Algorithm for Detecting Duplicate/Previously Encountered Strings

Environment: Straight C++

Overview

A challenge I've seen pop up often while writing search tools is the ability to verify that your code is not traversing the same source multiple times. A spider traversing the web, for example, would never wish to traverse the same URL twice in a session - to do so would merely allow circular loops and endless headaches.

To address this issue, I wrote an N-tree based search algorithm that breaks strings down by their leading characters. Identical portions are 'clipped' and stored in a single node, with the remainder of the string in similarly divided child nodes.

Example

Inserting the following text:
1234
12345
123456
1234567
www.another.site.org/index.html
www.site.org/index.html
www.test.com/index2.html
www.test.com/index.html
www.test.com

Results in this data structure
+-NULL
  +-1234
    +-NULL
    +-5
      +-NULL
      +-6
        +-NULL
        +-7
  +-www.
    +-another.site.org/index.html
    +-site.org/index.html
    +-test.com
     +-NULL
     +-/index
        +-2.html
        +-.html

The advantage to this approach over typical linear searches is two-fold.

  1. The memory overhead can be reduced as redundant text is partially eliminated
  2. Since the tree is ordered, a search can be performed in O(Log(N)) time (someone please correct me if my estimate of this algorithm is incorrect!)

On a 600 MHz PIII system, the following benchmarks were obtained for inserting 3000 random strings of length 0-254 characters into an array that already contains 12,000 strings.
Linear search: 2.7 seconds
StringTree: 0.02 seconds
A reasonable improvement!

The attached code implements the algorithm as a CStringTree class, and provides a test / performance driver to demo the code.

Updates

Changes to current source (zip file below):

  • Fixed several small memory leaks
  • Added significant functionality for searching and File / CStringArray I/O
  • Compiler defs for Borland compilers
  • String counting and data association per node

Downloads

Download demo and source code - 12 Kb


Comments

Leave a Comment
  • Your email address will not be published. All fields are required.

Top White Papers and Webcasts

  • CRM has long been seen as a must-have sales tool. However, much of the value of traditional CRM accrues to managers, not the reps that use them daily. Learn how CRM designed for the individual benefits the entire sales organization from increased data quality to more predictable revenue pipelines.

  • When individual departments procure cloud service for their own use, they usually don't consider the hazardous organization-wide implications. Read this paper to learn best practices for setting up an internal, IT-based cloud brokerage function that service the entire organization. Find out how this approach enables you to retain top-down visibility and control of network security and manage the impact of cloud traffic on your WAN.

Most Popular Programming Stories

More for Developers

RSS Feeds

Thanks for your registration, follow us on our social networks to keep up-to-date