In Rudin's book "Principles of Mathematical Analysis" there is a definition for sets to be disconnected: $A\subset X$ is disconnected if there exists an open cover $\{U,V\} $ of $A$ such that $A\subset U\cup V$ and $\bar{U}\cap V=U\cap\bar{V} =\emptyset $, under the hypothesis that $X$ is a metric space. Does this holds for an arbitrary topological space, i.e. for $X$ to be a topological space, a set $A\subset X$ is disconnected if and only if there exists an open cover $\{U,V\} $ of $A$ in $X$ such that $\bar{U}\cap V=V\cap\bar{U} =\emptyset $ ,$A\cap U\neq\emptyset $ and $A\cap V\neq\emptyset $? If it is wrong, is there an counterexample?
--------------------------------------------------Updated
I have made two main mistakes. What I actually want is a conclusion that from the disconnectedness of a subspace $F$ of a given topological space $X$ I can find two disjoint open sets in $X$ such that they forms an open cover of $F$ and $F$ intersects with both sets, but it is wrong because the property is equivalent to the complete normality of a topological space(in Munkres' book "Topology" there is an exercise talking about this, see section 32), so it cannot be true for an arbitrary topological space. Another mistake I make is that I have misread the definition of connectedness in Rudin's book.
Counter example. A = (0,1) = U, V = (2,3).
As U and V are open, it is prefered to simply require
them to be disjoint instead of separated.
Did not Rudin require a point of A be in U and
another point of A to be in V?