Compression Algorithms

AXON automatically applies the most effective compression based on your data patterns

Automatic Selection: You don't need to choose compression algorithms manually. AXON analyzes your data and applies the best combination automatically.

1. RLE (Run-Length Encoding)

Best For

Repeated values

204×

efficiency

How it works:

Compresses sequences of identical values by storing the value once plus a count.

Example:

JSON (verbose)

age: [30, 30, 30, 30, 30]

AXON (compressed)

age:30*5

Savings: 99.5% reduction for 100 repeated values

2. Dictionary Compression

Best For

Limited unique values (high cardinality)

4.3×

efficiency

How it works:

Creates a lookup table of unique values, then references them by index.

Example:

JSON

role: ["admin", "user", "admin", "guest", "user"]

AXON

@dict(admin,user,guest)
role:[0,1,0,2,1]

Savings: 65-75% reduction for strings with limited unique values

3. Delta Encoding

Best For

Sequential or time-series data

50-76%

reduction

How it works:

Stores differences between consecutive values instead of absolute values.

Example:

JSON

price: [10.00, 10.50, 10.75, 11.25]

AXON

price:Δ10.00[0.50,0.25,0.50]

4. Bit Packing

Best For

Boolean arrays

22.7×

efficiency

How it works:

Compresses boolean values to single bits (8 booleans = 1 byte).

Example:

JSON

active: [true, true, false, true, false, false, true, true]

AXON

active:@bitpack(8):0b11010011

Savings: 95.6% reduction for boolean arrays

5. Varint Encoding

Best For

Integers with variable magnitude

How it works:

Variable-length integer encoding that uses fewer bytes for smaller numbers. Small numbers (0-127) use just 1 byte, while larger numbers use more bytes only when needed.

Example:

id:@varint[1,100,1000000]

The number 1 uses 1 byte, 100 uses 1 byte, 1000000 uses 3 bytes. Much more efficient than fixed-width encoding.

Automatic Algorithm Selection

AXON analyzes your data and chooses the best algorithm(s) automatically:

import { encode } from '@axon-format/core';

// AXON automatically detects patterns and applies optimal compression
const data = {
  users: [
    { id: 1, role: "admin", active: true },
    { id: 2, role: "user", active: true },
    { id: 3, role: "user", active: true }
  ]
};

const axon = encode(data);
// Automatically uses:
// - RLE for repeated "true" values
// - Dictionary for role field (admin/user)
// - Varint for sequential IDs

Next: Learn about Encoding Modes and how AXON adapts to different data structures.