May 18, 2017 1:00:30 PM | How to Debug: Best Practices

A brief glimpse at some of the best practices on how to properly debug, including breakpoints, steps, conditions, locals, inspection, and more!

Knowing how to debug is a critical aspect of every application development life cycle. Debugging allows developers to not only recognize that an exception has occurred, but to also then systematically traverse through the application's execution until the culprit code can be located and fixed. Whether the resolution requires repairing a minor typo or even rewriting a massive component in the system, the simple act of debugging allows developers to (eventually) solve nearly every issue, given enough time and manpower to do so.

However, as powerful as debugging is, it can also be somewhat overwhelming. With hundreds of active programming languages and frameworks used across dozens of code editors and integrated development environments (IDEs), knowing exactly how to get started when trying to debug your own projects can be somewhat staggering at first.

Therefore, this series of articles attempts to answer the simple question of: How to Debug?. In this first article we'll take a broader approach in answering that question by providing a handful of fundamental tips and best practices that apply across numerous languages and development environments. We'll then dig a bit deeper for future articles, looking in more detail at specific programming languages and code editing tools to promote additional tips and tricks. Let's get debuggin'!

Work the Steps

The first important distinction to make between the process of debugging and normal application execution is that debugging allows for some form of additional interaction on your part as the developer. When you normally execute your application, it runs on its own and executes all code by its own volition, based on the logic and instructions provided in the codebase, commonly without the need for user interaction.

On the other hand, a crucial component of debugging is that ability for an individual to interact with the application while it is executing. This run-time interaction opens up a number of useful features and functionalities that would normally be unavailable.

The most critical of these features provided by debugging is the ability to control the flow of execution. Just as with a normal application, you can issue some basic commands to the application to get it to start and stop execution, but debugging allows for much more control over the execution through additional commands.

To illustrate, we'll use a very simple snippet of C# code:

DateTime birthday = new DateTime(2000, 1, 1);
int daysOld = (DateTime.Today - birthday).Days;
Logging.Log($"Age is {daysOld} days old.");

As with most applications this code would be executed line by line, in order from top to bottom. Therefore, we'd first create a birthday DateTime, then get the difference in days between DateTime.Today and birthday, then finally output the result as the number of days old:

Age is 6332 days old.

For normal execution, this process occurs extremely quickly and begins when we issue the start command to the application, telling it to begin execution. However, debugging introduces a few additional commands that allow us to halt execution and step through it line by line, if desired. The rough list of debugging execution flow commands can be found below:

  • Start
  • Pause
  • Continue
  • Step Into
  • Step Over
  • Step Out
  • Stop

Start and stop are pretty self-explanatory. However, once we gain the ability to pause midway through execution things really start to get interesting. Consider our simple three-line application above. For now, as soon as we begin execution we'll hit pause. When that occurs, the debugger process that is running in the background halts the current execution at the exact line of code it was at when we issued that pause command. Perhaps we halted execution on the second line of our application: int daysOld = (DateTime.Today - birthday).Days;. This means that our application has already executed the first line, so we have a DateTime value called birthday, but the second and third line have not yet executed (meaning our daysOld variable is 0, the default value of an int).

Now we can start using the step capabilities of our debugger.

Step Into

Most applications consist of many lines of code that often make calls to other functions or methods. A given method might have many lines of code to execute inside it, and it may even call additional methods or functions inside itself. This process can continue ad nauseam, deeper and deeper through many levels of execution. The Step Into command allows us to continue line-by-line execution while traversing into all those sub-function and sub-method calls. In our case, if we are on line 2 of our application and use Step Into, we'd just move ahead to line 3. However, issuing a Step Into on line 3 steps into our Logging.Log method, which is part of a simple helper class to output information during debugging:

public static void Log(string value)
{
#if DEBUG
Debug.WriteLine(value);
#else
Console.WriteLine(value);
#endif
}

After a few more Step Into commands we'd traverse through all the executed lines of Logging.Log, then we'd jump back out to our calling main method before the application shuts down.

Step Over

While Step Into is the most robust member of the step commands by digging into every single procedure call that the code makes and following it down the rabbit hole, the Step Over command does not traverse into method or function calls when it reaches them. Instead, as the name implies, the Step Over command simply steps to the next statement within the current procedure. Thus, when we reach line 3 of our simple application above where there's a call to the Logging.Log method, a Step Over command will tell the application to continue execution (and will therefore run the Logging.Log call as normal), but it will do so behind the scenes, without stepping into that method via the debugger.

Step Out

Finally, Step Out takes this trend one step further (no pun intended). When issuing a Step Out command, all remaining lines of the procedure which contains the current execution point are executed, after which the debugger jumps out to the next statement, following the procedure call that placed execution inside the inner procedure in the first place. In other words, Step Out jumps outside of the current (local) method or function, executing the rest of that local scope code automatically and without halting.

For example, if we're inside the Logging.Log method and we're sitting at the Debug.WriteLine(value) line, when we issue a Step Out command the application will immediately execute all remaining lines inside that calling procedure (Logging.Log), then it will suspend execution and wait at the end of our application, following the third line of code that contained our: Logging.Log($"Age is {daysOld} days old.");.

Break It Down With Breakpoints

Steps are all well and good, but when applications become much larger than a few lines of code, halting execution at the proper spot by timing a pause command becomes nearly impossible. This is where breakpoints come in handy.

