Develop Reference

How to detect that the parseFloat() stopped on "other" character? [duplicate]

I'm hoping there's something in the same conceptual space as the old VB6 IsNumeric() function?

2nd October 2020: note that many bare-bones approaches are fraught with subtle bugs (eg. whitespace, implicit partial parsing, radix, coercion of arrays etc.) that many of the answers here fail to take into account. The following implementation might work for you, but note that it does not cater for number separators other than the decimal point ".":
function isNumeric(str) {
if (typeof str != "string") return false // we only process strings!
return !isNaN(str) && // use type coercion to parse the _entirety_ of the string (`parseFloat` alone does not do this)...
!isNaN(parseFloat(str)) // ...and ensure strings of whitespace fail
}
To check if a variable (including a string) is a number, check if it is not a number:
This works regardless of whether the variable content is a string or number.
isNaN(num) // returns true if the variable does NOT contain a valid number
Examples
isNaN(123) // false
isNaN('123') // false
isNaN('1e10000') // false (This translates to Infinity, which is a number)
isNaN('foo') // true
isNaN('10px') // true
isNaN('') // false
isNaN(' ') // false
isNaN(false) // false
Of course, you can negate this if you need to. For example, to implement the IsNumeric example you gave:
function isNumeric(num){
return !isNaN(num)
}
To convert a string containing a number into a number:
Only works if the string only contains numeric characters, else it returns NaN.
+num // returns the numeric value of the string, or NaN
// if the string isn't purely numeric characters
Examples
+'12' // 12
+'12.' // 12
+'12..' // NaN
+'.12' // 0.12
+'..12' // NaN
+'foo' // NaN
+'12px' // NaN
To convert a string loosely to a number
Useful for converting '12px' to 12, for example:
parseInt(num) // extracts a numeric value from the
// start of the string, or NaN.
Examples
parseInt('12') // 12
parseInt('aaa') // NaN
parseInt('12px') // 12
parseInt('foo2') // NaN These last three may
parseInt('12a5') // 12 be different from what
parseInt('0x10') // 16 you expected to see.
Floats
Bear in mind that, unlike +num, parseInt (as the name suggests) will convert a float into an integer by chopping off everything following the decimal point (if you want to use parseInt() because of this behaviour, you're probably better off using another method instead):
+'12.345' // 12.345
parseInt(12.345) // 12
parseInt('12.345') // 12
Empty strings
Empty strings may be a little counter-intuitive. +num converts empty strings or strings with spaces to zero, and isNaN() assumes the same:
+'' // 0
+' ' // 0
isNaN('') // false
isNaN(' ') // false
But parseInt() does not agree:
parseInt('') // NaN
parseInt(' ') // NaN

If you're just trying to check if a string is a whole number (no decimal places), regex is a good way to go. Other methods such as isNaN are too complicated for something so simple.
function isNumeric(value) {
return /^-?\d+$/.test(value);
}
console.log(isNumeric('abcd')); // false
console.log(isNumeric('123a')); // false
console.log(isNumeric('1')); // true
console.log(isNumeric('1234567890')); // true
console.log(isNumeric('-23')); // true
console.log(isNumeric(1234)); // true
console.log(isNumeric(1234n)); // true
console.log(isNumeric('123.4')); // false
console.log(isNumeric('')); // false
console.log(isNumeric(undefined)); // false
console.log(isNumeric(null)); // false
To only allow positive whole numbers use this:
function isNumeric(value) {
return /^\d+$/.test(value);
}
console.log(isNumeric('123')); // true
console.log(isNumeric('-23')); // false

The accepted answer for this question has quite a few flaws (as highlighted by couple of other users). This is one of the easiest & proven way to approach it in javascript:
function isNumeric(n) {
return !isNaN(parseFloat(n)) && isFinite(n);
}
Below are some good test cases:
console.log(isNumeric(12345678912345678912)); // true
console.log(isNumeric('2 ')); // true
console.log(isNumeric('-32.2 ')); // true
console.log(isNumeric(-32.2)); // true
console.log(isNumeric(undefined)); // false
// the accepted answer fails at these tests:
console.log(isNumeric('')); // false
console.log(isNumeric(null)); // false
console.log(isNumeric([])); // false

And you could go the RegExp-way:
var num = "987238";
if(num.match(/^-?\d+$/)){
//valid integer (positive or negative)
}else if(num.match(/^\d+\.\d+$/)){
//valid float
}else{
//not valid number
}

If you really want to make sure that a string contains only a number, any number (integer or floating point), and exactly a number, you cannot use parseInt()/ parseFloat(), Number(), or !isNaN() by themselves. Note that !isNaN() is actually returning true when Number() would return a number, and false when it would return NaN, so I will exclude it from the rest of the discussion.
The problem with parseFloat() is that it will return a number if the string contains any number, even if the string doesn't contain only and exactly a number:
parseFloat("2016-12-31") // returns 2016
parseFloat("1-1") // return 1
parseFloat("1.2.3") // returns 1.2
The problem with Number() is that it will return a number in cases where the passed value is not a number at all!
Number("") // returns 0
Number(" ") // returns 0
Number(" \u00A0 \t\n\r") // returns 0
The problem with rolling your own regex is that unless you create the exact regex for matching a floating point number as Javascript recognizes it you are going to miss cases or recognize cases where you shouldn't. And even if you can roll your own regex, why? There are simpler built-in ways to do it.
However, it turns out that Number() (and isNaN()) does the right thing for every case where parseFloat() returns a number when it shouldn't, and vice versa. So to find out if a string is really exactly and only a number, call both functions and see if they both return true:
function isNumber(str) {
if (typeof str != "string") return false // we only process strings!
// could also coerce to string: str = ""+str
return !isNaN(str) && !isNaN(parseFloat(str))
}

Try the isNan function:
The isNaN() function determines whether a value is an illegal number (Not-a-Number).
This function returns true if the value equates to NaN. Otherwise it returns false.
This function is different from the Number specific Number.isNaN() method.
The global isNaN() function, converts the tested value to a Number, then tests it.
Number.isNan() does not convert the values to a Number, and will not return true for any value that is not of the type Number...

2019: Including ES3, ES6 and TypeScript Examples
Maybe this has been rehashed too many times, however I fought with this one today too and wanted to post my answer, as I didn't see any other answer that does it as simply or thoroughly:
ES3
var isNumeric = function(num){
return (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num);
}
ES6
const isNumeric = (num) => (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num);
Typescript
const isNumeric = (num: any) => (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num as number);
This seems quite simple and covers all the bases I saw on the many other posts and thought up myself:
// Positive Cases
console.log(0, isNumeric(0) === true);
console.log(1, isNumeric(1) === true);
console.log(1234567890, isNumeric(1234567890) === true);
console.log('1234567890', isNumeric('1234567890') === true);
console.log('0', isNumeric('0') === true);
console.log('1', isNumeric('1') === true);
console.log('1.1', isNumeric('1.1') === true);
console.log('-1', isNumeric('-1') === true);
console.log('-1.2354', isNumeric('-1.2354') === true);
console.log('-1234567890', isNumeric('-1234567890') === true);
console.log(-1, isNumeric(-1) === true);
console.log(-32.1, isNumeric(-32.1) === true);
console.log('0x1', isNumeric('0x1') === true); // Valid number in hex
// Negative Cases
console.log(true, isNumeric(true) === false);
console.log(false, isNumeric(false) === false);
console.log('1..1', isNumeric('1..1') === false);
console.log('1,1', isNumeric('1,1') === false);
console.log('-32.1.12', isNumeric('-32.1.12') === false);
console.log('[blank]', isNumeric('') === false);
console.log('[spaces]', isNumeric(' ') === false);
console.log('null', isNumeric(null) === false);
console.log('undefined', isNumeric(undefined) === false);
console.log([], isNumeric([]) === false);
console.log('NaN', isNumeric(NaN) === false);
You can also try your own isNumeric function and just past in these use cases and scan for "true" for all of them.
Or, to see the values that each return:

Old question, but there are several points missing in the given answers.
Scientific notation.
!isNaN('1e+30') is true, however in most of the cases when people ask for numbers, they do not want to match things like 1e+30.
Large floating numbers may behave weird
Observe (using Node.js):
> var s = Array(16 + 1).join('9')
undefined
> s.length
16
> s
'9999999999999999'
> !isNaN(s)
true
> Number(s)
10000000000000000
> String(Number(s)) === s
false
>
On the other hand:
> var s = Array(16 + 1).join('1')
undefined
> String(Number(s)) === s
true
> var s = Array(15 + 1).join('9')
undefined
> String(Number(s)) === s
true
>
So, if one expects String(Number(s)) === s, then better limit your strings to 15 digits at most (after omitting leading zeros).
Infinity
> typeof Infinity
'number'
> !isNaN('Infinity')
true
> isFinite('Infinity')
false
>
Given all that, checking that the given string is a number satisfying all of the following:
non scientific notation
predictable conversion to Number and back to String
finite
is not such an easy task. Here is a simple version:
function isNonScientificNumberString(o) {
if (!o || typeof o !== 'string') {
// Should not be given anything but strings.
return false;
}
return o.length <= 15 && o.indexOf('e+') < 0 && o.indexOf('E+') < 0 && !isNaN(o) && isFinite(o);
}
However, even this one is far from complete. Leading zeros are not handled here, but they do screw the length test.

