Tokens

Convert sequential database IDs into compact, non-enumerable, type-prefixed tokens.

User ID 1  → U_71VHDH
User ID 2  → U_1MXBJ1
User ID 3  → U_ZR5KA5

Each token type has its own prefix, length, and seed, so the same integer ID produces different tokens for different entity types:

UserToken(1).value     // "U_71VHDH"     (6-char)
InvoiceToken(1).value  // "INV_0RXG6MSN" (8-char)

Tokens are reversible, you can always recover the original ID:

UserToken(42).toId()  // 42

How it works

1. A Feistel cipher permutes integers reversibly within a fixed domain, so that sequential inputs produce unique, unrelated outputs.

2. A Base32 encoder encodes each integer as a short, human-friendly string using Crockford's Base32 (minus ambiguous characters like I, L, O). The encoded length is configurable per token type (1-12 characters).

3. A typed token wrapper pairs a type prefix (e.g. U_, INV_) with an encoded ID, so tokens are self-describing and can't be mixed across entity types.

Input canonicalisation

Tokens are permissive of common transcription errors. When constructed from a string, confusable characters are automatically replaced with their canonical equivalents following Crockford's Base32 rules, and lowercase is normalised to uppercase:

Input	Canonical
O, o	0
I, i, L, l	1
U, u	V
a–z	A–Z

UserToken("U_abcol2").value  // "U_ABC012"
UserToken("U_abcol2").toId() == UserToken("U_ABC012").toId()  // true

Defining new token types

class ProjectToken : Token {
    constructor(value: String) : super(PREFIX, LENGTH, SEED, value)
    constructor(id: Long) : super(PREFIX, LENGTH, SEED, id)

    companion object {
        const val PREFIX = "P"
        const val LENGTH = 6
        const val SEED = 0xYOUR_SEED_HEREL
    }
}

Use a different seed for each token type to ensure their output spaces don't collide. Length can be between 1 and 12 characters.

Running tests

./gradlew test

Sources

• Base32 (crockford.com)
• Feistel cipher (asecuritysite.com)