зеркало из https://github.com/github/ruby.git
[ruby/securerandom] Add support for UUID version 7
Although the specification for UUIDv7 is still in draft, the UUIDv7 algorithm has been relatively stable as it progresses to completion. Version 7 UUIDs can be very useful, because they are lexographically sortable, which can improve e.g: database index locality. See section 6.10 of the draft specification for further explanation: https://datatracker.ietf.org/doc/draft-ietf-uuidrev-rfc4122bis/ The specification allows up to 12 bits of extra timestamp precision, to make UUID generation closer to monotonically increasing. This provides between 1ms and ~240ns of timestamp precision. At the cost of some code complexity and a small performance penalty, a kwarg may specify any arbitrary precision between 0 and 12 extra bits. Any stronger guarantees of monotonicity have considerably larger tradeoffs, so nothing more is implemented. This limitation is documented. Ruby issue: https://bugs.ruby-lang.org/issues/19735 https://github.com/ruby/securerandom/commit/34ed1a2ec3
This commit is contained in:
Родитель
e77c766b7a
Коммит
dfb2b4cbc9
|
@ -174,6 +174,125 @@ module Random::Formatter
|
|||
"%08x-%04x-%04x-%04x-%04x%08x" % ary
|
||||
end
|
||||
|
||||
alias uuid_v4 uuid
|
||||
|
||||
# Generate a random v7 UUID (Universally Unique IDentifier).
|
||||
#
|
||||
# require 'random/formatter'
|
||||
#
|
||||
# Random.uuid_v7 # => "0188d4c3-1311-7f96-85c7-242a7aa58f1e"
|
||||
# Random.uuid_v7 # => "0188d4c3-16fe-744f-86af-38fa04c62bb5"
|
||||
# Random.uuid_v7 # => "0188d4c3-1af8-764f-b049-c204ce0afa23"
|
||||
# Random.uuid_v7 # => "0188d4c3-1e74-7085-b14f-ef6415dc6f31"
|
||||
# # |<--sorted-->| |<----- random ---->|
|
||||
#
|
||||
# # or
|
||||
# prng = Random.new
|
||||
# prng.uuid_v7 # => "0188ca51-5e72-7950-a11d-def7ff977c98"
|
||||
#
|
||||
# The version 7 UUID starts with the least significant 48 bits of a 64 bit
|
||||
# Unix timestamp (milliseconds since the epoch) and fills the remaining bits
|
||||
# with random data, excluding the version and variant bits.
|
||||
#
|
||||
# This allows version 7 UUIDs to be sorted by creation time. Time ordered
|
||||
# UUIDs can be used for better database index locality of newly inserted
|
||||
# records, which may have a significant performance benefit compared to random
|
||||
# data inserts.
|
||||
#
|
||||
# The result contains 74 random bits (9.25 random bytes).
|
||||
#
|
||||
# Note that this method cannot be made reproducable with Kernel#srand, which
|
||||
# can only affect the random bits. The sorted bits will still be based on
|
||||
# Process.clock_gettime.
|
||||
#
|
||||
# See draft-ietf-uuidrev-rfc4122bis[https://datatracker.ietf.org/doc/draft-ietf-uuidrev-rfc4122bis/]
|
||||
# for details of UUIDv7.
|
||||
#
|
||||
# ==== Monotonicity
|
||||
#
|
||||
# UUIDv7 has millisecond precision by default, so multiple UUIDs created
|
||||