I have tested and Michael's solution is best. Vote for his answer above (search this page for "If you really want to make sure that a string" to find it). In essence, his answer is this:
function isNumeric(num){
num = "" + num; //coerce num to be a string
return !isNaN(num) && !isNaN(parseFloat(num));
}
It works for every test case, which I documented here:
https://jsfiddle.net/wggehvp9/5/
Many of the other solutions fail for these edge cases:
' ', null, "", true, and [].
In theory, you could use them, with proper error handling, for example:
return !isNaN(num);
or
return (+num === +num);
with special handling for
/\s/, null, "", true, false, [] (and others?)

TL;DR
It depends largely on what you want to parse as a number.
Comparison Between Built-in Functions
As none of the existing sources satisfied my soul, I tried to figure out what actually was happening with these functions.
Three immediate answers to this question felt like:
!isNaN(input) (which gives the same output as +input === +input)
!isNaN(parseFloat(input))
isFinite(input)
But are any of them correct in every scenario?
I tested these functions in several cases, and generated output as markdown. This is what it looks like:
input
!isNaN(input) or +input===+input
!isNaN(parseFloat(input))
isFinite(input)
Comment
123
✔️
✔️
✔️
-
'123'
✔️
✔️
✔️
-
12.3
✔️
✔️
✔️
-
'12.3'
✔️
✔️
✔️
-
'   12.3   '
✔️
✔️
✔️
Empty whitespace trimmed, as expected.
1_000_000
✔️
✔️
✔️
Numeric separator understood, also expected.
'1_000_000'
❌
✔️
❌
Surprise! JS just won't parse numeric separator inside a string. For details, check this issue. (Why then parsing as float worked though? Well, it didn't. 😉)
'0b11111111'
✔️
✔️
✔️
Binary form understood, as it should've.
'0o377'
✔️
✔️
✔️
Octal form understood too.
'0xFF'
✔️
✔️
✔️
Of course hex is understood. Did anybody think otherwise? 😒
''
✔️
❌
✔️
Should empty string be a number?
'    '
✔️
❌
✔️
Should a whitespace-only string be a number?
'abc'
❌
❌
❌
Everybody agrees, not a number.
'12.34Ab!##$'
❌
✔️
❌
Ah! Now it's quite understandable what parseFloat() does. Not impressive to me, but may come handy in certain cases.
'10e100'
✔️
✔️
✔️
10100 is a number indeed. But caution! It's way more larger than the maximum safe integer value 253 (about 9×1015). Read this for details.
'10e1000'
✔️
✔️
❌
Say with me, help! Though not as crazy as it may seem. In JavaScript, a value larger than ~10308 is rounded to infinity, that's why. Look here for details. And yes, isNaN() considers infinity as a number, and parseFloat() parses infinity as infinity.
null
✔️
❌
✔️
Now this is awkward. In JS, when a conversion is needed, null becomes zero, and we get a finite number. Then why parseFloat(null) should return a NaN here? Someone please explain this design concept to me.
undefined
❌
❌
❌
As expected.
Infinity
✔️
✔️
❌
As explained before, isNaN() considers infinity as a number, and parseFloat() parses infinity as infinity.
So...which of them is "correct"?
Should be clear by now, it depends largely on what we need. For example, we may want to consider a null input as 0. In that case isFinite() will work fine.
Again, perhaps we will take a little help from isNaN() when 1010000000000 is needed to be considered a valid number (although the question remains—why would it be, and how would we handle that)!
Of course, we can manually exclude any of the scenarios.
Like in my case, I needed exactly the outputs of isFinite(), except for the null case, the empty string case, and the whitespace-only string case. Also I had no headache about really huge numbers. So my code looked like this:
/**
* My necessity was met by the following code.
*/
if (input === null) {
// Null input
} else if (input.trim() === '') {
// Empty or whitespace-only string
} else if (isFinite(input)) {
// Input is a number
} else {
// Not a number
}
And also, this was my JavaScript to generate the table:
/**
* Note: JavaScript does not print numeric separator inside a number.
* In that single case, the markdown output was manually corrected.
* Also, the comments were manually added later, of course.
*/
let inputs = [
123, '123', 12.3, '12.3', ' 12.3 ',
1_000_000, '1_000_000',
'0b11111111', '0o377', '0xFF',
'', ' ',
'abc', '12.34Ab!##$',
'10e100', '10e1000',
null, undefined, Infinity];
let markdownOutput = `| \`input\` | \`!isNaN(input)\` or <br>\`+input === +input\` | \`!isNaN(parseFloat(input))\` | \`isFinite(input)\` | Comment |
| :---: | :---: | :---: | :---: | :--- |\n`;
for (let input of inputs) {
let outputs = [];
outputs.push(!isNaN(input));
outputs.push(!isNaN(parseFloat(input)));
outputs.push(isFinite(input));
if (typeof input === 'string') {
// Output with quotations
console.log(`'${input}'`);
markdownOutput += `| '${input}'`;
} else {
// Output without quotes
console.log(input);
markdownOutput += `| ${input}`;
}
for (let output of outputs) {
console.log('\t' + output);
if (output === true) {
markdownOutput += ` | <div style="color:limegreen">true</div>`;
// markdownOutput += ` | ✔️`; // for stackoverflow
} else {
markdownOutput += ` | <div style="color:orangered">false</div>`;
// markdownOutput += ` | ❌`; // for stackoverflow
}
}
markdownOutput += ` ||\n`;
}
// Replace two or more whitespaces with $nbsp;
markdownOutput = markdownOutput.replaceAll(` `, ` `);
// Print markdown to console
console.log(markdownOutput);

The JavaScript global isFinite() checks if a value is a valid (finite) number.
See MDN for the difference between Number.isFinite() and global isFinite().
let a = isFinite('abc') // false
let b = isFinite('123') // true
let c = isFinite('12a') // false
let d = isFinite(null) // true
console.log(a, b, c, d)

Someone may also benefit from a regex based answer. Here it is:
One liner isInteger:
const isInteger = num => /^-?[0-9]+$/.test(num+'');
One liner isNumeric: Accepts integers and decimals
const isNumeric = num => /^-?[0-9]+(?:\.[0-9]+)?$/.test(num+'');

You can use the result of Number when passing an argument to its constructor.
If the argument (a string) cannot be converted into a number, it returns NaN, so you can determinate if the string provided was a valid number or not.
Notes: Note when passing empty string or '\t\t' and '\n\t' as Number will return 0; Passing true will return 1 and false returns 0.
Number('34.00') // 34
Number('-34') // -34
Number('123e5') // 12300000
Number('123e-5') // 0.00123
Number('999999999999') // 999999999999
Number('9999999999999999') // 10000000000000000 (integer accuracy up to 15 digit)
Number('0xFF') // 255
Number('Infinity') // Infinity
Number('34px') // NaN
Number('xyz') // NaN
Number('true') // NaN
Number('false') // NaN
// cavets
Number(' ') // 0
Number('\t\t') // 0
Number('\n\t') // 0

Maybe there are one or two people coming across this question who need a much stricter check than usual (like I did). In that case, this might be useful:
if(str === String(Number(str))) {
// it's a "perfectly formatted" number
}
Beware! This will reject strings like .1, 40.000, 080, 00.1. It's very picky - the string must match the "most minimal perfect form" of the number for this test to pass.
It uses the String and Number constructor to cast the string to a number and back again and thus checks if the JavaScript engine's "perfect minimal form" (the one it got converted to with the initial Number constructor) matches the original string.

