2 cents to improve Nickel
1. Nuisance with floats:
big integer property values in JSON, e.g. "linkid": 519596977 are returned as 5.19596977e+08.
2. Why there are no inner/outer joins? I understand that main loop logic uses _all_docs view (and joins would multiply the load for the same data). But it would be very easy to look up the doc by reference id, at least in the same bucket.
(Errors in N1QL docs: "Scalar Functions" section has copy/paste of Description from ROUND( value, digits ) to RTRIM, TRUNC and VALUE. Again, what VALUE() does?)
Yes, we need to improve number handling and distinguish floats from ints. We're currently using standard libraries for JSON number I/O, so we may need to do something custom for this. See http://www.couchbase.com/issues/browse/MB-9370
Inner/outer joins by primary key: these have been designed and will be added by the time of Beta / GA.
Docs: thanks, will be fixed.
Thanks for the feedback.