From 25460815ccbfcd6f86d39beddc6c0aa7005fc0a4 Mon Sep 17 00:00:00 2001 From: "Andrey M. Borodin" Date: Sun, 20 Aug 2023 23:55:31 +0300 Subject: [PATCH v9] Implement UUID v7 as per IETF draft This commit addes function to generate UUID v7. This function optionally accepts datetime used to generate next UUID. Also we add a function to extract timestamp from UUID v7. Authors: Andrey Borodin, Sergey Prokhorenko --- doc/src/sgml/func.sgml | 18 ++- src/backend/utils/adt/pseudotypes.c | 12 +- src/backend/utils/adt/uuid.c | 157 +++++++++++++++++++++++ src/include/catalog/pg_proc.dat | 14 ++ src/test/regress/expected/opr_sanity.out | 4 + src/test/regress/expected/uuid.out | 43 +++++++ src/test/regress/sql/uuid.sql | 22 ++++ 7 files changed, 267 insertions(+), 3 deletions(-) diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml index 210c7c0b02..d6c83cb13f 100644 --- a/doc/src/sgml/func.sgml +++ b/doc/src/sgml/func.sgml @@ -14130,13 +14130,29 @@ CREATE TYPE rainbow AS ENUM ('red', 'orange', 'yellow', 'green', 'blue', 'purple gen_random_uuid + + gen_uuid_v7 + + + + uuid_v7_time + + - PostgreSQL includes one function to generate a UUID: + PostgreSQL includes two functions to generate a UUID: gen_random_uuid () uuid This function returns a version 4 (random) UUID. This is the most commonly used type of UUID and is appropriate for most applications. + +gen_uuid_v7 () uuid + + This function returns a version 7 (time-ordered + random) UUID. + +uuid_v7_time (uuid) timestamptz + + This function extracts a timestamptz from UUID version 7. diff --git a/src/backend/utils/adt/pseudotypes.c b/src/backend/utils/adt/pseudotypes.c index a3a991f634..82c2d1309d 100644 --- a/src/backend/utils/adt/pseudotypes.c +++ b/src/backend/utils/adt/pseudotypes.c @@ -23,6 +23,7 @@ #include "postgres.h" #include "libpq/pqformat.h" +#include "miscadmin.h" #include "utils/array.h" #include "utils/builtins.h" #include "utils/rangetypes.h" @@ -332,11 +333,18 @@ shell_out(PG_FUNCTION_ARGS) * * We must disallow input of pg_node_tree values because the SQL functions * that operate on the type are not secure against malformed input. - * We do want to allow output, though. + * We do want to allow output, though. Also we need input during bootstrap. */ -PSEUDOTYPE_DUMMY_INPUT_FUNC(pg_node_tree); PSEUDOTYPE_DUMMY_RECEIVE_FUNC(pg_node_tree); +Datum +pg_node_tree_in(PG_FUNCTION_ARGS) +{ + if (!IsBootstrapProcessingMode()) + elog(ERROR, "cannot accept a value of type pg_node_tree_in"); + return textin(fcinfo); +} + Datum pg_node_tree_out(PG_FUNCTION_ARGS) { diff --git a/src/backend/utils/adt/uuid.c b/src/backend/utils/adt/uuid.c index 73dfd711c7..66d9672dd1 100644 --- a/src/backend/utils/adt/uuid.c +++ b/src/backend/utils/adt/uuid.c @@ -13,6 +13,9 @@ #include "postgres.h" +#include + +#include "access/xlog.h" #include "common/hashfn.h" #include "lib/hyperloglog.h" #include "libpq/pqformat.h" @@ -20,6 +23,7 @@ #include "utils/builtins.h" #include "utils/guc.h" #include "utils/sortsupport.h" +#include "utils/timestamp.h" #include "utils/uuid.h" /* sortsupport for uuid */ @@ -421,3 +425,156 @@ gen_random_uuid(PG_FUNCTION_ARGS) PG_RETURN_UUID_P(uuid); } + +static uint32_t sequence_counter; +static uint64_t previous_timestamp = 0; + + +Datum +gen_uuid_v7(PG_FUNCTION_ARGS) +{ + pg_uuid_t *uuid = palloc(UUID_LEN); + TimestampTz