Why is jQuery's implementation not good enough?
function isNumeric(a) {
var b = a && a.toString();
return !$.isArray(a) && b - parseFloat(b) + 1 >= 0;
};
Michael suggested something like this (although I've stolen "user1691651 - John"'s altered version here):
function isNumeric(num){
num = "" + num; //coerce num to be a string
return !isNaN(num) && !isNaN(parseFloat(num));
}
The following is a solution with most likely bad performance, but solid results. It is a contraption made from the jQuery 1.12.4 implementation and Michael's answer, with an extra check for leading/trailing spaces (because Michael's version returns true for numerics with leading/trailing spaces):
function isNumeric(a) {
var str = a + "";
var b = a && a.toString();
return !$.isArray(a) && b - parseFloat(b) + 1 >= 0 &&
!/^\s+|\s+$/g.test(str) &&
!isNaN(str) && !isNaN(parseFloat(str));
};
The latter version has two new variables, though. One could get around one of those, by doing:
function isNumeric(a) {
if ($.isArray(a)) return false;
var b = a && a.toString();
a = a + "";
return b - parseFloat(b) + 1 >= 0 &&
!/^\s+|\s+$/g.test(a) &&
!isNaN(a) && !isNaN(parseFloat(a));
};
I haven't tested any of these very much, by other means than manually testing the few use-cases I'll be hitting with my current predicament, which is all very standard stuff. This is a "standing-on-the-shoulders-of-giants" situation.

2019: Practical and tight numerical validity check
Often, a 'valid number' means a Javascript number excluding NaN and Infinity, ie a 'finite number'.
To check the numerical validity of a value (from an external source for example), you can define in ESlint Airbnb style :
/**
* Returns true if 'candidate' is a finite number or a string referring (not just 'including') a finite number
* To keep in mind:
* Number(true) = 1
* Number('') = 0
* Number(" 10 ") = 10
* !isNaN(true) = true
* parseFloat('10 a') = 10
*
* #param {?} candidate
* #return {boolean}
*/
function isReferringFiniteNumber(candidate) {
if (typeof (candidate) === 'number') return Number.isFinite(candidate);
if (typeof (candidate) === 'string') {
return (candidate.trim() !== '') && Number.isFinite(Number(candidate));
}
return false;
}
and use it this way:
if (isReferringFiniteNumber(theirValue)) {
myCheckedValue = Number(theirValue);
} else {
console.warn('The provided value doesn\'t refer to a finite number');
}

It is not valid for TypeScript as:
declare function isNaN(number: number): boolean;
For TypeScript you can use:
/^\d+$/.test(key)

I like the simplicity of this.
Number.isNaN(Number(value))
The above is regular Javascript, but I'm using this in conjunction with a typescript typeguard for smart type checking. This is very useful for the typescript compiler to give you correct intellisense, and no type errors.
Typescript typeguards
Warning: See Jeremy's comment below. This has some issues with certain values and I don't have time to fix it now, but the idea of using a typescript typeguard is useful so I won't delete this section.
isNotNumber(value: string | number): value is string {
return Number.isNaN(Number(this.smartImageWidth));
}
isNumber(value: string | number): value is number {
return Number.isNaN(Number(this.smartImageWidth)) === false;
}
Let's say you have a property width which is number | string. You may want to do logic based on whether or not it's a string.
var width: number|string;
width = "100vw";
if (isNotNumber(width))
{
// the compiler knows that width here must be a string
if (width.endsWith('vw'))
{
// we have a 'width' such as 100vw
}
}
else
{
// the compiler is smart and knows width here must be number
var doubleWidth = width * 2;
}
The typeguard is smart enough to constrain the type of width within the if statement to be ONLY string. This permits the compiler to allow width.endsWith(...) which it wouldn't allow if the type was string | number.
You can call the typeguard whatever you want isNotNumber, isNumber, isString, isNotString but I think isString is kind of ambiguous and harder to read.

parseInt(), but be aware that this function is a bit different in the sense that it for example returns 100 for parseInt("100px").

When guarding against empty strings and null
// Base cases that are handled properly
Number.isNaN(Number('1')); // => false
Number.isNaN(Number('-1')); // => false
Number.isNaN(Number('1.1')); // => false
Number.isNaN(Number('-1.1')); // => false
Number.isNaN(Number('asdf')); // => true
Number.isNaN(Number(undefined)); // => true
// Special notation cases that are handled properly
Number.isNaN(Number('1e1')); // => false
Number.isNaN(Number('1e-1')); // => false
Number.isNaN(Number('-1e1')); // => false
Number.isNaN(Number('-1e-1')); // => false
Number.isNaN(Number('0b1')); // => false
Number.isNaN(Number('0o1')); // => false
Number.isNaN(Number('0xa')); // => false
// Edge cases that will FAIL if not guarded against
Number.isNaN(Number('')); // => false
Number.isNaN(Number(' ')); // => false
Number.isNaN(Number(null)); // => false
// Edge cases that are debatable
Number.isNaN(Number('-0b1')); // => true
Number.isNaN(Number('-0o1')); // => true
Number.isNaN(Number('-0xa')); // => true
Number.isNaN(Number('Infinity')); // => false
Number.isNaN(Number('INFINITY')); // => true
Number.isNaN(Number('-Infinity')); // => false
Number.isNaN(Number('-INFINITY')); // => true
When NOT guarding against empty strings and null
Using parseInt:
// Base cases that are handled properly
Number.isNaN(parseInt('1')); // => false
Number.isNaN(parseInt('-1')); // => false
Number.isNaN(parseInt('1.1')); // => false
Number.isNaN(parseInt('-1.1')); // => false
Number.isNaN(parseInt('asdf')); // => true
Number.isNaN(parseInt(undefined)); // => true
Number.isNaN(parseInt('')); // => true
Number.isNaN(parseInt(' ')); // => true
Number.isNaN(parseInt(null)); // => true
// Special notation cases that are handled properly
Number.isNaN(parseInt('1e1')); // => false
Number.isNaN(parseInt('1e-1')); // => false
Number.isNaN(parseInt('-1e1')); // => false
Number.isNaN(parseInt('-1e-1')); // => false
Number.isNaN(parseInt('0b1')); // => false
Number.isNaN(parseInt('0o1')); // => false
Number.isNaN(parseInt('0xa')); // => false
// Edge cases that are debatable
Number.isNaN(parseInt('-0b1')); // => false
Number.isNaN(parseInt('-0o1')); // => false
Number.isNaN(parseInt('-0xa')); // => false
Number.isNaN(parseInt('Infinity')); // => true
Number.isNaN(parseInt('INFINITY')); // => true
Number.isNaN(parseInt('-Infinity')); // => true
Number.isNaN(parseInt('-INFINITY')); // => true
Using parseFloat:
// Base cases that are handled properly
Number.isNaN(parseFloat('1')); // => false
Number.isNaN(parseFloat('-1')); // => false
Number.isNaN(parseFloat('1.1')); // => false
Number.isNaN(parseFloat('-1.1')); // => false
Number.isNaN(parseFloat('asdf')); // => true
Number.isNaN(parseFloat(undefined)); // => true
Number.isNaN(parseFloat('')); // => true
Number.isNaN(parseFloat(' ')); // => true
Number.isNaN(parseFloat(null)); // => true
// Special notation cases that are handled properly
Number.isNaN(parseFloat('1e1')); // => false
Number.isNaN(parseFloat('1e-1')); // => false
Number.isNaN(parseFloat('-1e1')); // => false
Number.isNaN(parseFloat('-1e-1')); // => false
Number.isNaN(parseFloat('0b1')); // => false
Number.isNaN(parseFloat('0o1')); // => false
Number.isNaN(parseFloat('0xa')); // => false
// Edge cases that are debatable
Number.isNaN(parseFloat('-0b1')); // => false
Number.isNaN(parseFloat('-0o1')); // => false
Number.isNaN(parseFloat('-0xa')); // => false
Number.isNaN(parseFloat('Infinity')); // => false
Number.isNaN(parseFloat('INFINITY')); // => true
Number.isNaN(parseFloat('-Infinity')); // => false
Number.isNaN(parseFloat('-INFINITY')); // => true
Notes:
Only string, empty, and uninitialized values are considered in keeping with addressing the original question. Additional edge cases exist if arrays and objects are the values being considered.
Characters in binary, octal, hexadecimal, and exponential notation are not case-sensitive (ie: '0xFF', '0XFF', '0xfF' etc. will all yield the same result in the test cases shown above).
Unlike with Infinity (case-sensitive) in some cases, constants from the Number and Math objects passed as test cases in string format to any of the methods above will be determined to not be numbers.
See here for an explanation of how arguments are converted to a Number and why the edge cases for null and empty strings exist.

