Table of contents

Introduction

By now, you will have encountered various bugs when programming for this class. Most often, you will try to run your code and see something like this:

Traceback (most recent call last):
  File "<pyshell#29>", line 3, in <module>
    result = buggy(5)
  File "<pyshell#29>", line 5, in buggy
    return f + x
TypeError: unsupported operand type(s) for +: 'function' and 'int'

This is called a traceback message. It prints out the chain of function calls that led up to the error, with the most recent function call at the bottom. You can follow this chain to figure out which function(s) caused the problem.

Traceback Messages

Notice that the lines in the traceback appear to be paired together. The first line in such a pair has the following format:

File "<file name>", line <number>, in <function>

That line provides you with the following information:

The second line in the pair (it's indented farther in than the first) displays the actual line of code that makes the next function call. This gives you a quick look at what arguments were passed into the function, in what context the function was being used, etc.

Finally, remember that the traceback is organized with the "most recent call last."

Error Messages

The very last line in the traceback message is the error statement. An error statement has the following format:

<error type>: <error message>

This line provides you with two pieces of information:

Debugging Process

Running doctests

Python has a great way to quickly write tests for your code. These are called doctests, and look like this:

def foo(x):
    """A random function.

    >>> foo(4)
    4
    >>> foo(5)
    5
    """

The lines in the docstring that look like interpreter outputs are the doctests. To run them, go to your terminal and type:

python3 -m doctest file.py

This effectively loads your file into the Python interpreter, and checks to see if each doctest input (e.g. foo(4)) is the same as the specified output (e.g. 4). If it isn't, a message will tell you which doctests you failed.

The command line tool has a -v option that stands for "verbose."

python3 -m doctest file.py -v

In addition to telling you which doctests you failed, it will also tell you which doctests passed. I personally find that information unnecessary, so I usually leave -v out.

Usually, we will provide doctests for you in the starter files. Always run these doctests. Submitting an assignment without running doctests even once is practically throwing points away.

Also, do not manually type in doctests into the interpreter. The whole point of writing doctests is so you don't have to do that. Manually typing in doctests requires you to (1) open up a Python shell, (2) type in every doctest, (3) manually check if the output matches the doctest, (4) repeat. Running doctests from the command line requires you to (1) type in a single line, and (2) that's it!

Writing your own tests

In addition to doctests, you can write your own tests. There are two ways to do this: (1) write additional doctests, or (2) write testing functions.

Writing your own tests is good practice for the future. Remember, before the project deadlines, our autograder only runs sanity tests -- a subset of all the tests we will eventually run. In other words, passing the autograder does not mean you get full credit. As such, it is a very good idea to write your own test cases.

To write more doctests, simply follow the style of existing doctests. You can also write your own functions (much like the take_turn_test function from Project 1).

Some advice in writing tests:

Using print statements

Once the doctests tell you where the error is, you have to figure what went wrong. If the doctest gave you a traceback message, look at what type of error it is to help narrow your search. Also check that you aren't making any common mistakes.

When you first learn how to program, it can be hard to spot bugs in your code. One common practice is to add print statements. For example, let's say the following function foo keeps returning the wrong thing:

def foo(x):
    result = some_function(x)
    return result // 5

We can add a print statement before the return to check what some_function is returning:

def foo(x):
    result = some_function(x)
    print('result is', result)
    return other_function(result)

If it turns out result is not what we expect it to be, we would go look in some_function to see if it works properly. Otherwise, we might have to add a print statement before the return to check other_function:

def foo(x):
    result = some_function(x)
    print('result is', result)
    tmp = other_function(result)
    print('other_function returns', tmp)
    return tmp

Some advice:

Long-term debugging

The print statements described above are meant for quick debugging of one-time errors -- after figuring out the error, you would remove all the print statements.

However, sometimes we would like to leave the debugging code if we need to periodically test our file. It can get kind of annoying if every time we run our file, debugging messages pop up. One way to avoid this is to use a global debug variable:

debug = True

def foo(n):
    i = 0
    while i < n:
        i += func(i)
        if debug:
            print('i is', i)

Now, whenever we want to do some debugging, we can set the global debug variable to True, and when we don't want to see any debugging input, we can turn it to False (such a variable is called a "flag").

Error Types

The following are common error types that Python programmers run into.

  1. SyntaxError
  2. IndentationError
  3. TypeError
  4. NameError
  5. IndexError

Common Bugs