by author Craig Buckler
Your apps become more robust as you gain programming experience. Improvements to coding techniques play a large part but you'll also learn to consider the edge cases. If something can go wrong, it will go wrong: typically when the first user accesses your new system.
Some errors are avoidable:
A JavaScript linter or good editor can catch syntax errors such as misspelled statements and missing brackets.
You can use validation to catch errors in data from users or other systems. Never make presumptions about a user's ingenuity to cause havoc. You may be expecting a numeric integer when asking for someone's age but could they leave the box blank, enter a negative value, use a fractional value, or even fully type "twenty-one" in their native language.
Remember that server-side validation is essential. Browser-based client-side validation can improve the UI but the user could be using an application where JavaScript is disabled, fails to load, or fails to execute. It may not be a browser that's calling your API!
Other runtime errors are impossible to avoid:
- the network can fail
- a busy server or application could take too long to respond
- a script or other asset could time out
- the application can fail
- a disk or database can fail
- a server OS can fail
- a host's infrastructure can fail
These may be temporary. You cannot necessarily code your way out of these issues but you can anticipate problems, take appropriate actions, and make your application more resilient.
Showing an Error is the Last Resort
We've all encountered errors in applications. Some are useful:
"The file already exists. Would you like to overwrite it?"
Others are less so:
"ERROR 5969"
Showing an error to the user should be the last resort after exhausting all other options.
You may be able to ignore some less-critical errors such as an image failing to load. In other cases, remedial or recovery actions may be possible. For example, if you're unable to save data to a server because of a network failure, you could temporarily store it in IndexedDB or localStorage and try a few minutes later. It should only be necessary to show a warning when repeated saves fail and the user is at risk of losing data. Even then: ensure the user can take appropriate actions. They may be able to reconnect to the network but that won't help if your server is down.
Error Handling in JavaScript
Error handling is simple in JavaScript but it's often shrouded in mystery and can become complicated when considering asynchronous code.
An "error" is a object you can throw to raise an exception -- which could halt the program if it's not captured and dealt with appropriately. You can create an Error
object by passing an optional message to the constructor:
const e = new Error('An error has occurred');
You can also use Error
like a function without new
-- it still returns an Error
object identical to that above:
const e = Error('An error has occurred');
You can pass a filename and a line number as the second and third parameters:
const e = new Error('An error has occurred', 'script.js', 99);
This is rarely necessary since these default to the line where you created the Error
object in the current file.
Once created, an Error
object has the following properties which you can read and write:
.name
: the name of theError
type ("Error"
in this case).message
: the error message
The following read/write properties are also supported in Firefox:
.fileName
: the file where the error occurred.lineNumber
: the line number where the error occurred.columnNumber
: the column number on the line where the error occurred.stack
: a stack trace -- the list of functions calls made to reach the error.
Error Types
As well as a generic Error
, JavaScript supports specific error types:
EvalError
: caused by aneval()
RangeError
: a value outside a valid rangeReferenceError
: occurs when de-referencing an invalid referenceSyntaxError
: invalid syntaxTypeError
: a value is not a valid typeURIError
: invalid parameters passed toencodeURI()
ordecodeURI()
AggregateError
: several errors wrapped in a single error typically raised when calling an operation such asPromise.all()
.
The JavaScript interpreter will raise appropriate errors as necessary. In most cases, you will use Error
or perhaps TypeError
in your own code.
Throwing an Exception
Creating an Error
object does nothing on its own. You must throw
an Error
to raise an exception:
throw new Error('An error has occurred');
This sum()
function throws a TypeError
when either argument is not a number -- the return
is never executed:
function sum(a, b) {
if (isNaN(a) || isNaN(b)) {
throw new TypeError('Value is not a number.');
}
return a + b;
}
It's practical to throw
an Error
object but you can use any value or object:
throw 'Error string';
throw 42;
throw true;
throw { message: 'An error', name: 'CustomError' };
When you throw
an exception it bubbles up the call stack -- unless it's caught. Uncaught exceptions eventually reach the top of the stack where the program will halt and show an error in the DevTools console, e.g.
Uncaught TypeError: Value is not a number.
sum https://mysite.com/js/index.js:3
Catching Exceptions
You can catch exceptions in a try ... catch
block:
try {
console.log( sum(1, 'a') );
}
catch (err) {
console.error( err.message );
}
This executes the code in the try {}
block but, when an exception occurs, the catch {}
block receives the object returned by the throw
.
