You've got string $*s*$, consisting of only lowercase English letters. Find its lexicographically maximum subsequence.

We'll call a non-empty string $*s*[*p*_{1}*p*_{2}... *p*_{*k*}] = *s*_{*p*_{1}}*s*_{*p*_{2}}... *s*_{*p*_{*k*}}(1 ≤ *p*_{1} < *p*_{2} < ... < *p*_{*k*} ≤ |*s*|)$ a subsequence of string $*s* = *s*_{1}*s*_{2}... *s*_{|*s*|}$.

String $*x* = *x*_{1}*x*_{2}... *x*_{|*x*|}$ is lexicographically larger than string $*y* = *y*_{1}*y*_{2}... *y*_{|*y*|}$, if either $|*x*| > |*y*|$ and $*x*_{1} = *y*_{1}, *x*_{2} = *y*_{2}, ... , *x*_{|*y*|} = *y*_{|*y*|}$, or exists such number $*r*$ $(*r* < |*x*|, *r* < |*y*|)$, that $*x*_{1} = *y*_{1}, *x*_{2} = *y*_{2}, ... , *x*_{*r*} = *y*_{*r*}$ and $*x*_{*r* + 1} > *y*_{*r* + 1}$. Characters in lines are compared like their ASCII codes.

Input

The single line contains a non-empty string $*s*$, consisting only of lowercase English letters. The string's length doesn't exceed $10^{5}$.

Output

Print the lexicographically maximum subsequence of string $*s*$.

Note

Let's look at samples and see what the sought subsequences look like (they are marked with uppercase bold letters).

The first sample: aBaBBA

The second sample: abbCbCCaCbbCBaaBA