# within the same millisecond are not issued in monotonically increasing
|
||||
# order. To create UUIDs that are time-ordered with sub-millisecond
|
||||
# precision, up to 12 bits of additional timestamp may added with
|
||||
# +extra_timestamp_bits+. The extra timestamp precision comes at the expense
|
||||
# of random bits. Setting <tt>extra_timestamp_bits: 12</tt> provides ~244ns
|
||||
# of precision, but only 62 random bits (7.75 random bytes).
|
||||
#
|
||||
# prng = Random.new
|
||||
# Array.new(4) { prng.uuid_v7(extra_timestamp_bits: 12) }
|
||||
# # =>
|
||||
# ["0188d4c7-13da-74f9-8b53-22a786ffdd5a",
|
||||
# "0188d4c7-13da-753b-83a5-7fb9b2afaeea",
|
||||
# "0188d4c7-13da-754a-88ea-ac0baeedd8db",
|
||||
# "0188d4c7-13da-7557-83e1-7cad9cda0d8d"]
|
||||
# # |<--- sorted --->| |<-- random --->|
|
||||
#
|
||||
# Array.new(4) { prng.uuid_v7(extra_timestamp_bits: 8) }
|
||||
# # =>
|
||||
# ["0188d4c7-3333-7a95-850a-de6edb858f7e",
|
||||
# "0188d4c7-3333-7ae8-842e-bc3a8b7d0cf9", # <- out of order
|
||||
# "0188d4c7-3333-7ae2-995a-9f135dc44ead", # <- out of order
|
||||
# "0188d4c7-3333-7af9-87c3-8f612edac82e"]
|
||||
# # |<--- sorted -->||<---- random --->|
|
||||
#
|
||||
# Any rollbacks of the system clock will break monotonicity. UUIDv7 is based
|
||||
# on UTC, which excludes leap seconds and can rollback the clock. To avoid
|
||||
# this, the system clock can synchronize with an NTP server configured to use
|
||||
# a "leap smear" approach. NTP or PTP will also be needed to synchronize
|
||||
# across distributed nodes.
|
||||
#
|
||||
# Counters and other mechanisms for stronger guarantees of monotonicity are
|
||||
# not implemented. Applications with stricter requirements should follow
|
||||
# {Section 6.2}[https://www.ietf.org/archive/id/draft-ietf-uuidrev-rfc4122bis-07.html#monotonicity_counters]
|
||||
# of the specification.
|
||||
#
|
||||
def uuid_v7(extra_timestamp_bits: 0)
|
||||
case (extra_timestamp_bits = Integer(extra_timestamp_bits))
|
||||
when 0 # min timestamp precision
|
||||
ms = Process.clock_gettime(Process::CLOCK_REALTIME, :millisecond)
|
||||
rand = random_bytes(10)
|
||||
rand.setbyte(0, rand.getbyte(0) & 0x0f | 0x70) # version
|
||||
rand.setbyte(2, rand.getbyte(2) & 0x3f | 0x80) # variant
|
||||
"%08x-%04x-%s" % [
|
||||
(ms & 0x0000_ffff_ffff_0000) >> 16,
|
||||
(ms & 0x0000_0000_0000_ffff),
|
||||
rand.unpack("H4H4H12").join("-")
|
||||
]
|
||||
|
||||
when 12 # max timestamp precision
|
||||
ms, ns = Process.clock_gettime(Process::CLOCK_REALTIME, :nanosecond)
|
||||
.divmod(1_000_000)
|
||||
extra_bits = ns * 4096 / 1_000_000
|
||||
rand = random_bytes(8)
|
||||
rand.setbyte(0, rand.getbyte(0) & 0x3f | 0x80) # variant
|
||||
"%08x-%04x-7%03x-%s" % [
|
||||
(ms & 0x0000_ffff_ffff_0000) >> 16,
|
||||
(ms & 0x0000_0000_0000_ffff),
|
||||
extra_bits,
|
||||
rand.unpack("H4H12").join("-")
|
||||
]
|
||||
|
||||
when (0..12) # the generic version is slower than the special cases above
|
||||
