Limiting the consecutive occurrence

[ntk]

you would create the ' abcd_8 then use this
rrep 'a' 'z' abcd_8

it's blazing quick even for 15GB file size, on 1.86GHz machine
k@k-DIXONSXP:~/dict$ time rrep 'd' 'z' abcd_8
abcd_8: Pattern replaced

real 0m0.133s
user 0m0.120s
sys 0m0.000s

---

tried many ways. more complex, but hard to beat that clean, simple way. If only generating by sed/grep filter and the wee calculation procedures were that quick, nothing could stand you.