|
TopCoder Competitions vs. The Real World
Tuesday, April 8, 2003
By nicomp
TopCoder Member
Reality Television has become extremely popular. TV networks combine groups of 'average' people, place them under modest duress, and film the proceedings. Of course the stress of surviving on a jungle island or selecting a beau from 50 attractive strangers pales in comparison to cranking out the 1000 point problem in the last 15 minutes of an SRM. Since most TopCoder competitors prefer ones and zeroes to TV, it's interesting to consider how the SRM paradigm compares to software development in the real world.
TopCoder Competitions relate positively to the real world
The SRM arena offers positive comparisons to software development in the real world. SRM-style problems require competitors to recognize and apply classic data structures and algorithms. Programming on the job makes similar demands. A TopCoder problem is relatively brief but still contains liberal helpings of ancillary verbiage and downright unnecessary information. Life is also replete with red-herrings. From my perspective the actual coding is less important than the ability to visualize the problem solution. I appreciate the skill required to filter through excess information in order to identify algorithms and data structures.
Deadlines are a fact of life. Pressure applied by programming against the clock approximates the pressure of on the job deadlines. A real world deadline, however, is hopefully derived by considering the complexity of the job, the skill of the team, and the requirements of the customer. We should never see an arbitrary time limit applied to a job, but it has been known to occur. Don't expect your supervisor burst into your cubicle and insist that "You've got 35 minutes to deliver that linked list" but don't become flustered if it does happen.
Every decent software engineering text book emphasizes reusable code. I press such a philosophy into my students no matter what language I'm teaching. In order to complete 3 SRM problems in 75 minutes one must have a set of utilities/objects at the ready. In order to develop commercial applications similar types of code must be available. A programmer may choose to either create or obtain. Professional programmers generally disdain writing code from scratch unless absolutely necessary. That's not to imply that the world would be complete with only one binary tree class. On the contrary, seasoned TopCoders can bolster their resumes with their own unique implementations of classic algorithms and data structures. As someone who has hired and fired programmers, I know I'm impressed with those candidates with the skill to write code that can be reused to solve multiple problems.
What's missing from an SRM?
Before I took part in my first TopCoder competition I visited many of the practice rooms and studied the solutions of other competitors. My first reaction (being an teacher) was "What's up with these variable names? Where are the comments?". I shortly realized that no points were awarded for conforming to coding standards. One letter variable names take less time to type than more descriptive ones. Comments only help others during the challenge phase. On the other hand, the mindset of a professional programmer must be that code is written once to be read many times. An effective SRM solution is written once to be tested shortly thereafter. As I write production code I tend to ask myself if I'll be able to understand it in 6 months. I understand that I'll often be required to revisit it. My employees labor under the same expectations. None of us wants our cell phone to ring on Sunday morning because something we wrote is dividing by zero. If such a problem does arise, however, it's comforting to know that our code will still be readable. A future SRM might be structured to require competitors to make specific modifications to past submissions and resubmit for new rounds of system testing.
Most software breaks down into 3 broad components: algorithms, input/output, and user interfaces. TopCoder exercises emphasize algorithmic programming. User interfaces are problematic as they cannot be objectively tested. Low level I/O programming quickly becomes tedious (how many ways can you query a port?). At the high end we find specialized languages such as SQL rather than C++ or Java.
We live in a microwave society. We can prepare popcorn in 3 minutes or we can retrieve 42 biographies of Grace Hopper by simply asking Google. TopCoder provides almost instant gratification with a submission score, system test, and rating points all in less time than the average commute. The development cycle of most production code lasts a bit longer. Gratification tends to come in smaller increments over extended periods. Your code depends on the work of others including salespeople (argh!) and supervisors. Your best work may suffer interminable delays or never reach its intended audience through no fault of your own. Hopefully you will still get paid.
Conclusion
I conclude that TopCoder provides an extremely useful metric for programmers. SRMs do not challenge all aspects of software development but the algorithmic portion of programming is implemented in such way as to mirror Real World endeavors. Any job seeker would be prudent to enhance their portfolio by leveraging the services offered by TopCoder. It's also a lot of fun for geeks like me.
Biography
Bill Nicholson graduated from the University of Cincinnati Engineering College in 1982 with a degree in Engineering Science. His first real job was writing engine control software in 6801 assembly language for General Motors. From there he found work developing warehouse management systems in C and then C++. Currently he owns Nicholson Computer Consulting in Cincinnati, Ohio. He also teaches C++ and Visual Basic at Cincinnati State Technical and Community College.
Would you like to write a feature?
