1.txt

{ "ap_mac" : "00:18:0a:xx:xx:xx", "rssi" : 0, "is_associated" : true, "client_mac" : "xx:xx:xx:xx:xx:xx", "last_seen" : { "$date" : 1396571076936 }, "secret" : "xxx", "created_at" : "1396571075209", "_id" : { "$oid" : "533dfbc3fca8c126209a3ce0" }, "__v" : 0 }

last_seen needs reformatting.

ID should be mapped on import. Currently _id doesn't import.

yes, you have to transform that thing.
created_at should become a number, not a string, otherwise the timestamp semantics won't work.
yes, you have to extract the $oid thing from the id field and the $date thing from the last_seen field.

I would suggest using the java client, issuing bulk insert statements (insert into t (id, col) values (1, 'hfjg'), (2, 'some more'), (...)) via the transport. This is the fastest thing you can do besided using copy to. And you would get correct results. Maybe parallelize this file reading, request issueing stuff, split up your dump, you know...