For the m entries following k, just start a scan at k and read m entries. (Be sure to set the batch size on the scanner.) The m entries before k is harder, unless your keys are very predictable. You might want a separate reverse-ordered index of the keys. To get the reverse order you'll have to flip the bits of the keys manually, before you insert and query and on the results you get back.
Billie On Aug 7, 2013 5:32 PM, "Jeff Kubina" <[email protected]> wrote: > > I have records <key; value> in an Accumulo table where the key is about a 50 long byte string. Given a new key k, I want to find the m records that would precede and succeed the record <k;v> if it were inserted into the table. Any ideas on how I can do this efficiently? The record <k;v> will eventually be inserted into the table. > > -Jeff >
