## Search by Hashing

5 stars based on 53 reviews

Searching a list of values is a common task. An application program might retrieve a student record, bank account record, credit record, or any other type of record using a search algorithm. Some of the most common search algorithms are serial searchbinary search and search by hashing. The tool for comparing the performance between the different algorithms is called run-time analysis.

Here, we present search by hashingand discuss the performance of this method. But first, we present a simple search method, the serial search and its run-time analysis.

In a serial search, we step through an array or list one item at a time looking for a desired item. The search stops when the item is found or when the search has examined each item without success. This technique is probably the easiest to implement and is applicable to many situations.

The running-time of serial search is easy serial and binary search program in cpp analyze. We will count the number of operations required by the algorithm, rather than measuring the actual time. For searching an array, a common approach is to count one operation each time that the algorithm accesses an element of the array. Usually, when we discuss running times, we consider the "hardest" inputs, for example, a search that requires the algorithm to access the largest number of array elements.

This is called the worst-case running time. For serial searchthe worst-case running time occurs when the desired item is not in the array.

In this case, the algorithm accesses every element. Serial and binary search program in cpp, for an array of n elements, the worst-case time for serial search requires n array accesses. An alternative to worst-case running time, is the average-case running time, which is obtained by averaging the different running times for all inputs of a particular kind.

For example, if our array contains ten elements, then if we are searching for the target that occurs at the first location, then there is just one array access.

If we are searching for the target that occurs at the second location, then there are two array accesses. And serial and binary search program in cpp on through the final target, which requires ten accesses. The average of all these searches is:. Both worst-case time and average-case time are O nbut nevertheless, the average case is about half the time of the worst-case. A third way to measure running time is called best-caseand as the name suggests, it takes the most optimistic view.

The best-case running time is defined as the smallest of all the running times on inputs of a particular serial and binary search program in cpp. For serial search, the best-case occurs when the target is found at the front of the array, requiring only one array access. Thus, for an array of n serial and binary search program in cpp, the best-case time for serial search requires just 1 array access.

Unless the best-case behavior occurs with high probability, the best-case running time is generally not used during analysis. Hashing has a worst-case behavior that is linear for finding a target, but with some care, hashing can be dramatically fast in the average-case.

Hashing also makes it easy to add and delete elements from the collection that is being searched. To be specific, suppose the information about each student is an object of the following form, with the student ID stored in the key field:. We call each of these objects a record. Of course, there might be other information in each student record. Serial and binary search program in cpp student IDs are all in the range The record for student ID k can be retrieved immediately since we know it is in data[k].

What, however, if the student IDs do not form a neat range like Suppose that we only know that there will be a hundred or fewer and that they will be distributed in the range We could then use an array with 10, components, but that seems wasteful since only a small fraction of the array will be used. It appears that we have to store the records in an array with elements and to use a serial search through this array whenever we wish to find a particular student Serial and binary search program in cpp.

If we are clever, we can store the records in a relatively small array and still retrieve students by ID much faster than we could by serial search. In this case, we can store the records in an array called data with only components. We'll store the record with student ID k at location:. The record for student ID is stored in array component data[7]. This general technique is called hashing.

Each record requires a unique value called its key. In our example the student ID is the key, but other, more complex keys are sometimes used. A function called the hash functionmaps keys to array indices.

Suppose we name our hash function hash. If a record has a key of kthen we will try to store that record at location data[hash k ]. Using the hash function to compute the correct array index is called hashing the key to an array index. The hash function must be chosen so that its return value is always a valid index for the array. Given this hash function and keys that are multiples ofevery key produces a different index serial and binary search program in cpp it was hashed.

Thus, hash is a perfect hash function. Unfortunately, a perfect hash function cannot always be found. Suppose we no longer have a student IDbut we have instead. The record with student ID will be stored in data[3] as before, but where will student ID be placed?

So there are now two different records that belong in data[3]. This situation is known as a collision. In this case, we could redefine our hash function to avoid the collision, but in practice you do not know the exact numbers that will occur as keys, and therefore, you cannot design a hash function that is guaranteed serial and binary search program in cpp be free of collisions.

Typically, though, you do know an upper bound on how many keys there will be. The usual approach is to use an array size that is larger than needed. The extra array positions make the collisions less likely. A good hash function will distribute the keys uniformly throughout the locations of the array.

If the array indices range from 0 to 99, then you might use the following hash function to produce an array index for a record with a given key:. One way to resolve collisions is to place the colliding serial and binary search program in cpp in another location that is still open. This storage algorithm is called open-addressing. Open addressing requires that the array be initialized so that the program can test if an array position already contains serial and binary search program in cpp record.

With this method of resolving collisions, we still must decide how to choose the locations to search for an open position when a collision occurs There are 2 main ways to do so. There is a problem with linear probing. When several different keys hash to the same location, the result is a cluster of elements, one after another. As the table approaches its capacity, these clusters tend to merge into larger and larger clusters.

This is the problem of clustering. Clustering makes insertions take longer because the insert function must step all the way through a cluster to find a vacant location. Searches require more time for the same reason. The most common technique to avoid clustering is called double hashing. With double hashing, we could return to our starting position before we have examined every available location.

An easy way to avoid this problem is to make sure that the array size is relatively prime with respect to the value returned by hash2 in other words, these two numbers must not have any common factor, apart from 1. Two possible implementations are:. In open addressing, each array element can hold just one entry.

When the array is full, no more records can be added to the table. One possible solution is to resize the array and rehash all the entries. This would require a careful choice of new size and probably require each entry to have a new hash value computed. A better approach is to use a different collision resolution method called chained hashingor simply chainingin which each component of the hash table's array can hold more than one entry.

We still hash the key of each entry, but upon collision, we simply place the new entry in its proper array component along with other entries that happened to hash to the same array index. The most common way to implement chaining is to have each array element be a linked list. The nodes in a particular linked list will each have a key that hashes to the same value.

The worst-case for hashing occurs when every key hashes to the same array index. In this case, we may end up searching through all the records to find the target just as in serial search. The average-case performance of hashing is complex, particularly if deletions are allowed.

We will give three different formulas for the three versions of hashing: The three formulas depend on how many records are in the table.

When the table has many records, there are many collisions and the average time for a search is longer. We define the load factor alpha as follows:. For open address serial and binary search program in cpp, each array element holds at most one item, so the load factor can never exceed 1.

But with chaining, each array position can hold many records, and the load factor might be higher than 1. In the following table, we give formulas for the average-case performance of the three hashing schemes along with numerical examples. You are given a template implementation of a hash table using open addressing with linear probing.

Here is the source code:.