Quote:
isNaN(num) // returns true if the variable does NOT contain a valid number
is not entirely true if you need to check for leading/trailing spaces - for example when a certain quantity of digits is required, and you need to get, say, '1111' and not ' 111' or '111 ' for perhaps a PIN input.
Better to use:
var num = /^\d+$/.test(num)

This is built on some of the previous answers and comments. The following covers all the edge cases and fairly concise as well:
const isNumRegEx = /^-?(\d*\.)?\d+$/;
function isNumeric(n, allowScientificNotation = false) {
return allowScientificNotation ?
!Number.isNaN(parseFloat(n)) && Number.isFinite(n) :
isNumRegEx.test(n);
}

If anyone ever gets this far down, I spent some time hacking on this trying to patch moment.js (https://github.com/moment/moment). Here's something that I took away from it:
function isNumeric(val) {
var _val = +val;
return (val !== val + 1) //infinity check
&& (_val === +val) //Cute coercion check
&& (typeof val !== 'object') //Array/object check
}
Handles the following cases:
True! :
isNumeric("1"))
isNumeric(1e10))
isNumeric(1E10))
isNumeric(+"6e4"))
isNumeric("1.2222"))
isNumeric("-1.2222"))
isNumeric("-1.222200000000000000"))
isNumeric("1.222200000000000000"))
isNumeric(1))
isNumeric(0))
isNumeric(-0))
isNumeric(1010010293029))
isNumeric(1.100393830000))
isNumeric(Math.LN2))
isNumeric(Math.PI))
isNumeric(5e10))
False! :
isNumeric(NaN))
isNumeric(Infinity))
isNumeric(-Infinity))
isNumeric())
isNumeric(undefined))
isNumeric('[1,2,3]'))
isNumeric({a:1,b:2}))
isNumeric(null))
isNumeric([1]))
isNumeric(new Date()))
Ironically, the one I am struggling with the most:
isNumeric(new Number(1)) => false
Any suggestions welcome. :]

I recently wrote an article about ways to ensure a variable is a valid number: https://github.com/jehugaleahsa/artifacts/blob/master/2018/typescript_num_hack.md The article explains how to ensure floating point or integer, if that's important (+x vs ~~x).
The article assumes the variable is a string or a number to begin with and trim is available/polyfilled. It wouldn't be hard to extend it to handle other types, as well. Here's the meat of it:
// Check for a valid float
if (x == null
|| ("" + x).trim() === ""
|| isNaN(+x)) {
return false; // not a float
}
// Check for a valid integer
if (x == null
|| ("" + x).trim() === ""
|| ~~x !== +x) {
return false; // not an integer
}

function isNumberCandidate(s) {
const str = (''+ s).trim();
if (str.length === 0) return false;
return !isNaN(+str);
}
console.log(isNumberCandidate('1')); // true
console.log(isNumberCandidate('a')); // false
console.log(isNumberCandidate('000')); // true
console.log(isNumberCandidate('1a')); // false
console.log(isNumberCandidate('1e')); // false
console.log(isNumberCandidate('1e-1')); // true
console.log(isNumberCandidate('123.3')); // true
console.log(isNumberCandidate('')); // false
console.log(isNumberCandidate(' ')); // false
console.log(isNumberCandidate(1)); // true
console.log(isNumberCandidate(0)); // true
console.log(isNumberCandidate(NaN)); // false
console.log(isNumberCandidate(undefined)); // false
console.log(isNumberCandidate(null)); // false
console.log(isNumberCandidate(-1)); // true
console.log(isNumberCandidate('-1')); // true
console.log(isNumberCandidate('-1.2')); // true
console.log(isNumberCandidate(0.0000001)); // true
console.log(isNumberCandidate('0.0000001')); // true
console.log(isNumberCandidate(Infinity)); // true
console.log(isNumberCandidate(-Infinity)); // true
console.log(isNumberCandidate('Infinity')); // true
if (isNumberCandidate(s)) {
// use +s as a number
+s ...
}

Well, I'm using this one I made...
It's been working so far:
function checkNumber(value) {
return value % 1 == 0;
}
If you spot any problem with it, tell me, please.

Checking the number in JS:
Best way for check if it's a number:
isFinite(20)
//True
Read a value out of a string. CSS *:
parseInt('2.5rem')
//2
parseFloat('2.5rem')
//2.5
For an integer:
isInteger(23 / 0)
//False
If value is NaN:
isNaN(20)
//False

This way it works for me.
function isNumeric(num){
let value1 = num.toString();
let value2 = parseFloat(num).toString();
return (value1 === value2);
}
console.log(
isNumeric(123), //true
isNumeric(-123), //true
isNumeric('123'), //true
isNumeric('-123'), //true
isNumeric(12.2), //true
isNumeric(-12.2), //true
isNumeric('12.2'), //true
isNumeric('-12.2'), //true
isNumeric('a123'), //false
isNumeric('123a'), //false
isNumeric(' 123'), //false
isNumeric('123 '), //false
isNumeric('a12.2'), //false
isNumeric('12.2a'), //false
isNumeric(' 12.2'), //false
isNumeric('12.2 '), //false
)

PFB the working solution:
function(check){
check = check + "";
var isNumber = check.trim().length>0? !isNaN(check):false;
return isNumber;
}

Save yourself the headache of trying to find a "built-in" solution.
There isn't a good answer, and the hugely upvoted answer in this thread is wrong.
npm install is-number
In JavaScript, it's not always as straightforward as it should be to reliably check if a value is a number. It's common for devs to use +, -, or Number() to cast a string value to a number (for example, when values are returned from user input, regex matches, parsers, etc). But there are many non-intuitive edge cases that yield unexpected results:
console.log(+[]); //=> 0
console.log(+''); //=> 0
console.log(+' '); //=> 0
console.log(typeof NaN); //=> 'number'

Check if String can be converted to number

I created the following Typescript extension to convert a string to Number:
declare global {
interface String {
toNumber(): number | null;
}
}
String.prototype.toNumber = function(this: string) {
return parseFloat(this);
}
When it is not possible to parse the string to number either because it is invalid, null, undefined, etc I would always like to return null.
How can I do this?

I am assuming you already understand the differences between parseFloat / Number as conversion mechanisms.
Effectively all you need to do is check if the output is NaN. You can do this by:
String.prototype.toNumber = function(this: string) {
const num = parseFloat(this);
return Number.isNaN(num) ? null : num;
}

If you want to return either a non-zero valid number (well, note that NaN is a number, but I think I know what you mean), then check for what you don't want before returning:
Object.defineProperty(String.prototype, "toNumber", {
value: function(str) {
let num = Number(str);
return num === 0 || isNaN(num) ? null : num;
}
});
(Defining properties directly on the prototype is a bad habit and can lead to weird behavior; using .defineProperty gives you a property that is not enumerable.)
Oh, and that's JavaScript, obviously, not Typescript.

A simple answer would be to use return Number(this) || null;
The Number function will convert to a number or NaN, NaN || null will return null (because NaN is falsey).
Updated added testing for zero condition, which with the above code would have also returned null. If that is not what you want, this code will allow zero to return. (Note that this can be done many different ways!):
const parsedValue = Number(this);
return parsedValue === 0 ? parsedValue : parsedValue || null;
Updated to use the parseFloat function, example of early exit for string of '0'. Very similar to the previous updated example.
if (this === '0') {
return 0;
}
return parseFloat(this) || null;

Issue with using isNaN()

Actually this is a simple program to check weather the number is divisible by 2 or not divisible by 2 or input given is not a number. According to my information isNaN('berry) should give me true as'berry' is a string but in my code this goes quite wrong.
The code is :
var isNum = function(number) {
// My code goes here!
if (number%2===0){
return true
}
else if (isNaN(number)){
console.log("enter the number not the string");
return number
}
else{
return false
}
};
isNum('berry');
The code above returns me false when I run it. Any help will be appreciated.
This is the screenshot

You can try using parseInt() or parseFloat() to convert variables to their integer or float equivalent before use elsewhere.
You can try using typeof to determine the variable type.
var isNum = function(number) {
var number_parsed = parseFloat( number );
if (isNaN(number) || typeof number == 'string' ){
console.log("enter the number not the string");
return number
} elseif(number_parsed % 2 ===0){
return true
} else{
return false
}
};
isNum('berry');

Your code gives the following output (run it here) and it's correctly working
var isNum = function(number) {
// My code goes here!
if (number%2===0){
return true
}
else if (isNaN(number)){
console.log("enter the number not the string");
return number
}
else{
return false
}
};
isNum('berry');

isNan() returns true when the argument is actually NaN. You provided 'berry', not NaN and so isNan() returns false.
Quoting the documentation:
The isNaN() function determines whether a value is NaN or not.
.
Return value
true if the given value is NaN; otherwise, false.

