CSCI 230 -- Data Structures and Algorithms
Project 3 -- Jog Phone Company -- Additional Notes

 

March 17, 1999

In programming the Jog Phone Company simulation and hash table implementation, here are some other things to watch out for:
 
  1. Searching a map with the find member function. In the above example note that in the connections table the entries appear with their keys in increasing order. Sorted order is maintained by an STL map so that searches can be conducted with a binary search algorithm, which is much faster than linear search. (The time required is proportional to logn rather than n.) The main function for searching a map is the find member function, which takes a key k to be searched for as its argument and returns an iterator referring to the entry with that key, if there is one, or the iterator returned by the end member function if there is no entry with key equal to k, as in the following code
    2.   typedef map<double, pair<double, float> > table_type;
  ...
  table_type connections;
  table_type::iterator i;

  double number = ....

  i = connections.find(number);
  if (i != connections.end())
    ... action to take if the number was found
  else
    ... action to take if the number wasn't found
    In case (i != connections.end()), it is valid to use *i to retrieve the key and its associated value, which are stored as a pair.
     
  2. Inserting into a map using array-like bracket notation. There are several ways in which new entries can be inserted in a map. One of the easiest is to use the same bracket subscript notation as you would with an array, as in the following code (which assumes the same typedef and declarations as in the previous part).
    3.   string number1;
  string number2;
  float start_time;

  ... code to give values to number1, number2, start_time

  connections[number1] = pair<double, float>(number2, start_time);
    This means store the value created by the pair<string, float> constructor in the table so that it can be found using key number1. On the left hand side of the assignment, it looks like you are treating connections as an array; this works because the map class overloads the definition of operator[].
     
  3. The map insert member function. Another way of inserting a new entry into a map is with the map class's insert member function. This function takes a pair consisting of a key and its associated value. It returns a pair consisting of an iterator and a bool. The bool value is true if the pair was successfully inserted in the table, false if it wasn't (which would be the case if there was already an entry in the table with the given key). If the insertion was successful, the iterator refers to the new entry; otherwise it refers to the existing entry with the given key. Here is some sample code (again assuming the same declarations as in part 1).
    4.    typedef pair<double, float> associated_type;
   typedef pair<const double, associated_type> entry_type;

   typedef pair<table_type::iterator, bool> pib;

   pib p = connections.insert(
             entry_type(number1, associated_type(number_2, start_time)));
   if (p.second)
     ... action to take if the insertion was successful,
     ... using p.first to refer to the entry with key == number1
   else
     ... action to take if there was already an entry with key == number1,
     ... using p.first to refer to that entry
    Thus insert allows you to check if a key is in the table and, if not, put a new entry with that key into the table in one step--that is, with one search of the table. This is faster than using find and then operator[], which requires searching the map twice, but either way is acceptable for this assignment.
     
  4. Using drand48. For randomly generating phone numbers and simulated times you'll need to use a pseudo-random number generator from the C++ library. Unfortunately the only random number generator required to be in all C++ libraries is rand, which is not a very good generator. If you are using the g++ compiler, a better random number generator is drand48. The rand function has associated with it another function called srand (and drand48 similarly has srand48) which is called to "seed" the random number generator; i.e., to initialize it. The initial seed value determines the entire sequence of values generated, so one can repeat an experiment by reusing the same seed, or one can ensure that a different sequence of values will be generated by using a different seed.