Puid

Simple, fast, flexible and efficient generation of probably unique identifiers (puid, aka random strings) of intuitively specified entropy using pre-defined or custom characters.

  iex> defmodule(RandId, do: use(Puid, chars: :alpha, total: 1.0e5, risk: 1.0e12))
  iex> RandId.generate()
  "YAwrpLRqXGlny"

Hex Version   License: MIT

<a name="TOC"></a>TOC

<a name="Overview"></a>Overview

A general overview of PUID provides information relevant to all PUID implementations.

TOC

<a name="Usage"></a>Usage

Creating a random ID generator using Puid is a simple as:

  iex> defmodule(RandId, do: use(Puid))
  iex> RandId.generate()
  "8nGA2UaIfaawX-Og61go5A"

Options allow easy and complete control of ID generation.

Entropy Source

Puid uses :crypto.strong_rand_bytes/1 as the default entropy source. The rand_bytes option can be used to specify any function of the form (non_neg_integer) -> binary as the source:

  iex > defmodule(PrngPuid, do: use(Puid, rand_bytes: &:rand.bytes/1))
  iex> PrngPuid.generate()
  "bIkrSeU6Yr8_1WHGvO0H3M"

Characters

By default, Puid use the RFC 4648 file system & URL safe characters. The chars option can by used to specify any of 16 pre-defined character sets or custom characters, including Unicode:

  iex> defmodule(HexPuid, do: use(Puid, chars: :hex))
  iex> HexPuid.generate()
  "13fb81e35cb89e5daa5649802ad4bbbd"

  iex> defmodule(DingoskyPuid, do: use(Puid, chars: "dingosky"))
  iex> DingoskyPuid.generate()
  "yiidgidnygkgydkodggysonydodndsnkgksgonisnko"

  iex> defmodule(DingoskyUnicodePuid, do: use(Puid, chars: "dîñgø$kyDÎÑGØßK¥", total: 2.5e6, risk: 1.0e15))
  iex> DingoskyUnicodePuid.generate()
  "øßK$ggKñø$dyGîñdyØøØÎîk"

Captured Entropy

Generated IDs have 128-bit entropy by default. Puid provides a simple, intuitive way to specify ID randomness by declaring a total number of possible IDs with a specified risk of a repeat in that many IDs:

To generate up to 10 million random IDs with 1 in a trillion chance of repeat:

  iex> defmodule(MyPuid, do: use(Puid, total: 10.0e6, risk: 1.0e15))
  iex> MyPuid.generate()
  "T0bFZadxBYVKs5lA"

The bits option can be used to directly specify an amount of ID randomness:

  iex> defmodule(Token, do: use(Puid, bits: 256, chars: :hex_upper))
  iex> Token.generate()
  "6E908C2A1AA7BF101E7041338D43B87266AFA73734F423B6C3C3A17599F40F2A"

TOC

<a name="Installation"></a>Installation

Add puid to mix.exs dependencies:

def deps,
  do: [
    {:puid, "~> 2.1"}
  ]

Update dependencies

mix deps.get

<a name="ModuleAPI"></a>Module API

Puid modules have two functions:

generate/0

Generates a puidinfo/0

Returns a Puid.Info structure consisting of

<a name="Chars"></a>Chars

There are 16 pre-defined character sets:

Name Characters
:alpha ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz
:alpha_lower abcdefghijklmnopqrstuvwxyz
:alpha_upper ABCDEFGHIJKLMNOPQRSTUVWXYZ
:alphanum ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789
:alphanum_lower abcdefghijklmnopqrstuvwxyz0123456789
:alphanum_upper ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789
:base32 ABCDEFGHIJKLMNOPQRSTUVWXYZ234567
:base32_hex 0123456789abcdefghijklmnopqrstuv
:base32_hex_upper 0123456789ABCDEFGHIJKLMNOPQRSTUV
:decimal 0123456789
:hex 0123456789abcdef
:hex_upper 0123456789ABCDEF
:safe_ascii !#$%&()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[]^_abcdefghijklmnopqrstuvwxyz{|}~
:safe32 2346789bdfghjmnpqrtBDFGHJLMNPQRT
:safe64 ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_
:symbol !#$%&()*+,-./:;<=>?@[]^_{|}~

Any String or charlist of up to 256 unique characters can be used for puid generation. Custom characters set are optimized in the same manner as the pre-defined character sets.

TOC

<a name="Comparisons"></a>Comparisons

As described in the overview, PUID aims to be a general, flexible mechanism for creating random string for use as random IDs. The following comparisons to other Elixir random ID generators is with respect to the issues of random ID generation described in that overview.

TOC

<a name="Common_Solution"></a>Common Solution

Comments

Timing

PUID is much faster.

Generate 100000 random IDs with 128 bits of entropy using alphanumeric characters

  Common Solution   (PRNG) : 4.977226
  Puid              (PRNG) : 0.831748

  Common Solution (CSPRNG) : 8.435073
  Puid            (CSPRNG) : 0.958437

TOC

<a name="misc_random"></a>misc_random

Comments

Timing

Quite slow compared to PUID

Generate 50000 random IDs with 128 bits of entropy using alphanum characters

  Misc.Random (PRNG) : 12.196646
  Puid        (PRNG) : 0.295741

  Misc.Random (CSPRNG) : 11.9858
  Puid        (CSPRNG) : 0.310417

TOC

<a name="nanoid"></a>nanoid

Comments:

Timing:

nanoid is much slower than PUID

  Generate 100000 random IDs with 126 bits of entropy using safe64 characters

    Nanoid (CSPRNG) : 8.480194
    Puid   (CSPRNG) : 0.353484

    Nanoid (PRNG) : 1.603285
    Puid   (PRNG) : 0.425961

  Generate 100000 random IDs with 195 bits of entropy using safe32 characters

    Nanoid (CSPRNG) : 6.117834
    Puid   (CSPRNG) : 0.366509

TOC

<a name="Randomizer"></a>Randomizer

Comments

Timing

Slower than PUID

Generate 100000 random IDs with 128 bits of entropy using alphanum characters

  Randomizer   (PRNG) : 1.201281
  Puid         (PRNG) : 0.829199

  Randomizer (CSPRNG) : 4.329881
  Puid       (CSPRNG) : 0.807226

TOC

<a name="SecureRandom"></a>SecureRandom

Comments

Timing

About the same as PUID when using CSPRNG

Generate 500000 random IDs with 128 bits of entropy using hex characters

  SecureRandom (CSPRNG) : 1.19713
  Puid         (CSPRNG) : 1.187726

Generate 500000 random IDs with 128 bits of entropy using safe64 characters

  SecureRandom (CSPRNG) : 2.103798
  Puid         (CSPRNG) : 1.806514

TOC

<a name="ulid"></a>ulid

Comments

A significant characteristic of ulid is the generation of lexicographically sortable IDs. This is not a goal for PUID; however, one could use PUID to generate such IDs by prefixing a timestamp to a generated puid. Such a solution would be similar to ulid while still providing full control to entropy source, characters, and captured entropy per timestamp context.

Timing

ulid and PUID are not directly comparable with regard to speed.

TOC

<a name="UUID"></a>UUID

Comments

Timing

Similar to PUID when using CSPRNG

Generate 500000 random IDs with 122 bits of entropy using hex
  UUID     : 1.925131
  Puid hex : 1.823116

Generate 500000 random IDs with 122 bits of entropy using safe64
  UUID        : 1.751625
  Puid safe64 : 1.367201

TOC