This is more of a comment/addendum to Arthur's answer, which is however much to long for a comment. I think the argument in case of the non-terminating decimal expansion needs to be exapnded a bit.
The thing is, if you have say $x=0.50\ldots0\overline{1234}$, and there are $1000$ zero digits between the initial $5$ and the periodical part, then decreasing the number by $\Delta x=10^{-1001}$ will give $x-\Delta x=0.49\ldots9\overline{1234}$ and $f(x-\Delta x)=0.409\ldots$, that means the very small change by $\Delta x$ caused a very big change in $f$.
The argument goes that for a non-terminating expansion of $x$, for each integer $n$ you can find an integer $N > n$ such the the $N$-th decimal digit of $x$ $d_N$ is neither
$0$ nor $9$. That means if you change $x$ by $\Delta x$ with $\lvert\Delta x\rvert < 10^{-N}$, you are changing $d_N$ by at most $1$ (in either direction), and since $d_N \neq 0,9$, that cannot cause an overcarry into the previuous digit (if $x$ was increased) or an undercarry into the previous digit (if $x$ was decreased).
So in either cases, the decimal expansion of $x+\Delta x$ will be the same as for $x$ up to at least the $(N-1)$-th decimal, so their $f$ values will be the same at least until the $(2N-2)$-th decimal digit, as Arthur said.