General tips for representing large numbers

Question

Sometimes, while golfing, one needs to represent large number(s) in their code. Writing them as is can significantly increase the byte-count.

What general¹ tips do you have for representing long numbers concisely in code?

Please post one tip per answer.

---

¹With general, I mean tips that can be applied to more than a single language. For language-specific tips, post in their respective thread.

Answer 1

Look out for special numbers

Some languages have built-in functions for squares, exponentiation with base 2, n-th prime, factorial, or other procedures that can generate large numbers. Check if you number happens to fall into any of those categories.

And if it doesn't, it may happen that a larger number that does will fit for your purposes and can be used instead.

Answer 2

Use bitwise boolean operators

Some languages have bitwise AND, OR, XOR, and sometimes NOT.

Expressing a specific large number as a bitwise combination of a result of an exponentiation or left shift and another number can get you to precisely the number you need. This is usually only worth it if the numbers get quite big.

For example, 2147483722 is 10 bytes, but 2<<30^74 (2^31 bitwise-XORed with 74) is only 8.

Answer 3

Use Scientific Notation

Scientific notation can save bytes in case of long numbers. For Example:

3564e-8 // returns 0.00003564 (saves 3 bytes!)

Answer 4

Use Strings for repetitive numbers

For numbers that are very repetitive in nature, you can use Strings and cast them to Integer. For Example, in JavaScript

+"1".repeat(100) // returns 100 1s (saves 84 bytes!)

Answer 5

Base Compression

Base decompression code can be fairly complex, but if you have a truly enormous number sometimes it can help to compress it in some base higher than 10.

It also helps that in some languages, the base compression code is very simple. For example, PHP has base64_decode(_), Python has int(_,36), JavaScript has parseInt(_,36), and many golfing languages have base decompression builtins. For example, in CJam:

"+ÜTbô±"256b

This contains an unprintable. Try it online!

This yields:

12345678987654321

Answer 6

Look for another number to use instead

This may sound like a non-answer, but it is not always obvious that a larger number can be calculated by shorter code. An example I remember is Output a googol copies of a string, where the obvious answers require computing 10¹⁰⁰. As it turns out, computing any multiple of 10¹⁰⁰ leads to an equally correct, but in some languages, shorter answer. Dennis's answer there uses 100¹⁰⁰, my own uses 250²⁵⁵.

Answer 7

Use Bitwise Left Shift for 2's exponentiation

Although, there are many languages that support operator for exponentiation, some do not. And those which don't, usually require calling functions (or Class/Object methods), which can cost a few bytes.

But you can save some bytes when you need to raise 2 to the power n by using the Bitwise Left Shift operator << as 1<<n. Note that this will only save you bytes if n is greater than or equal to 17. However, this will always save you bytes if n is dynamic. Few Examples:

1<<2 // returns 4 (3 bytes more :( )
1<<3 // returns 8 (3 bytes more :( )
1<<6 // returns 64 (2 bytes more :( )
1<<14 // returns 16384 (no bytes saved)
1<<17 // returns 131072 (saves 1 byte!)
1<<18 // returns 262114 (saves 1 byte!)

Answer 8

Use exponential fractions for large repetitive numbers

Say you wanted to generate the number made of 100 1's. You could use int("1"*100), +"1".repeat(100), etc. but you can also take advantage of the fact that it's very close to

1e100/9

This works best for very repetitive numbers, such as those made of a single digit. A couple repeated digits also works fairly well:

12e100/99  // Generates 121212121212... (100 digits)

Occasionally you'll find some other weird pattern which also can be represented fairly tersely in this method. If you happened to need int("123456790"*11), for example:

1e100/81

Be careful, though: numbers such as int("1234567890"*10) don't have such an easy representation.

Answer 9

Chinese Remainder Theorem

If arbitrary big integers frequently appear, or big integer representation in target programming language costs too many bytes, you can consider using Chinese Remainder Theorem.

Choose some pairwise relatively prime integers m_i >=2, and you can express a big number from 0 to lcm(m₁, m₂, ... , m_i) -1

For example, I choose 2, 3, 5, 11, 79, 83, 89, 97, then I can express number less than 18680171730 uniquely. 10000000000 (1e10) can be expressed as 0,1,0,1,38,59,50,49 (1e10 mod 2, 3 ... , 97) which need not be expressed as special Big Integer class/struct which might save some bytes in some programming language.

Addition and substraction can be done directly using this representation. Example:

(0,1,0,1,38,59,50,49)+(0,2,0,6,23,20,16,53) = 1e10 + 5000 
                                            = (0+0 mod 2, 1+2 mod 3, 0+0 mod 5, 1+6 mod 11, 38+23 mod 79, 59+20 mod 83, 50+16 mod 89, 49+53 mod 97)

Answer 10

Use String Padding (where possible)

If a large number includes a repeating digit at the beginning or end, you may be able to save bytes by using your one of your language's padding methods to construct a string of the number you're looking for, which you can then convert to an integer.

---

Example

To generate the number 1111111111111111111111112 (25 bytes) in JavaScript (ES8):

+"2".padStart(25,1) // 19 bytes

Answer 11

Use Exponents

If your language has an exponent operator you might be able to use it to generate, if not the number you want, at least a number you can perform a simple calculation or 2 on to arrive at your number. Even without an operator, you may still be able to save bytes with a built-in function or method.

Example

The maximum safe integer in JavaScript is 9007199254740991, which is 16 digits long. In ES7, this can be calculated with the following 7 bytes:

2**53-1

The equivalent in ES6 and earlier, while the same length as the integer itself in this instance, demonstrates that using a more verbose method might not necessarily cost you any bytes.

Math.pow(2,53)-1

The above, though, may work out shorter if, for example, you already have Math aliased to a single character elsewhere in your code.

Answer 12

Use fractions in the place of float

Example: 1./3 in place of 0.333333333