So far, when something has gone wrong in a script, MATLAB has stopped and printed an error. That default behavior is often exactly what you want while developing a script, but sometimes you want more control: a script that checks its own inputs and fails with a clear message, or one that keeps going after something unexpected happens instead of crashing outright. This page covers throwing and catching errors, issuing warnings, checking for NaN and Inf, and configuring MATLAB to pause automatically when any of these happen.

The MATLAB Help Center has a guide on Errors and Warnings.

Throwing Errors

The error function stops execution immediately and displays a message. Its syntax works like fprintf, accepting a format string with values substituted in:

error('Altitude cannot be negative, got %d ft.', altitude)

If this line runs, the script stops right there. Nothing after it executes, unless the error is caught (see below). Errors can also be given an identifier, a short component:mnemonic label that comes before the message, which makes a specific error easier to catch or look up later:

error('InvalidInput:altitude', 'Altitude cannot be negative, got %d ft.', altitude)

The MATLAB Help Center has documentation on error.

Catching Errors

A try-catch block lets a script recover from an error instead of stopping:

try
    statements
catch ME
    recovery statements
end

MATLAB runs statements. If one of them throws an error, execution jumps immediately to catch, and ME is set to an object describing what happened. ME.message is the error message, and ME.identifier is its identifier (empty if the error didn’t set one). If nothing in statements throws an error, the catch block is skipped entirely.

Example: Guarding Against a Bad Mass Reading

Question

A script computes a rocket’s thrust-to-weight ratio from its thrust and mass. If a sensor glitch reports a mass of zero, dividing by its weight (which would also be zero) produces Inf instead of a useful number, silently propagating a bad value through the rest of the script.

Write a MATLAB script that checks for a non-positive mass before computing anything, and reports a clear error message instead of letting Inf slip through.

Solution

thrust = 5000; % N
mass = 0; % kg, invalid sensor reading
g = 9.81; % m/s^2

try
    if mass <= 0
        error('InvalidInput:mass', 'Mass must be positive, got %d kg.', mass)
    end
    weight = mass * g;
    TWR = thrust / weight;
    disp(TWR)
catch ME
    disp(ME.identifier)
    disp(ME.message)
end


InvalidInput:mass
Mass must be positive, got 0 kg.

Warnings

The warning function prints a message but, unlike error, does not stop execution. Use it when something looks wrong but the script can reasonably continue, often by falling back to a default value:

warning('off', 'backtrace');

mass = 0; % kg, sensor returned zero unexpectedly
if mass == 0
    warning('Mass reading is zero, using default value instead.')
    mass = 1;
end

disp(mass)


[Warning: Mass reading is zero, using default value instead.] 
     1

By default, a warning also prints which line triggered it. warning('off', 'backtrace') turns that off, leaving just the message. Warnings can be disabled entirely with warning('off', 'all') and re-enabled with warning('on', 'all'), though it’s usually better to fix what’s causing the warning than to silence it.

The MATLAB Help Center has documentation on warning.

NaN and Inf

Some operations don’t have a defined numeric result. Dividing zero by zero produces NaN (Not a Number), and dividing a nonzero number by zero produces Inf or -Inf. Neither of these throws an error or a warning on its own, which means they can silently work their way through the rest of a calculation unless you check for them. The isnan and isinf functions test for exactly this:

thrust = 0; % N
weight = 0; % N
TWR = thrust / weight;

disp(TWR)
disp(isnan(TWR))

climb_rate = 500 / 0; % ft/min, divided by zero time

disp(climb_rate)
disp(isinf(climb_rate))


   NaN

   1

   Inf

   1

The MATLAB Help Center has documentation on isnan and isinf.

Breakpoints and Pausing on Errors

A breakpoint pauses execution at a specific line, so you can inspect variables in the Workspace before continuing. To set one, click the gray area to the left of the line number in the Editor; a red dot appears, and running the script will pause right before that line runs. Click the red dot again to clear it.

Instead of picking a specific line, you can also tell MATLAB to pause automatically whenever an error, a warning, or a NaN/Inf occurs anywhere in your code. In the Editor, open the Debugger panel (Editor tab → Analyze section → Debugger), and under Breakpoints choose Pause on Errors, Pause on Warnings, or Pause on NaN or Inf. The same options are available from the Command Window with dbstop if error, dbstop if warning, and dbstop if naninf, respectively. Use dbclear all to turn all of these back off.

Pausing on errors is especially useful for a bug that’s otherwise hard to reproduce: rather than guessing where things went wrong from the error message alone, MATLAB stops mid-script with every variable exactly as it was at the moment of the error.

The MATLAB Help Center has documentation on dbstop.

Reading Questions

  1. What is the difference in behavior between error and warning?
  2. What does ME.message contain inside a catch block? What about ME.identifier?
  3. If nothing inside a try block throws an error, does the catch block run?
  4. What value does 0/0 produce in MATLAB? What about 1/0?
  5. Which functions check whether a value is NaN or Inf?
  6. How do you set a breakpoint at a specific line in the MATLAB Editor?
  7. How would you configure MATLAB to automatically pause whenever a warning occurs, without setting a breakpoint at a specific line?