What causes isNaN to malfunction? [duplicate]

This question already has answers here:
Validate decimal numbers in JavaScript - IsNumeric()
(52 answers)
Closed 9 years ago.
I'm simply trying to evaluate if an input is a number, and figured isNaN would be the best way to go. However, this causes unreliable results. For instance, using the following method:
function isNumerical(value) {
var isNum = !isNaN(value);
return isNum ? "<mark>numerical</mark>" : "not numerical";
}
on these values:
isNumerical(123)); // => numerical
isNumerical("123")); // => numerical
isNumerical(null)); // => numerical
isNumerical(false)); // => numerical
isNumerical(true)); // => numerical
isNumerical()); // => not numerical
shown in this fiddle: http://jsfiddle.net/4nm7r/1
Why doesn't isNaN always work for me?

isNaN returns true if the value passed is not a number(NaN)(or if it cannot be converted to a number, so, null, true and false will be converted to 0), otherwise it returns false. In your case, you have to remove the ! before the function call!
It is very easy to understand the behaviour of your script. isNaN simply checks if a value can be converted to an int. To do this, you have just to multiply or divide it by 1, or subtract or add 0. In your case, if you do, inside your function, alert(value * 1); you will see that all those passed values will be replaced by a number(0, 1, 123) except for undefined, whose numerical value is NaN.
You can't compare any value to NaN because it will never return false(even NaN === NaN), I think that's because NaN is dynamically generated... But I'm not sure.
Anyway you can fix your code by simply doing this:
function isNumerical(value) {
var isNum = !isNaN(value / 1); //this will force the conversion to NaN or a number
return isNum ? "<mark>numerical</mark>" : "not numerical";
}

Your ternary statement is backward, if !isNaN() returns true you want to say "numerical"
return isNum ? "not numerical" : "<mark>numerical</mark>";
should be:
return isNum ? "<mark>numerical</mark>" : "not numerical";
See updated fiddle:
http://jsfiddle.net/4nm7r/1/

Now that you already fixed the reversed logic pointed out on other answers, use parseFloat to get rid of the false positives for true, false and null:
var isNum = !isNaN(parseFloat(value));
Just keep in mind the following kinds of outputs from parseFloat:
parseFloat("200$"); // 200
parseFloat("200,100"); // 200
parseFloat("200 foo"); // 200
parseFloat("$200"); // NaN
(i.e, if the string starts with a number, parseFloat will extract the first numeric part it can find)

I suggest you use additional checks:
function isNumerical(value) {
var isNum = !isNaN(value) && value !== null && value !== undefined;
return isNum ? "<mark>numerical</mark>" : "not numerical";
}
If you would like treat strings like 123 like not numerical than you should add one more check:
var isNum = !isNaN(value) && value !== null && value !== undefined && (typeof value === 'number');

How can I check if a string is a valid number?

I'm hoping there's something in the same conceptual space as the old VB6 IsNumeric() function?

2nd October 2020: note that many bare-bones approaches are fraught with subtle bugs (eg. whitespace, implicit partial parsing, radix, coercion of arrays etc.) that many of the answers here fail to take into account. The following implementation might work for you, but note that it does not cater for number separators other than the decimal point ".":
function isNumeric(str) {
if (typeof str != "string") return false // we only process strings!
return !isNaN(str) && // use type coercion to parse the _entirety_ of the string (`parseFloat` alone does not do this)...
!isNaN(parseFloat(str)) // ...and ensure strings of whitespace fail
}
To check if a variable (including a string) is a number, check if it is not a number:
This works regardless of whether the variable content is a string or number.
isNaN(num) // returns true if the variable does NOT contain a valid number
Examples
isNaN(123) // false
isNaN('123') // false
isNaN('1e10000') // false (This translates to Infinity, which is a number)
isNaN('foo') // true
isNaN('10px') // true
isNaN('') // false
isNaN(' ') // false
isNaN(false) // false
Of course, you can negate this if you need to. For example, to implement the IsNumeric example you gave:
function isNumeric(num){
return !isNaN(num)
}
To convert a string containing a number into a number:
Only works if the string only contains numeric characters, else it returns NaN.
+num // returns the numeric value of the string, or NaN
// if the string isn't purely numeric characters
Examples
+'12' // 12
+'12.' // 12
+'12..' // NaN
+'.12' // 0.12
+'..12' // NaN
+'foo' // NaN
+'12px' // NaN
To convert a string loosely to a number
Useful for converting '12px' to 12, for example:
parseInt(num) // extracts a numeric value from the
// start of the string, or NaN.
Examples
parseInt('12') // 12
parseInt('aaa') // NaN
parseInt('12px') // 12
parseInt('foo2') // NaN These last three may
parseInt('12a5') // 12 be different from what
parseInt('0x10') // 16 you expected to see.
Floats
Bear in mind that, unlike +num, parseInt (as the name suggests) will convert a float into an integer by chopping off everything following the decimal point (if you want to use parseInt() because of this behaviour, you're probably better off using another method instead):
+'12.345' // 12.345
parseInt(12.345) // 12
parseInt('12.345') // 12
Empty strings
Empty strings may be a little counter-intuitive. +num converts empty strings or strings with spaces to zero, and isNaN() assumes the same:
+'' // 0
+' ' // 0
isNaN('') // false
isNaN(' ') // false
But parseInt() does not agree:
parseInt('') // NaN
parseInt(' ') // NaN

If you're just trying to check if a string is a whole number (no decimal places), regex is a good way to go. Other methods such as isNaN are too complicated for something so simple.
function isNumeric(value) {
return /^-?\d+$/.test(value);
}
console.log(isNumeric('abcd')); // false
console.log(isNumeric('123a')); // false
console.log(isNumeric('1')); // true
console.log(isNumeric('1234567890')); // true
console.log(isNumeric('-23')); // true
console.log(isNumeric(1234)); // true
console.log(isNumeric(1234n)); // true
console.log(isNumeric('123.4')); // false
console.log(isNumeric('')); // false
console.log(isNumeric(undefined)); // false
console.log(isNumeric(null)); // false
To only allow positive whole numbers use this:
function isNumeric(value) {
return /^\d+$/.test(value);
}
console.log(isNumeric('123')); // true
console.log(isNumeric('-23')); // false

The accepted answer for this question has quite a few flaws (as highlighted by couple of other users). This is one of the easiest & proven way to approach it in javascript:
function isNumeric(n) {
return !isNaN(parseFloat(n)) && isFinite(n);
}
Below are some good test cases:
console.log(isNumeric(12345678912345678912)); // true
console.log(isNumeric('2 ')); // true
console.log(isNumeric('-32.2 ')); // true
console.log(isNumeric(-32.2)); // true
console.log(isNumeric(undefined)); // false
// the accepted answer fails at these tests:
console.log(isNumeric('')); // false
console.log(isNumeric(null)); // false
console.log(isNumeric([])); // false

And you could go the RegExp-way:
var num = "987238";
if(num.match(/^-?\d+$/)){
//valid integer (positive or negative)
}else if(num.match(/^\d+\.\d+$/)){
//valid float
}else{
//not valid number
}

If you really want to make sure that a string contains only a number, any number (integer or floating point), and exactly a number, you cannot use parseInt()/ parseFloat(), Number(), or !isNaN() by themselves. Note that !isNaN() is actually returning true when Number() would return a number, and false when it would return NaN, so I will exclude it from the rest of the discussion.
The problem with parseFloat() is that it will return a number if the string contains any number, even if the string doesn't contain only and exactly a number:
parseFloat("2016-12-31") // returns 2016
parseFloat("1-1") // return 1
parseFloat("1.2.3") // returns 1.2
The problem with Number() is that it will return a number in cases where the passed value is not a number at all!
Number("") // returns 0
Number(" ") // returns 0
Number(" \u00A0 \t\n\r") // returns 0
The problem with rolling your own regex is that unless you create the exact regex for matching a floating point number as Javascript recognizes it you are going to miss cases or recognize cases where you shouldn't. And even if you can roll your own regex, why? There are simpler built-in ways to do it.
However, it turns out that Number() (and isNaN()) does the right thing for every case where parseFloat() returns a number when it shouldn't, and vice versa. So to find out if a string is really exactly and only a number, call both functions and see if they both return true:
function isNumber(str) {
if (typeof str != "string") return false // we only process strings!
// could also coerce to string: str = ""+str
return !isNaN(str) && !isNaN(parseFloat(str))
}

Try the isNan function:
The isNaN() function determines whether a value is an illegal number (Not-a-Number).
This function returns true if the value equates to NaN. Otherwise it returns false.
This function is different from the Number specific Number.isNaN() method.
The global isNaN() function, converts the tested value to a Number, then tests it.
Number.isNan() does not convert the values to a Number, and will not return true for any value that is not of the type Number...

