uuid

For creating, comparing, parsing and printing Universally Unique Identifiers

https://github.com/haskell-hvr/uuid

Version on this page:1.3.15@rev:3
LTS Haskell 22.41:1.3.15@rev:3
Stackage Nightly 2024-11-12:1.3.16@rev:1
Latest on Hackage:1.3.16@rev:1

See all snapshots uuid appears in

BSD-3-Clause licensed by Antoine Latter
Maintained by Oleg Grenrus
This version can be pinned in stack with:uuid-1.3.15@sha256:7d553f95a750502228fc8411a5ad7e672d9b95272e3a61e5c711f19a49143821,2670

Module documentation for 1.3.15

This library is useful for creating, comparing, parsing and printing Universally Unique Identifiers.

See http://en.wikipedia.org/wiki/UUID for the general idea.

Changes

1.3.15

  • Add (Template Haskell) Lift UUID instance

1.3.14

  • Use more compact heap object representation which saves 16 bytes on 64bit platforms.
  • Add toWords64/fromWords64 functions

1.3.13

  • Optimize V4.nextRandom (~3x speed increase)
  • Optimize UUID V3 & V5 generation (~2x speed increase)
  • Use cryptohash-md5/cryptohash-sha1/entropy instead of memory/cryptonite for better performance and stability, but GHCJS is now no longer supported.
  • Update cabal-spec to version 1.10

1.3.12

  • Update package dependencies
  • Use cryptonite for crypto
  • The function V4.nextRandom is now implemented with functions from Crypto.Random in package cryptonite. This does slow-down random UUID generation but provides for far greater randomness. The Random instance for UUID can be used to re-gain the old behavior.

1.3.11

  • Add toText/fromText

1.3.10

  • Update dependencies in tests and benchmarks.

1.3.9

  • Split definition of UUID data type into separate Cabal package to enable 3rd party libraries to avoid some of the larger dependencies.

1.3.8

  • Allow building against newer deepseq.

1.3.7

  • Allow building against newer time.

1.3.6

  • Move to GitHub.

1.3.5

  • Allow building against newer random.

1.3.4

  • Allow building against newer QuickCheck.

1.3.3

  • More complex version constraints on hashable, to avoid building against versions less than 1.1.1.0 or equal to 1.2.0.*.

1.3.2

  • Fix for building against bytestring 0.9.*

1.3.1

  • Allow building against hashable 1.1.* in addition to 1.2.*

1.3.0

  • New functions for parsing and printing UUIDs to and from ASCII BytesStrings
  • New module Data.UUID.Util. This module includes the type UnpackedUUID, whose fields correspond to the UUID fields described in RFC 4122.
  • The Storable instance now stores a UUID in host byte-order instead of big endian.
  • There is now an instance for Hashable UUID.

1.2.13

Benchmark only changes:

  • Allow criterion 0.8.*

1.2.12

Test only changes:

  • Allow QuickCheck 2.6.*

1.2.11

  • Allow binary 0.7.*

1.2.10

  • Allow cryptohash version 0.9.*
  • Cleanup tests

1.2.9

  • Bumped dependency on cryptohash.

1.2.8

  • Bumped various dependencies and cleaned up dependencies in general.

1.2.7

  • Added stricter constraints on random package.

1.2.6

  • Add module V4 to direct attention to our Random instance

  • In module V1 seed the generator with a random number if the hardware MAC address could not be discovered.

  • Fix and cleanup various haddocks.

  • In module docs, warn about MD5 use in Data.UUID.V3 and encourage the reader to use Data.UUID.V5 instead.

1.2.5

  • Use cryptohash package for MD5 and SHA1 instead of Crypto

1.2.4

  • Unpack Word32 values into UUID constructor.

  • Update test suite to QuickCheck 2

  • Bump other dependencies in tests/benchmarks

1.2.3

  • The Read instance now drops leading spaces in the string to be parsed. Thanks to Marc Ziegert for reporting this bug.

  • The tests have moved over to the new Cabal test running framework.

1.2.2

  • Add functions fromWords/toWords

The goal was to have a total function that can be used to construct a UUID, primarily for use by uuid-th.

1.2.1

  • Fix concurrency bug in Data.UUID.V1 (thanks to Neil Mitchell for reporting and a test case)

1.2.0 (Contributors: Antoine Latter & Mark Lentczner)

  • added functions toByteString and fromByteString
  • added nil UUID
  • added unit tests and benchmarks, built when configured -ftest
  • major speed up of to/from functions (as well as in general)
  • added version-3 generation (deterministic based on MD5)
  • major changes to internal representation
    • now uses four strict Word32 values
    • internal ByteSource classes for easy construction (see Builder.hs)
  • Storable instance now stores in memory as system libraries in C do: 16 bytes derived from the network order of the fields, no matter what the host native endianess is.
  • fixed bugs in V1 time and clock stepping, and V1 generated values
  • builds cleanly under GHC’s -Wall
  • added CHANGES file

1.1.1

  • no longer exporting null from the prelude
  • add null predicate on UUIDs
  • documentation fix (thanks Mark Lentczner)