Lecture 18


MCS 260 Fall 2021
David Dumas


  • Work on Project 2, due 6pm Fri Oct 8
  • Project 2 autograder now open!
  • Error handling

    Programs sometimes encounter unexpected events:

    • Data has unexpected format
    • File operation impossible (missing, permissions, ...)
    • Variable name does not exist
    • ...many more

    Making a program robust means ensuring it can serve its function even after certain errors occur.

    Error handling approaches

    Three main approaches:

    • Do nothing. Behavior when an error occurs depends on OS and language. Not good!
    • Explicitly check for error at every step (often using return values), report to caller if in a function.
    • Exceptions (to be explained soon)

    Explicit checks at each step

    Build functions that return information, and an indication of whether an error occurred.

            retval, errcode = load_data()
            if errcode != 0:
                # Some error occurred
                print("Unable to load data due to error: ",errmsg[errcode])

    When functions call other functions, this gets complicated. Each one needs to detect and report errors to its caller.


    An exception signals that an unexpected event has occurred, and control should jump to code that is meant to handle it. We say the error "raises" an exception, and other code "catches" it.

    In Python, an exception behaves a bit like break. Just as break searches for an enclosing loop, after an exception Python searches for an enclosing try block that will catch it.


                # code that does something that may raise an
                # exception we want to handle
                # code to start executing if an error occurs
            # line that will execute after the try-except

    Handle input string that is not a number.

            while True:
                s = input()
                    n = float(s)
                    print("Please enter a number.")
            print("Got a number:",n)

    Exceptions are Python's preferred error handling mechanism.

    Often described as "it is easier to ask forgiveness than permission"* or EAFP.

    The contrasting approach, of anticipating error conditions and checking beforehand, is often called "look before you leap" or LBYL.

    * A phrase popularized by US Navy admiral and computer scientist Grace Hopper.

    Uncaught exceptions

    If no try...except block catches an exception, the program ends.

    An error message is printed that also describes what type of exception occurred.

    >>> int(input())
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    ValueError: invalid literal for int() with base 10: 'walrus'

    Some built-in exceptions

    • ValueError - Function got the right type, but an inappropriate value
      e.g. int("apple")
    • IndexError - Valid index requested, but that item does not exist
      e.g. ["a","b"][15]
    • KeyError - A requested key was not found in a dictionary
      e.g. {"a": 260, "b":330}["autumn"]
    • TypeError - Invalid argument type, e.g. non-integer list index:
      e.g. ["a","b"]["foo"]
    • OSError - The OS reported an error in a requested operation; includes many file-related errors (e.g. file not found, filename is a directory, permissions do not allow opening the file, ...)
    • NameError - Reference to unknown variable.

    Catching specific exceptions

                # code that does something that may raise an
                # exception we want to handle
            except ValueError:
                # code to handle a ValueError
            except OSError:
                # code to handle a OSError
                # code to handle any other exception
            # line that will execute after the try-except

    Catching exception objects

            except OSError as e:
                print("Unable to open foo.txt; the error was:\n",e)

    Printing an exception object gives some information about the error. Some exception types carry additional data, like OSError.filename to get the filename of the file the error involves.

    Raising exceptions yourself

    Your functions can raise exceptions using the raise keyword, followed by an exception type.

            raise ValueError("U+1F4A9 not allowed in homework")
            raise TypeError("This function cannot use a complex value")
            raise NotImplementedError("Loading map from JSON not working yet")
            raise Exception("Locusts.")


    Revision history

    • 2021-10-04 Initial publication