使用二叉树实现字典

Question

I read at multiple locations that Binary Trees should be preffered over Hash Tables where memory is limited as Binary Trees would keep the data in sorted order while Hash Tables won't. The tradeoff being Binary Trees would have a O(log n) lookup and insertion as compared to constant time insertion and lookup in case of hash tables.

I was wondering if I choose Binary Trees to implement my Address Book(always sorted), what should the key look like? would value be the name and number pair?

Answer 1

BTree != binary tree. I'll assume you mean the latter.

How your address book would look depends on what lookups you want to do. If you want to find the address of a person given their name, then the keys are names and the addresses/numbers are the values.

If you want lookup from addresses to names, then you just reverse the keys and the values. If you want two-way lookup, you'll need two trees per address book.

Note that binary tree-based dictionaries are available in the C++ standard library as std::map in the header <map> . Don't roll your own unless you want a programming exercise; std::map is very hard to beat in terms of performance and features.

Answer 2

When considering something like an address book, how the data will be used (lookups) should take precedence over how the data will be stored (physically in memory).

If your name is the key, how would that be stored? "First Last", "Last, First", etc. That will impact how the lookup can be performed. What if you want to lookup by number (for instance, if you start typing a phone number and you want matching contacts to popup)?

I'd suggest really thinking about how you plan to use the address book so that you don't inadvertently tie yourself down to some implementation before you've fully understood the requirements.

Given that, I'd have a look at boost::multi_index . That library provides a set of data structures that can be used to store the data once, and provide an arbitrary number of "lookup keys" against it. Very handy for cases like this.

Answer 3

I think the key is just the name. The number is the value. You search by name, right?

And, if you want good performance, use AVL trees. That way, all subtrees has the same (± 1) deepness.

Answer 4

For a dictionary a type trie and the data structure binary tree is a good choice. It'a also name crit-bit tree, radix-tree or patricia tree. A simplier trie with a hash-key is a kart-trie where the hash is used to define left and right but it's data structure is still a binary tree. Then there is a ternary trie but it's data structure is from a B-tree and it has 3 leafs.