Date

Development Tips

I have been writing programs since I was in 9th grade, that would be 35 years (Yikes!)  I have more than put in my 10,000 hours to attain some level of mastery, but I like to think of it as wisdom.  Studies have shown that there can be huge differences in the ability level of even the most experienced coders.  More than a factor of 20, in some cases! I think beginning and intermediate CS students really struggle with this. They see some of their classmates streaking on ahead of them effortlessly writing lines and lines of code. But for them its a constant struggle. I've watched some of my students really struggling with the latest homework assignment, and I felt compelled to write down some "words of wisdom" to try and help them through the rest of the semester.

I have been lucky enough to work with some of the best developers in the world during my career.  Here is a summary of  some tips and things that I have observed.  I practice all of these things in my own development work, and you should too.  In my opinion there is no more valuable advice than the first section on Incremental Development. But, I would love to hear from friends, and former students. What tips would you give beginning to intermediate students on writing software?

Incremental Development

  1. Think before you code.  Before you sit down and start coding, try drawing a picture that illustrates your objects, and the relationships between those objects.  Its important to understand not only the basic functionality of each object, but how they interact.
  2. Make a plan.  After you have an idea of the big picture now make a plan for what you want to do first, second, etc.  Nothing is more satisfying than knowing you have been productive.  I have known CEOs of companies that delight in creating a hand-written todo list each morning, for no other reason than crossing something off the list gives them a little boost of joy and keeps them going through the day.  Everyone likes that feeling of being productive.  Making progress on your plan makes you feel productive, and motivates you to keep going.
  3. Start Small.  Start with a small class, for example die, and write a little test program to make sure that the die class is working properly.
  4. Test as you go.  Nothing frees you up to work on the next thing, like knowing you can build on something you have already completed.  Doing a good job of testing is what gives you that freedom.  I know, it is tempting, to look at a small class and read the code and simply assume that it works.  Most of the time this is a bad assumption, even in simple examples.
  5. Use Stubs.  Even in a relatively small class it is tempting to try to implement all of the methods and then compile.  This may not be a very good idea.   What you can do is write a stub for each class.  For example if you have a method for your cup called shake, and you are not ready to write everything in shake just make a function called shake with an empty body.  This will compile fine, but it won't do anything.  If your method is supposed to return a value, you may need to have it return some bogus value for the time being.
  6. Build on success.  Once you have one class working then go on to the next.  Test that class, and test its interaction with the class or classes that came before it.
  7. If you still have classes to write GOTO 3
  8. Integrate your classes and write your main, but again, start small.  Don't try to add everything at once.  For example in the Yahtzee command line game, start with the roll command and make sure you can roll all of your dice.  Then add each command one at a time.
  9. Always keep a working program working.  Notice that if you use this approach you always have something that compiles and runs that you can hand in.  Even a simple example program of a die class with a die that you can roll is worth more points that nothing at all, or a program that simply does not compile.

Use The Tools

  1. Early on in the development, you should create a directory for this project.  This keeps everything together and reduces the clutter.
  2. Use git.  -- Start right away with a git init in your directory, and commit your changes often.   This always gives you a road back.
  3. Use a Makefile
  4. When debugging, make sure you sprinkle in plenty of cout statements.  If you are getting unexpected seg faults, use gdb to find out where the program is crashing.
  5. For goodness sake, use the internet.  Stackoverflow is your friend. cplusplus.com is your friend,  You should not feel like you ought to have command of everything off the top of your head.  Maybe after several years of C++ programming you will have all of the options memorized, but until then, use the documentation.  Even googling large parts of a compiler error message may lead you to a solution.

Debugging

  1. Talk it out. Either to yourself, or even better, to someone else The number of times that I sit in my office and simply ask a student to explain their code to me is amazing. Very often, while that explanation is in progress a lightbulb goes on, and the student sees the problem. I wish it was because of my mystical professorial powers, but more often than not it is by telling someone else that we are forced to...
  2. Confront your assumptions. Way too often we are conviced our code is correct because we don't see what we actually wrote. In most cases we see what we think we wrote. By explaining to someone else we are forced to see what we actually put in the file. In some extreme cases I find that students are convinced that there must be some other evil force in the universe that is causing their program to fail, becuase they have done everything right. This has never been the case yet. The programs that we write for class are not going to be the programs that find new bugs in the C++ compiler or the Python programming language.
  3. Sometimes you just can't beat good old fashioned pen and paper style tracing. In this world of fast turnarounds, instant compiles, and fancy IDE's we too often get caught up in making quick changes to the program just to see if that fixes the problem. Usually this is just masking some logic error or assumption we are making. Pen and paper style tracing is another way to force our minds into focusing on what is there instead of what we assume is there.

Those Pesky Compiler Errors

  1. Yes, C++ error messages are voluminous, cryptic, and downright frustrating in many cases!!  There's nothing you can do about that. except:
  2. Did I mention you should start small.  If you write your code in small chunks, and compile frequently this will reduce the number of compiler errors, and will help you focus on where the error is likely to be.. Hint:  in the new stuff you have just written.
  3. Be systematic. If the code compiled two minutes ago, and you have just added three new lines of code, and now it does not compile you should have a pretty good idea of where the error was introduced.
  4. Start with the first error!!   Its tempting to look at the error message at the end of the list. because thats what you see as soon as the compiler is done.  But, scroll back and start with the first one.  Many of the errors later on are the result of an error at the start.  If you focus on fixing the wrong error, you are likely wasting your time.
  5. Pay attention to the parameters and their Types.   Very often the questions I get boil down to the fact that someone is trying to call a function but is passing a parameter of the wrong type.  Some words and phrases to pay attention to that may help you identify these errors:   "candidate function not viable" ,  "no known conversion from  XXXX to YYY"
  6. Use redirection!  Sometimes its hard to find the first error.  You can change your compile command, or run it manually and send the output to a file.  For example: g++ -c cup.cpp > compile.out 2>&1
  7. Sometimes the best way back is to comment some things out.  If you have followed the start small philosophy this should not be an issue.  But maybe you didn't and now you have a big pile of code that doesn't compile.  Once again you need to break it up some how.  Sometimes the way to do this is to comment out a big chunk of your code and see if the error goes away.

Dealing with Complexity

  1. There is no doubt that the programs we write in this class are more complicated than any you have ever written.  Often by an order of magnitude or two.  This can feel overwhelming, and is yet another reason for you to go back and reread the first section of this document again.
  2. Embrace Abstraction. At the time you are creating a die class you may feel its a waste of time.  Why not just a variable that we assign a random number to?  Or the cup class, why not just use a vector?  But as the project progresses you will see that using the abstraction has some  huge benefits:
    1. It makes the interesting parts of your code much more readable.
    2. It reduces the amount of repetition you need to do.
    3. If you follow good practice, it gives you confidence that the little things in your program are working correctly.
  3. Learn and embrace the KISS (keep it simple stupid) principle. Too often I see people making things way harder than they need to be. You don't have to reinvent the wheel for every assignment. Learn to reuse code from previous assignments.
  4. Part of dealing with complexity is having some confidence that what you have done is correct.  Working in small chunks, and testing in small chunks helps you develop that confidence.
  5. The tools we have talked about in this class are there to help you deal with complexity.  Make, gdb, grep, and others are there to make your life better.
  6. Don't worry if you feel like you don't know everything.  having knowledge available to you comes after time, and frequent practice.  You are not going to be there after one semester.


Comments

comments powered by Disqus