A catch
block can analyse the error and react accordingly, e.g.
try {
console.log( sum(1, 'a') );
}
catch (err) {
if (err instanceof TypeError) {
console.error( 'wrong type' );
}
else {
console.error( err.message );
}
}
You can define an optional finally {}
block if you require code to run whether the try
or catch
code executes. This can be useful when cleaning up, e.g. to close a database connection in Node.js or Deno:
try {
console.log( sum(1, 'a') );
}
catch (err) {
console.error( err.message );
}
finally {
// this always runs
console.log( 'program has ended' );
}
A try
block requires either a catch
block, a finally
block, or both.
Note that when a finally
block contains a return
, that value becomes the return value for the whole try ... catch ... finally
regardless of any return
statements in the try
and catch
blocks.
Open Source Session Replay
Debugging a web application in production may be challenging and time-consuming. OpenReplay is an Open-source alternative to FullStory, LogRocket and Hotjar. It allows you to monitor and replay everything your users do and shows how your app behaves for every issue. It’s like having your browser’s inspector open while looking over your user’s shoulder. OpenReplay is the only open-source alternative currently available.
Happy debugging, for modern frontend teams - Start monitoring your web app for free.
Nested try ... catch
Blocks and Rethrowing Errors
An exception bubbles up the stack and is caught once only by the nearest catch
block, e.g.
try {
try {
console.log( sum(1, 'a') );
}
catch (err) {
console.error('This error will trigger', err.message);
}
}
catch (err) {
console.error('This error will not trigger', err.message);
}
Any catch
block can throw
a new exception which is caught by the outer catch
. You can pass the first Error
object to a new Error
in the cause
property of an object passed to the constructor. This makes it possible to raise and examine a chain of errors.
Both catch
blocks execute in this example because the first error throws a second:
try {
try {
console.log( sum(1, 'a') );
}
catch (err) {
console.error('First error caught', err.message);
throw new Error('Second error', { cause: err });
}
}
catch (err) {
console.error('Second error caught', err.message);
console.error('Error cause:', err.cause.message);
}
Throwing Exceptions in Asynchronous Functions
You cannot catch
an exception thrown by an asynchronous function because it's raised after the try ... catch
block has completed execution. This will fail:
function wait(delay = 1000) {
setTimeout(() => {
throw new Error('I am never caught!');
}, delay);
}
try {
wait();
}
catch(err) {
// this will never run
console.error('caught!', err.message);
}
After one second has elapsed, the console displays:
Uncaught Error: I am never caught!
wait http://localhost:8888/:14
If you're using a callback, the convention presumed in frameworks and runtimes such as Node.js is to return an error as the first parameter to that function. This does not throw
an exception although you can manually do that when necessary:
function wait(delay = 1000, callback) {
setTimeout(() => {
callback('I am caught!');
}, delay);
}
wait(1000, (err) => {
if (err) {
throw new Error(err);
}
});
In modern ES6 it's often better to return a Promise when defining asynchronous functions. When an error occurs, the Promise's reject
method can return a new Error
object (although any value or object is possible):
function wait(delay = 1000) {
return new Promise((resolve, reject) => {
if (isNaN(delay) || delay < 0) {
reject(new TypeError('Invalid delay'));
}
else {
setTimeout(() => {
resolve(`waited ${ delay } ms`);
}, delay);
}
})
}
The Promise.catch()
method executes when passing an invalid delay
parameter so it can react to the returned Error
object:
// this fails - the catch block is run
wait('x')
.then( res => console.log( res ))
.catch( err => console.error( err.message ))
.finally(() => console.log('done'));
Any function which returns a Promise
can be called by an async
function using an await
statement. You can contain this in a try ... catch
block which runs identically to the .then
/.catch
Promise example above but can be a little easier to read:
// Immediately-Invoked (Asynchronous) Function Expression
(async () => {
try {
console.log( await wait('x') );
}
catch (err) {
console.error( err.message );
}
finally {
console.log('done');
}
})();
Errors are Inevitable
Creating error objects and raising exceptions is easy in JavaScript. Reacting appropriately and building resilient applications is somewhat more difficult! The best advice: expect the unexpected and deal with errors as soon as possible.