What the author means by implementing a HashTable as a BST is simply implementing a BST with $insert(), \space delete() \space and \space search()$ with slight modifications
- The node of the BST would have the following structure. $Node \space := (Key,Value)$.
- Insert a key-value pair by performing comparisons on the keys of two different pairs.
- Avoids nodes with redundant keys i.e. before insertion, if a node in the BST with the same key value exists, do not insert it. Else if no node with the same key exists, insert the key-value pair.
- No change in delete function.
- Search returns $null$ if no node's key matches, else return the value associated with the key which was searched.
Then use these functions assuming that these perform the corresponding operations on a HashTable.
Using a self-balancing BST would give $O(log\space n)$ time complexity for all the three functions defined above. And iterating through all the key-value pairs would simply mean doing an in-order traversal of the BST.