2019: Including ES3, ES6 and TypeScript Examples
Maybe this has been rehashed too many times, however I fought with this one today too and wanted to post my answer, as I didn't see any other answer that does it as simply or thoroughly:
ES3
var isNumeric = function(num){
return (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num);
}
ES6
const isNumeric = (num) => (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num);
Typescript
const isNumeric = (num: any) => (typeof(num) === 'number' || typeof(num) === "string" && num.trim() !== '') && !isNaN(num as number);
This seems quite simple and covers all the bases I saw on the many other posts and thought up myself:
// Positive Cases
console.log(0, isNumeric(0) === true);
console.log(1, isNumeric(1) === true);
console.log(1234567890, isNumeric(1234567890) === true);
console.log('1234567890', isNumeric('1234567890') === true);
console.log('0', isNumeric('0') === true);
console.log('1', isNumeric('1') === true);
console.log('1.1', isNumeric('1.1') === true);
console.log('-1', isNumeric('-1') === true);
console.log('-1.2354', isNumeric('-1.2354') === true);
console.log('-1234567890', isNumeric('-1234567890') === true);
console.log(-1, isNumeric(-1) === true);
console.log(-32.1, isNumeric(-32.1) === true);
console.log('0x1', isNumeric('0x1') === true); // Valid number in hex
// Negative Cases
console.log(true, isNumeric(true) === false);
console.log(false, isNumeric(false) === false);
console.log('1..1', isNumeric('1..1') === false);
console.log('1,1', isNumeric('1,1') === false);
console.log('-32.1.12', isNumeric('-32.1.12') === false);
console.log('[blank]', isNumeric('') === false);
console.log('[spaces]', isNumeric(' ') === false);
console.log('null', isNumeric(null) === false);
console.log('undefined', isNumeric(undefined) === false);
console.log([], isNumeric([]) === false);
console.log('NaN', isNumeric(NaN) === false);
You can also try your own isNumeric function and just past in these use cases and scan for "true" for all of them.
Or, to see the values that each return:

Old question, but there are several points missing in the given answers.
Scientific notation.
!isNaN('1e+30') is true, however in most of the cases when people ask for numbers, they do not want to match things like 1e+30.
Large floating numbers may behave weird
Observe (using Node.js):
> var s = Array(16 + 1).join('9')
undefined
> s.length
16
> s
'9999999999999999'
> !isNaN(s)
true
> Number(s)
10000000000000000
> String(Number(s)) === s
false
>
On the other hand:
> var s = Array(16 + 1).join('1')
undefined
> String(Number(s)) === s
true
> var s = Array(15 + 1).join('9')
undefined
> String(Number(s)) === s
true
>
So, if one expects String(Number(s)) === s, then better limit your strings to 15 digits at most (after omitting leading zeros).
Infinity
> typeof Infinity
'number'
> !isNaN('Infinity')
true
> isFinite('Infinity')
false
>
Given all that, checking that the given string is a number satisfying all of the following:
non scientific notation
predictable conversion to Number and back to String
finite
is not such an easy task. Here is a simple version:
function isNonScientificNumberString(o) {
if (!o || typeof o !== 'string') {
// Should not be given anything but strings.
return false;
}
return o.length <= 15 && o.indexOf('e+') < 0 && o.indexOf('E+') < 0 && !isNaN(o) && isFinite(o);
}
However, even this one is far from complete. Leading zeros are not handled here, but they do screw the length test.

The JavaScript global isFinite() checks if a value is a valid (finite) number.
See MDN for the difference between Number.isFinite() and global isFinite().
let a = isFinite('abc') // false
let b = isFinite('123') // true
let c = isFinite('12a') // false
let d = isFinite(null) // true
console.log(a, b, c, d)

I have tested and Michael's solution is best. Vote for his answer above (search this page for "If you really want to make sure that a string" to find it). In essence, his answer is this:
function isNumeric(num){
num = "" + num; //coerce num to be a string
return !isNaN(num) && !isNaN(parseFloat(num));
}
It works for every test case, which I documented here:
https://jsfiddle.net/wggehvp9/5/
Many of the other solutions fail for these edge cases:
' ', null, "", true, and [].
In theory, you could use them, with proper error handling, for example:
return !isNaN(num);
or
return (+num === +num);
with special handling for
/\s/, null, "", true, false, [] (and others?)

TL;DR
It depends largely on what you want to parse as a number.
Comparison Between Built-in Functions
As none of the existing sources satisfied my soul, I tried to figure out what actually was happening with these functions.
Three immediate answers to this question felt like:
!isNaN(input) (which gives the same output as +input === +input)
!isNaN(parseFloat(input))
isFinite(input)
But are any of them correct in every scenario?
I tested these functions in several cases, and generated output as markdown. This is what it looks like:
input
!isNaN(input) or +input===+input
!isNaN(parseFloat(input))
isFinite(input)
Comment
123
✔️
✔️
✔️
-
'123'
✔️
✔️
✔️
-
12.3
✔️
✔️
✔️
-
'12.3'
✔️
✔️
✔️
-
'   12.3   '
✔️
✔️
✔️
Empty whitespace trimmed, as expected.
1_000_000
✔️
✔️
✔️
Numeric separator understood, also expected.
'1_000_000'
❌
✔️
❌
Surprise! JS just won't parse numeric separator inside a string. For details, check this issue. (Why then parsing as float worked though? Well, it didn't. 😉)
'0b11111111'
✔️
✔️
✔️
Binary form understood, as it should've.
'0o377'
✔️
✔️
✔️
Octal form understood too.
'0xFF'
✔️
✔️
✔️
Of course hex is understood. Did anybody think otherwise? 😒
''
✔️
❌
✔️
Should empty string be a number?
'    '
✔️
❌
✔️
Should a whitespace-only string be a number?
'abc'
❌
❌
❌
Everybody agrees, not a number.
'12.34Ab!##$'
❌
✔️
❌
Ah! Now it's quite understandable what parseFloat() does. Not impressive to me, but may come handy in certain cases.
'10e100'
✔️
✔️
✔️
10100 is a number indeed. But caution! It's way more larger than the maximum safe integer value 253 (about 9×1015). Read this for details.
'10e1000'
✔️
✔️
❌
Say with me, help! Though not as crazy as it may seem. In JavaScript, a value larger than ~10308 is rounded to infinity, that's why. Look here for details. And yes, isNaN() considers infinity as a number, and parseFloat() parses infinity as infinity.
null
✔️
❌
✔️
Now this is awkward. In JS, when a conversion is needed, null becomes zero, and we get a finite number. Then why parseFloat(null) should return a NaN here? Someone please explain this design concept to me.
undefined
❌
❌
❌
As expected.
Infinity
✔️
✔️
❌
As explained before, isNaN() considers infinity as a number, and parseFloat() parses infinity as infinity.
So...which of them is "correct"?
Should be clear by now, it depends largely on what we need. For example, we may want to consider a null input as 0. In that case isFinite() will work fine.
Again, perhaps we will take a little help from isNaN() when 1010000000000 is needed to be considered a valid number (although the question remains—why would it be, and how would we handle that)!
Of course, we can manually exclude any of the scenarios.
Like in my case, I needed exactly the outputs of isFinite(), except for the null case, the empty string case, and the whitespace-only string case. Also I had no headache about really huge numbers. So my code looked like this:
/**
* My necessity was met by the following code.
*/
if (input === null) {
// Null input
} else if (input.trim() === '') {
// Empty or whitespace-only string
} else if (isFinite(input)) {
// Input is a number
} else {
// Not a number
}
And also, this was my JavaScript to generate the table:
/**
* Note: JavaScript does not print numeric separator inside a number.
* In that single case, the markdown output was manually corrected.
* Also, the comments were manually added later, of course.
*/
let inputs = [
123, '123', 12.3, '12.3', ' 12.3 ',
1_000_000, '1_000_000',
'0b11111111', '0o377', '0xFF',
'', ' ',
'abc', '12.34Ab!##$',
'10e100', '10e1000',
null, undefined, Infinity];
let markdownOutput = `| \`input\` | \`!isNaN(input)\` or <br>\`+input === +input\` | \`!isNaN(parseFloat(input))\` | \`isFinite(input)\` | Comment |
| :---: | :---: | :---: | :---: | :--- |\n`;
for (let input of inputs) {
let outputs = [];
outputs.push(!isNaN(input));
outputs.push(!isNaN(parseFloat(input)));
outputs.push(isFinite(input));
if (typeof input === 'string') {
// Output with quotations
console.log(`'${input}'`);
markdownOutput += `| '${input}'`;
} else {
// Output without quotes
console.log(input);
markdownOutput += `| ${input}`;
}
for (let output of outputs) {
console.log('\t' + output);
if (output === true) {
markdownOutput += ` | <div style="color:limegreen">true</div>`;
// markdownOutput += ` | ✔️`; // for stackoverflow
} else {
markdownOutput += ` | <div style="color:orangered">false</div>`;
// markdownOutput += ` | ❌`; // for stackoverflow
}
}
markdownOutput += ` ||\n`;
}
// Replace two or more whitespaces with $nbsp;
markdownOutput = markdownOutput.replaceAll(` `, ` `);
// Print markdown to console
console.log(markdownOutput);

