Debugging in PyCharm and pdb

Debugging a program in PyCharm

A debugger lets you pause execution and inspect variable values before the next statement. First reproduce the error on a small data set. Then set a breakpoint on the line before the suspicious calculation by clicking the gutter to the left of the line number, and start Debug (Shift+F9). The highlighted line at a pause has not yet executed.

Main actions in the default Windows keymap: Step Over (F8) executes a line without entering the called function; Step Into (F7) enters its body; Step Out (Shift+F8) continues until the current call returns; Resume (F9) continues to the next stop. Check names and shortcuts in your PyCharm keymap's menu: https://www.jetbrains.com/help/pycharm/debugging-code.html.

Example 4. A mean grade with a logic error

The following code has an intentionally incorrect denominator. Save the file as average.py. The scores list contains four numbers; sum calculates their sum, and len their count. The program prints a number rather than a traceback, so merely checking that it runs is insufficient.

def average(scores: list[int]) -> float:
    if not scores:
        raise ValueError("score list is empty")
    total = 0
    for score in scores:
        total += score
    return total / (len(scores) - 1)  # Intentional error.


scores = [60, 70, 80, 90]
print(f"Mean grade: {average(scores):.2f}")

Mean grade: 100.00

The expected result is (60 + 70 + 80 + 90) / 4 = 75. Set a breakpoint on return. The variables should show total = 300 and len(scores) = 4. Thus, accumulation is correct; the problem is the final division. Replacing the denominator with len(scores) gives Mean grade: 75.00. For one grade, the corrected function returns the grade itself, while the old one raises ZeroDivisionError.

Screenshot

average.py: total = 300; scores contains four numbers.

Figure 4.5. Pausing before calculating the mean grade.

In the Frames panel, select a call to view its local variables (Variables or Threads & Variables, depending on the version). Watches retain expressions for repeated inspection, such as len(scores). Evaluate Expression (Alt+F8) evaluates an entered expression in the current frame. Enter sum(scores) / len(scores) and compare it with the program's result. Do not call functions that modify data in this window: the observation itself may affect the error.

Screenshot

Evaluate Expression: sum(scores) / len(scores) → 75.0.

Figure 4.6. Independently checking a formula in the current frame.

A long loop calls for a conditional breakpoint: use its context menu to set a Condition, such as score == 80. For a loop with index i, the condition might be i == 7; it must use a name that already exists in that frame. If you only need to monitor values, disable Suspend and enable Log evaluated expression. The program will then continue while leaving a log entry.

Screenshot

Condition: score == 80; Suspend enabled.

Figure 4.7. A conditional stop at the desired loop value.

An exception breakpoint locates where an exception is raised even if it is caught later. Open Run → View Breakpoints (Ctrl+Shift+F8), add a Python Exception Breakpoint for ZeroDivisionError, and test the old function with one grade. Options for stopping when raised or when unhandled depend on the selected debugger; follow the actual labels in your version rather than the icon color.

Screenshot

View Breakpoints; add a Python Exception Breakpoint.

Figure 4.8. Stopping on ZeroDivisionError before error handling.

The pdb debugger and a method for finding errors

The built-in pdb module works in the terminal. For the same file, run python -m pdb average.py. The command l shows text near the current line, n executes the next line, s enters a function, c continues, p expression prints a value, w shows the stack, and q ends debugging. b 7 sets a breakpoint at line 7 of the example file. Reference: https://docs.python.org/3.14/library/pdb.html.

After starting, enter b 7, then c. Execution stops before the incorrect return; p total shows 300, p len(scores) shows 4, and p total / len(scores) shows 75.0. These are the same observations as in PyCharm. Adjust line numbers to match your file if you added comments or blank lines. A breakpoint() call directly in the program also opens pdb under default settings; remove it before submission.

Screenshot

b 7; c; p total; p len(scores); p total / len(scores); q.

Figure 4.9. Inspecting variables with pdb.

Finding a logic error consists of four steps. First, reproduce it and record the expected result. Then locate the first incorrect state by comparing intermediate values. Form a hypothesis, such as “the denominator is one too small”. Finally, test it with the smallest change and repeat reference examples. Changing many formulas at once does not show which one fixed the failure.

For the mean grade, test four grades, one grade, all equal grades, and an empty list. The first three cases should give correct means; the last should raise an explicit ValueError. After fixing the bug, keep these data in the test table: they document the discovered cause.

Common mistakes

• A bare except: catches even KeyboardInterrupt and SystemExit. Identify the expected type and preserve the ability to stop the program.
• A large try hides the failure source. Protect the smallest operation for which you have a specific recovery strategy.
• A general handler before a specific one makes the latter unreachable. Order them from specific to base type.
• except: pass turns failure into apparent success. Explain the failure or raise the exception again.
• A return in finally may destroy a result or hide an exception. Use this block only for final cleanup actions.
• Input checks using assert disappear with -O. Check data limits explicitly.
• Debugging one successful example does not test the algorithm. Add boundaries, invalid formats, failures, and retries after failures.