rand_a, rand_b1, rand_b2, rand_b3 = random_bytes(10).unpack("nnnN")
|
||||
rand_mask_bits = 12 - extra_timestamp_bits
|
||||
ms, ns = Process.clock_gettime(Process::CLOCK_REALTIME, :nanosecond)
|
||||
.divmod(1_000_000)
|
||||
"%08x-%04x-%04x-%04x-%04x%08x" % [
|
||||
(ms & 0x0000_ffff_ffff_0000) >> 16,
|
||||
(ms & 0x0000_0000_0000_ffff),
|
||||
0x7000 |
|
||||
((ns * (1 << extra_timestamp_bits) / 1_000_000) << rand_mask_bits) |
|
||||
rand_a & ((1 << rand_mask_bits) - 1),
|
||||
0x8000 | (rand_b1 & 0x3fff),
|
||||
rand_b2,
|
||||
rand_b3
|
||||
]
|
||||
|
||||
else
|
||||
raise ArgumentError, "extra_timestamp_bits must be in 0..12"
|
||||
end
|
||||
end
|
||||
|
||||
private def gen_random(n)
|
||||
self.bytes(n)
|
||||
end
|
||||
|
|
|
@ -75,6 +75,54 @@ module Random::Formatter
|
|||
assert_match(/\A\h{8}-\h{4}-\h{4}-\h{4}-\h{12}\z/, uuid)
|
||||
end
|
||||
|
||||
def test_uuid_v7(extra_timestamp_bits)
|
||||
t1 = current_uuid7_time
|
||||
uuid = @it.uuid_v7
|
||||
t3 = current_uuid7_time
|
||||
|
||||
assert_match(/\A\h{8}-\h{4}-7\h{3}-[89ab]\h{3}-\h{12}\z/, uuid)
|
||||
|
||||
t2 = get_uuid7_time(uuid)
|
||||
assert_operator(t1, :<=, t2)
|
||||
assert_operator(t2, :<=, t3)
|
||||
end
|
||||
|
||||
def test_uuid_v7_extra_timestamp_bits
|
||||
0.upto(12) do |extra_timestamp_bits|
|
||||
t1 = current_uuid7_time extra_timestamp_bits: extra_timestamp_bits
|
||||
uuid = @it.uuid_v7 extra_timestamp_bits: extra_timestamp_bits
|
||||
t3 = current_uuid7_time extra_timestamp_bits: extra_timestamp_bits
|
||||
|
||||
assert_match(/\A\h{8}-\h{4}-7\h{3}-[89ab]\h{3}-\h{12}\z/, uuid)
|
||||
|
||||
t2 = get_uuid7_time uuid, extra_timestamp_bits: extra_timestamp_bits
|
||||
assert_operator(t1, :<=, t2)
|
||||
assert_operator(t2, :<=, t3)
|
||||
end
|
||||
end
|
||||
|
||||
# It would be nice to simply use Time#floor here. But that is problematic
|
||||
# due to the difference between decimal vs binary fractions.
|
||||
def current_uuid7_time(extra_timestamp_bits: 0)
|
||||
denominator = (1 << extra_timestamp_bits).to_r
|
||||
Process.clock_gettime(Process::CLOCK_REALTIME, :nanosecond)
|
||||
.then {|ns| ((ns / 1_000_000r) * denominator).floor / denominator }
|
||||
.then {|ms| Time.at(ms / 1000r, in: "+00:00") }
|
||||
end
|
||||
|
||||
def get_uuid7_time(uuid, extra_timestamp_bits: 0)
|
||||
denominator = (1 << extra_timestamp_bits) * 1000r
|
||||
extra_chars = extra_timestamp_bits / 4
|
||||
last_char_bits = extra_timestamp_bits % 4
|
||||
extra_chars += 1 if last_char_bits != 0
|
||||
timestamp_re = /\A(\h{8})-(\h{4})-7(\h{#{extra_chars}})/
|
||||
timestamp_chars = uuid.match(timestamp_re).captures.join
|
||||
timestamp = timestamp_chars.to_i(16)
|
||||
timestamp >>= 4 - last_char_bits unless last_char_bits == 0
|
||||
timestamp /= denominator
|
||||
Time.at timestamp, in: "+00:00"
|
||||
end
|
||||
|
||||
def test_alphanumeric
|
||||
65.times do |n|
|
||||
an = @it.alphanumeric(n)
|
||||
|
|
Загрузка…
Ссылка в новой задаче