Someone may also benefit from a regex based answer. Here it is:
One liner isInteger:
const isInteger = num => /^-?[0-9]+$/.test(num+'');
One liner isNumeric: Accepts integers and decimals
const isNumeric = num => /^-?[0-9]+(?:\.[0-9]+)?$/.test(num+'');

You can use the result of Number when passing an argument to its constructor.
If the argument (a string) cannot be converted into a number, it returns NaN, so you can determinate if the string provided was a valid number or not.
Notes: Note when passing empty string or '\t\t' and '\n\t' as Number will return 0; Passing true will return 1 and false returns 0.
Number('34.00') // 34
Number('-34') // -34
Number('123e5') // 12300000
Number('123e-5') // 0.00123
Number('999999999999') // 999999999999
Number('9999999999999999') // 10000000000000000 (integer accuracy up to 15 digit)
Number('0xFF') // 255
Number('Infinity') // Infinity
Number('34px') // NaN
Number('xyz') // NaN
Number('true') // NaN
Number('false') // NaN
// cavets
Number(' ') // 0
Number('\t\t') // 0
Number('\n\t') // 0

Maybe there are one or two people coming across this question who need a much stricter check than usual (like I did). In that case, this might be useful:
if(str === String(Number(str))) {
// it's a "perfectly formatted" number
}
Beware! This will reject strings like .1, 40.000, 080, 00.1. It's very picky - the string must match the "most minimal perfect form" of the number for this test to pass.
It uses the String and Number constructor to cast the string to a number and back again and thus checks if the JavaScript engine's "perfect minimal form" (the one it got converted to with the initial Number constructor) matches the original string.

Why is jQuery's implementation not good enough?
function isNumeric(a) {
var b = a && a.toString();
return !$.isArray(a) && b - parseFloat(b) + 1 >= 0;
};
Michael suggested something like this (although I've stolen "user1691651 - John"'s altered version here):
function isNumeric(num){
num = "" + num; //coerce num to be a string
return !isNaN(num) && !isNaN(parseFloat(num));
}
The following is a solution with most likely bad performance, but solid results. It is a contraption made from the jQuery 1.12.4 implementation and Michael's answer, with an extra check for leading/trailing spaces (because Michael's version returns true for numerics with leading/trailing spaces):
function isNumeric(a) {
var str = a + "";
var b = a && a.toString();
return !$.isArray(a) && b - parseFloat(b) + 1 >= 0 &&
!/^\s+|\s+$/g.test(str) &&
!isNaN(str) && !isNaN(parseFloat(str));
};
The latter version has two new variables, though. One could get around one of those, by doing:
function isNumeric(a) {
if ($.isArray(a)) return false;
var b = a && a.toString();
a = a + "";
return b - parseFloat(b) + 1 >= 0 &&
!/^\s+|\s+$/g.test(a) &&
!isNaN(a) && !isNaN(parseFloat(a));
};
I haven't tested any of these very much, by other means than manually testing the few use-cases I'll be hitting with my current predicament, which is all very standard stuff. This is a "standing-on-the-shoulders-of-giants" situation.

2019: Practical and tight numerical validity check
Often, a 'valid number' means a Javascript number excluding NaN and Infinity, ie a 'finite number'.
To check the numerical validity of a value (from an external source for example), you can define in ESlint Airbnb style :
/**
* Returns true if 'candidate' is a finite number or a string referring (not just 'including') a finite number
* To keep in mind:
* Number(true) = 1
* Number('') = 0
* Number(" 10 ") = 10
* !isNaN(true) = true
* parseFloat('10 a') = 10
*
* #param {?} candidate
* #return {boolean}
*/
function isReferringFiniteNumber(candidate) {
if (typeof (candidate) === 'number') return Number.isFinite(candidate);
if (typeof (candidate) === 'string') {
return (candidate.trim() !== '') && Number.isFinite(Number(candidate));
}
return false;
}
and use it this way:
if (isReferringFiniteNumber(theirValue)) {
myCheckedValue = Number(theirValue);
} else {
console.warn('The provided value doesn\'t refer to a finite number');
}

It is not valid for TypeScript as:
declare function isNaN(number: number): boolean;
For TypeScript you can use:
/^\d+$/.test(key)

I like the simplicity of this.
Number.isNaN(Number(value))
The above is regular Javascript, but I'm using this in conjunction with a typescript typeguard for smart type checking. This is very useful for the typescript compiler to give you correct intellisense, and no type errors.
Typescript typeguards
Warning: See Jeremy's comment below. This has some issues with certain values and I don't have time to fix it now, but the idea of using a typescript typeguard is useful so I won't delete this section.
isNotNumber(value: string | number): value is string {
return Number.isNaN(Number(this.smartImageWidth));
}
isNumber(value: string | number): value is number {
return Number.isNaN(Number(this.smartImageWidth)) === false;
}
Let's say you have a property width which is number | string. You may want to do logic based on whether or not it's a string.
var width: number|string;
width = "100vw";
if (isNotNumber(width))
{
// the compiler knows that width here must be a string
if (width.endsWith('vw'))
{
// we have a 'width' such as 100vw
}
}
else
{
// the compiler is smart and knows width here must be number
var doubleWidth = width * 2;
}
The typeguard is smart enough to constrain the type of width within the if statement to be ONLY string. This permits the compiler to allow width.endsWith(...) which it wouldn't allow if the type was string | number.
You can call the typeguard whatever you want isNotNumber, isNumber, isString, isNotString but I think isString is kind of ambiguous and harder to read.

parseInt(), but be aware that this function is a bit different in the sense that it for example returns 100 for parseInt("100px").

