K
kirchner
Guest
Hello Octavio, I have never seen formal documentation on this matter, but based on some experience evaluating this stuff I believe the queries you showed us will be resolved on the server. It will envolve lots of reads and residual filtering but the million records won't transfer to the client. About your proposed D index, make it D(field3, fieldX), where fieldX is the most selective of field1 and field2. No point in adding both columns for this query unless you think the index can be used elsewhere.
Continue reading...
Continue reading...