Many databases store the data in the character fields (and especially indices) using prefix compression. This technique compresses a sequence of strings A1, ..., ANby the following method: if there are strings Ai = ai,1ai,2...ai,pand Ai + 1= ai+1,1ai+1,2...ai+1,q
such that for some j <= min(p, q) ai,1= ai+1,1, ai,2= ai+1,2, ... ai,j= ai+1,j, then the second string is stored as [j]ai+1,j+1ai+1,j+2... ai+1,q, where [j] is a single character with code j.
If j = 0, that is, strings do not have any common prefix, then the second string is prefixed with zero byte, and so the total length actually increases.