When guarding against empty strings and null
// Base cases that are handled properly
Number.isNaN(Number('1')); // => false
Number.isNaN(Number('-1')); // => false
Number.isNaN(Number('1.1')); // => false
Number.isNaN(Number('-1.1')); // => false
Number.isNaN(Number('asdf')); // => true
Number.isNaN(Number(undefined)); // => true
// Special notation cases that are handled properly
Number.isNaN(Number('1e1')); // => false
Number.isNaN(Number('1e-1')); // => false
Number.isNaN(Number('-1e1')); // => false
Number.isNaN(Number('-1e-1')); // => false
Number.isNaN(Number('0b1')); // => false
Number.isNaN(Number('0o1')); // => false
Number.isNaN(Number('0xa')); // => false
// Edge cases that will FAIL if not guarded against
Number.isNaN(Number('')); // => false
Number.isNaN(Number(' ')); // => false
Number.isNaN(Number(null)); // => false
// Edge cases that are debatable
Number.isNaN(Number('-0b1')); // => true
Number.isNaN(Number('-0o1')); // => true
Number.isNaN(Number('-0xa')); // => true
Number.isNaN(Number('Infinity')); // => false
Number.isNaN(Number('INFINITY')); // => true
Number.isNaN(Number('-Infinity')); // => false
Number.isNaN(Number('-INFINITY')); // => true
When NOT guarding against empty strings and null
Using parseInt:
// Base cases that are handled properly
Number.isNaN(parseInt('1')); // => false
Number.isNaN(parseInt('-1')); // => false
Number.isNaN(parseInt('1.1')); // => false
Number.isNaN(parseInt('-1.1')); // => false
Number.isNaN(parseInt('asdf')); // => true
Number.isNaN(parseInt(undefined)); // => true
Number.isNaN(parseInt('')); // => true
Number.isNaN(parseInt(' ')); // => true
Number.isNaN(parseInt(null)); // => true
// Special notation cases that are handled properly
Number.isNaN(parseInt('1e1')); // => false
Number.isNaN(parseInt('1e-1')); // => false
Number.isNaN(parseInt('-1e1')); // => false
Number.isNaN(parseInt('-1e-1')); // => false
Number.isNaN(parseInt('0b1')); // => false
Number.isNaN(parseInt('0o1')); // => false
Number.isNaN(parseInt('0xa')); // => false
// Edge cases that are debatable
Number.isNaN(parseInt('-0b1')); // => false
Number.isNaN(parseInt('-0o1')); // => false
Number.isNaN(parseInt('-0xa')); // => false
Number.isNaN(parseInt('Infinity')); // => true
Number.isNaN(parseInt('INFINITY')); // => true
Number.isNaN(parseInt('-Infinity')); // => true
Number.isNaN(parseInt('-INFINITY')); // => true
Using parseFloat:
// Base cases that are handled properly
Number.isNaN(parseFloat('1')); // => false
Number.isNaN(parseFloat('-1')); // => false
Number.isNaN(parseFloat('1.1')); // => false
Number.isNaN(parseFloat('-1.1')); // => false
Number.isNaN(parseFloat('asdf')); // => true
Number.isNaN(parseFloat(undefined)); // => true
Number.isNaN(parseFloat('')); // => true
Number.isNaN(parseFloat(' ')); // => true
Number.isNaN(parseFloat(null)); // => true
// Special notation cases that are handled properly
Number.isNaN(parseFloat('1e1')); // => false
Number.isNaN(parseFloat('1e-1')); // => false
Number.isNaN(parseFloat('-1e1')); // => false
Number.isNaN(parseFloat('-1e-1')); // => false
Number.isNaN(parseFloat('0b1')); // => false
Number.isNaN(parseFloat('0o1')); // => false
Number.isNaN(parseFloat('0xa')); // => false
// Edge cases that are debatable
Number.isNaN(parseFloat('-0b1')); // => false
Number.isNaN(parseFloat('-0o1')); // => false
Number.isNaN(parseFloat('-0xa')); // => false
Number.isNaN(parseFloat('Infinity')); // => false
Number.isNaN(parseFloat('INFINITY')); // => true
Number.isNaN(parseFloat('-Infinity')); // => false
Number.isNaN(parseFloat('-INFINITY')); // => true
Notes:
Only string, empty, and uninitialized values are considered in keeping with addressing the original question. Additional edge cases exist if arrays and objects are the values being considered.
Characters in binary, octal, hexadecimal, and exponential notation are not case-sensitive (ie: '0xFF', '0XFF', '0xfF' etc. will all yield the same result in the test cases shown above).
Unlike with Infinity (case-sensitive) in some cases, constants from the Number and Math objects passed as test cases in string format to any of the methods above will be determined to not be numbers.
See here for an explanation of how arguments are converted to a Number and why the edge cases for null and empty strings exist.

Quote:
isNaN(num) // returns true if the variable does NOT contain a valid number
is not entirely true if you need to check for leading/trailing spaces - for example when a certain quantity of digits is required, and you need to get, say, '1111' and not ' 111' or '111 ' for perhaps a PIN input.
Better to use:
var num = /^\d+$/.test(num)

This is built on some of the previous answers and comments. It covers all the edge cases and also handles scientific notation optionally:
const NUMBER_REG_EXP = /^-?\d+(?:\.\d+)?$/;
const SCIENTIFIC_NOTATION_REG_EXP = /^-?\d+(?:\.\d+)?(?:[eE]\d+)?$/;
const isNumeric = (n, allowScientificNotation = false) => (
(typeof n === 'number' && !Number.isNaN(n)) ||
(typeof n === 'string' && (allowScientificNotation ?
SCIENTIFIC_NOTATION_REG_EXP : NUMBER_REG_EXP).test(n))
);

If anyone ever gets this far down, I spent some time hacking on this trying to patch moment.js (https://github.com/moment/moment). Here's something that I took away from it:
function isNumeric(val) {
var _val = +val;
return (val !== val + 1) //infinity check
&& (_val === +val) //Cute coercion check
&& (typeof val !== 'object') //Array/object check
}
Handles the following cases:
True! :
isNumeric("1"))
isNumeric(1e10))
isNumeric(1E10))
isNumeric(+"6e4"))
isNumeric("1.2222"))
isNumeric("-1.2222"))
isNumeric("-1.222200000000000000"))
isNumeric("1.222200000000000000"))
isNumeric(1))
isNumeric(0))
isNumeric(-0))
isNumeric(1010010293029))
isNumeric(1.100393830000))
isNumeric(Math.LN2))
isNumeric(Math.PI))
isNumeric(5e10))
False! :
isNumeric(NaN))
isNumeric(Infinity))
isNumeric(-Infinity))
isNumeric())
isNumeric(undefined))
isNumeric('[1,2,3]'))
isNumeric({a:1,b:2}))
isNumeric(null))
isNumeric([1]))
isNumeric(new Date()))
Ironically, the one I am struggling with the most:
isNumeric(new Number(1)) => false
Any suggestions welcome. :]

I recently wrote an article about ways to ensure a variable is a valid number: https://github.com/jehugaleahsa/artifacts/blob/master/2018/typescript_num_hack.md The article explains how to ensure floating point or integer, if that's important (+x vs ~~x).
The article assumes the variable is a string or a number to begin with and trim is available/polyfilled. It wouldn't be hard to extend it to handle other types, as well. Here's the meat of it:
// Check for a valid float
if (x == null
|| ("" + x).trim() === ""
|| isNaN(+x)) {
return false; // not a float
}
// Check for a valid integer
if (x == null
|| ("" + x).trim() === ""
|| ~~x !== +x) {
return false; // not an integer
}

function isNumberCandidate(s) {
const str = (''+ s).trim();
if (str.length === 0) return false;
return !isNaN(+str);
}
console.log(isNumberCandidate('1')); // true
console.log(isNumberCandidate('a')); // false
console.log(isNumberCandidate('000')); // true
console.log(isNumberCandidate('1a')); // false
console.log(isNumberCandidate('1e')); // false
console.log(isNumberCandidate('1e-1')); // true
console.log(isNumberCandidate('123.3')); // true
console.log(isNumberCandidate('')); // false
console.log(isNumberCandidate(' ')); // false
console.log(isNumberCandidate(1)); // true
console.log(isNumberCandidate(0)); // true
console.log(isNumberCandidate(NaN)); // false
console.log(isNumberCandidate(undefined)); // false
console.log(isNumberCandidate(null)); // false
console.log(isNumberCandidate(-1)); // true
console.log(isNumberCandidate('-1')); // true
console.log(isNumberCandidate('-1.2')); // true
console.log(isNumberCandidate(0.0000001)); // true
console.log(isNumberCandidate('0.0000001')); // true
console.log(isNumberCandidate(Infinity)); // true
console.log(isNumberCandidate(-Infinity)); // true
console.log(isNumberCandidate('Infinity')); // true
if (isNumberCandidate(s)) {
// use +s as a number
+s ...
}

Well, I'm using this one I made...
It's been working so far:
function checkNumber(value) {
return value % 1 == 0;
}
If you spot any problem with it, tell me, please.

Checking the number in JS:
Best way for check if it's a number:
isFinite(20)
//True
Read a value out of a string. CSS *:
parseInt('2.5rem')
//2
parseFloat('2.5rem')
//2.5
For an integer:
isInteger(23 / 0)
//False
If value is NaN:
isNaN(20)
//False

This way it works for me.
function isNumeric(num){
let value1 = num.toString();
let value2 = parseFloat(num).toString();
return (value1 === value2);
}
console.log(
isNumeric(123), //true
isNumeric(-123), //true
isNumeric('123'), //true
isNumeric('-123'), //true
isNumeric(12.2), //true
isNumeric(-12.2), //true
isNumeric('12.2'), //true
isNumeric('-12.2'), //true
isNumeric('a123'), //false
isNumeric('123a'), //false
isNumeric(' 123'), //false
isNumeric('123 '), //false
isNumeric('a12.2'), //false
isNumeric('12.2a'), //false
isNumeric(' 12.2'), //false
isNumeric('12.2 '), //false
)

PFB the working solution:
function(check){
check = check + "";
var isNumber = check.trim().length>0? !isNaN(check):false;
return isNumber;
}

Save yourself the headache of trying to find a "built-in" solution.
There isn't a good answer, and the hugely upvoted answer in this thread is wrong.
npm install is-number
In JavaScript, it's not always as straightforward as it should be to reliably check if a value is a number. It's common for devs to use +, -, or Number() to cast a string value to a number (for example, when values are returned from user input, regex matches, parsers, etc). But there are many non-intuitive edge cases that yield unexpected results:
console.log(+[]); //=> 0
console.log(+''); //=> 0
console.log(+' '); //=> 0
console.log(typeof NaN); //=> 'number'