A breakpoint is an intentional marker placed at a particular line of code that instructs the debugger to temporarily suspend execution when it reaches that line of code -- in other words, to break at that point. In most code editors breakpoints can be placed by simply clicking on the left-hand gutter next to a line of code. This usually creates a dot or other marker indicating that a breakpoint has been made.

Now, when executing the application in debug mode, execution will halt at the first line of code with an associated breakpoint on it. From there, normal step commands can be used to traverse through the relevant sections of code that need inspection, rather than the whole of the application.

See What Condition My Condition Was In

As useful as breakpoints are, there are many scenarios where we may wish to halt execution at a specific breakpoint only under certain circumstances. For example, if we have a snippet of code that loops through thousands of objects and we need to debug that code for only one specific iteration of that loop, it can be a painstaking process to sit there and step through that looped code thousands of times just to reach the single loop iteration that we care about.

That's why most debuggers allow us to create breakpoint conditions. A condition is usually just one or more boolean(true/false) statements, typically written in the code of the application in question, which is used to determine when a particular breakpoint should trigger and halt execution.

For example, here we have two lines of code that use our PizzaBuilder class to create a tasty little pizza, then we output the resulting instance to the logger:

// Set Medium size, add Sauce, add Provolone cheese, add Pepperoni, add Olives, then build.
var pizza = new PizzaBuilder(Size.Medium)
.AddSauce()
.AddCheese(Cheese.Provolone)
.AddPepperoni()
.AddOlives()
.Build();
Logging.Log(pizza);

Perhaps we're creating a bunch of pizzas and we have a breakpoint on the Logging.Log(pizza) line, but we only want that breakpoint to trigger (and suspend execution) if our pizza has provolone cheese on it. Thus, our breakpoint condition would contain this simple statement: pizza.Cheese == Cheese.Provolone

Now, when this breakpoint is reached during execution, our debugger will execute the statement above and confirm that our pizza does contain provolone cheese. Since this condition is true, our breakpoint triggers and the debugger halts at the Logging.Log(pizza) line, awaiting further commands from us.

Get to Know the Locals

Stepping through our code line by line, or skipping around as desired, is a great feature, but it's not very useful if we cannot examine and evaluate how each line of executed code is altering values and states of objects. This is where locals come into the debugging process.

For most debuggers, a local is simply a representation of a local variable (one that is defined within the local scope, such as the currently executing method or function). Thus, the collection of locals is a full, automatically populated list of all local variables within the current execution scope, as determined by whatever line of code execution is suspended at.

For example, pausing execution at the end of our three-line age output method shows us the following list of locals:

Name Value Type
args {string[0]} string[]
birthday {1/1/2000 12:00:00 AM} System.DateTime
daysOld 6332 int

Inline Run-Time Inspection

While automatically generated locals are certainly useful to check on variable values and the like, most debugging tools provide inline inspection, which allows us to examine objects, variables, statements, and so forth directly inline during run-time. For example, let's suspend execution at this line: Logging.Log($"Age is {daysOld} days old.");. The debugger I'm using allows me to simply mouse over the daysOld variable in that line and a tooltip immediately pops out showing me what the current value of daysOld is at this point in execution (6332).

Night Gathers and Now My Watch Begins

In many cases while debugging, it may be necessary to evaluate a variable, object, or line of code that isn't actually being executed at the time (either it wasn't written at all in the code base, or maybe it's being executed elsewhere but we need to check it right now). This is where the wonderful watch functionality comes to the rescue.

Most debuggers allow us to create watches, which are local code statements that are executed/evaluated automatically during the debugging process, every time execution steps to a new line of code.

For example, let's remind ourselves of our simple three-line program:

DateTime birthday = new DateTime(2000, 1, 1); // Break here.
int daysOld = (DateTime.Today - birthday).Days;
Logging.Log($"Age is {daysOld} days old.");

Let's start by suspending execution at the first line before anything has executed. Then, let's add a few items to our watch list: birthday, DateTime.Today, (DateTime.Today - birthday).Days, and daysOld.

Name Value Type
birthday {1/1/0001 12:00:00 AM} System.DateTime
DateTime.Today {5/3/2017 12:00:00 AM} System.DateTime
(DateTime.Today - birthday).Days 736451 int
daysOld 0 int

A few important things to note here. While not the case with all languages, for this example using C# and .NET, the common language runtime is smart enough to evaluate the type of objects by looking ahead at our code, even when it hasn't actually initialized those values yet. Thus, even though we haven't initialized our birthday variable yet, we already declared via the compiler, which automatically gives it a default value. In this case, since birthday is a System.DateTime, its default value is Midnight, January 1st of the year 1. A similar process of assigning default values occurs for our other variables.

However, as you may have noticed we can add not only variables but full, executable code statements to our watch list.(DateTime.Today - birthday).Days is just such a statement. It is executed and evaluated every time we break at a new line of code (that is, every time we execute a code statement). Therefore, while the value is currently showing as some 736,451 days (since it's counting from year 1), once we Step Over to the second line of code, all our watches update automatically:

Name Value Type
birthday {1/1/2000 12:00:00 AM} System.DateTime
DateTime.Today {5/3/2017 12:00:00 AM} System.DateTime
(DateTime.Today - birthday).Days 6332 int
daysOld 0 int

Now that the first line has executed, birthday was changed to our intended date of January 1st, 2000 and, therefore, our third watch statement also updated to a new value of 6332.

While debugging practices often include many more additional features, hopefully this brief introduction is a good starting point for your dive into debugging on your own projects. In future articles in this series we'll cover more topics and in greater detail, including specific languages and editors, so stay tuned for that!

Written By: Frances Banks