ts; + uint64_t tms; + struct timeval tp; + bool increment_counter; + + if (PG_NARGS() == 0 || PG_ARGISNULL(0)) + { + gettimeofday(&tp, NULL); + tms = ((uint64_t)tp.tv_sec) * 1000 + (tp.tv_usec) / 1000; + /* time from clock is protected from backward leaps */ + increment_counter = tms <= previous_timestamp; + } + else + { + ts = PG_GETARG_TIMESTAMPTZ(0); + tms = (ts + (POSTGRES_EPOCH_JDATE - UNIX_EPOCH_JDATE) * SECS_PER_DAY * USECS_PER_SEC) / 1000; + /* + * The time can leap backwards when provided by the user, so we use + * counter only when called with exactly same unix_ts_ms argument. + */ + increment_counter = (tms == previous_timestamp); + } + + if (increment_counter) + { + /* Time did not increment from the previous generation, we must increment counter */ + ++sequence_counter; + if (sequence_counter > 0x3ffff) + { + /* We only have 18-bit counter */ + sequence_counter = 0; + previous_timestamp++; + } + + /* protection from leap backward */ + tms = previous_timestamp; + + /* fill everything after the timestamp and counter with random bytes */ + if (!pg_strong_random(&uuid->data[8], UUID_LEN - 8)) + ereport(ERROR, + (errcode(ERRCODE_INTERNAL_ERROR), + errmsg("could not generate random values"))); + + /* most significant 4 bits of 18-bit counter */ + uuid->data[6] = (unsigned char)(sequence_counter >> 14); + /* next 8 bits */ + uuid->data[7] = (unsigned char)(sequence_counter >> 6); + /* least significant 6 bits */ + uuid->data[8] = (unsigned char)(sequence_counter); + } + else + { + /* fill everything after the timestamp with random bytes */ + if (!pg_strong_random(&uuid->data[6], UUID_LEN - 6)) + ereport(ERROR, + (errcode(ERRCODE_INTERNAL_ERROR), + errmsg("could not generate random values"))); + + /* + * Left-most counter bits are initialized as zero for the sole purpose + * of guarding against counter rollovers. + * See section "Fixed-Length Dedicated Counter Seeding" + * https://datatracker.ietf.org/doc/html/draft-ietf-uuidrev-rfc4122bis-09#monotonicity_counters + */ + uuid->data[6] = (uuid->data[6] & 0xf7); + + sequence_counter = ((uint32_t)uuid->data[8] & 0x3f) + + (((uint32_t)uuid->data[7]) << 6) + + (((uint32_t)uuid->data[6] & 0x0f) << 14); + + previous_timestamp = tms; + } + + /* Fill in time part */ + uuid->data[0] = (unsigned char)(tms >> 40); + uuid->data[1] = (unsigned char)(tms >> 32); + uuid->data[2] = (unsigned char)(tms >> 24); + uuid->data[3] = (unsigned char)(tms >> 16); + uuid->data[4] = (unsigned char)(tms >> 8); + uuid->data[5] = (unsigned char)tms; + + /* + * Set magic numbers for a "version 7" (pseudorandom) UUID, see + * http://tools.ietf.org/html/rfc ??? + * https://datatracker.ietf.org/doc/html/draft-peabody-dispatch-new-uuid-format#name-creating-a-uuidv7-value + */ + /* set version field, top four bits are 0, 1, 1, 1 */ + uuid->data[6] = (uuid->data[6] & 0x0f) | 0x70; + /* set variant field, top two bits are 1, 0 */ + uuid->data[8] = (uuid->data[8] & 0x3f) | 0x80; + + PG_RETURN_UUID_P(uuid); +} + +Datum +uuid_v7_time(PG_FUNCTION_ARGS) +{ + pg_uuid_t *uuid = PG_GETARG_UUID_P(0); + TimestampTz ts; + uint64_t tms; + + if (((uuid->data[6] & 0xf0) != 0x70) + || ((uuid->data[8] & 0xc0) != 0x80)) + elog(ERROR,"uuid_v7_time() can only extract timestamp from UUID v7"); + + tms = uuid->data[5]; + tms += ((uint64_t)uuid->data[4]) << 8; + tms += ((uint64_t)uuid->data[3]) << 16; + tms += ((uint64_t)uuid->data[2]) << 24; + tms += ((uint64_t)uuid->data[1]) << 32; + tms += ((uint64_t)uuid->data[0]) << 40; + + ts = (TimestampTz) (tms * 1000) - /* convert ms to us, than adjust */ + (POSTGRES_EPOCH_JDATE - UNIX_EPOCH_JDATE) * SECS_PER_DAY * USECS_PER_SEC; + + PG_RETURN_TIMESTAMPTZ(ts); +} + +Datum +uuid_ver(PG_FUNCTION_ARGS) +{ + pg_uuid_t *uuid = PG_GETARG_UUID_P(0); + TimestampTz ts; + uint64_t tms; + uint16_t result; + + if ((uuid->data[8] & 0xc0) != 0x80) + elog(ERROR,"uuid_ver() is only defined for RFC 4122 variants"); + result = uuid->data[6] >> 4; + + PG_RETURN_UINT16(result); +} + +Datum +uuid_var(PG_FUNCTION_ARGS) +{ + pg_uuid_t *uuid = PG_GETARG_UUID_P(0); + TimestampTz ts; + uint64_t tms; + uint16_t result; + result = uuid->data[8] >> 6; + + PG_RETURN_UINT16(result); +} \ No newline at end of file diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index 58811a6530..c00cd1320f 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -9174,6 +9174,20 @@ { oid => '3432', descr => 'generate random UUID', proname => 'gen_random_uuid', proleakproof => 't', provolatile => 'v', prorettype => 'uuid', proargtypes => '', prosrc => 'gen_random_uuid' }, +{ oid => '9895', descr => 'generate UUID version 7', proisstrict => 'f', + proname => 'gen_uuid_v7', proleakproof => 't', provolatile => 'v', + prorettype => 'uuid', proargtypes => 'timestamptz', prosrc => 'gen_uuid_v7', + proargnames => '{unix_ts_ms}', pronargdefaults => 1, proargmodes => '{i}', + proargdefaults => '({CONST :consttype 1184 :consttypmod -1 :constcollid 0 :constlen 8 :constbyval true :constisnull true :location 46 :constvalue <>})' }, +{ oid => '9896', descr => 'extract timestamp from UUID version 7', + proname => 'uuid_v7_time', proleakproof => 't', provolatile => 'i', + prorettype => 'timestamptz', proargtypes => 'uuid', prosrc => 'uuid_v7_time' }, +{ oid => '9897', descr => 'extract version from RFC 4122 UUID', + proname => 'uuid_ver', proleakproof => 't', provolatile => 'i', + prorettype => 'int2', proargtypes => 'uuid', prosrc => 'uuid_ver' }, +{ oid => '9898', descr => 'extract variant from UUID', + proname => 'uuid_var', proleakproof => 't', provolatile => 'i', + prorettype => 'int2', proargtypes => 'uuid', prosrc => 'uuid_var' }, # pg_lsn { oid => '3229', descr => 'I/O', diff --git a/src/test/regress/expected/opr_sanity.out b/src/test/regress/expected/opr_sanity.out index 7610b011d6..163658f002 100644 --- a/src/test/regress/expected/opr_sanity.out +++ b/src/test/regress/expected/opr_sanity.out @@ -872,6 +872,10 @@ xid8ge(xid8,xid8) xid8eq(xid8,xid8) xid8ne(xid8,xid8) xid8cmp(xid8,xid8) +gen_uuid_v7(timestamp with time zone) +uuid_v7_time(uuid) +uuid_ver(uuid) +uuid_var(uuid) -- restore normal output mode \a\t -- List of functions used by libpq's fe-lobj.c diff --git a/src/test/regress/expected/uuid.out b/src/test/regress/expected/uuid.out index 8e7f21910d..b8426b1f8e 100644 --- a/src/test/regress/expected/uuid.out +++ b/src/test/regress/expected/uuid.out @@ -168,5 +168,48 @@ SELECT count(DISTINCT guid_field) FROM guid1; 2 (1 row) +-- generation test for v7 +TRUNCATE guid1; +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7()); +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7()); +SELECT count(DISTINCT guid_field) FROM guid1; + count +------- + 2 +(1 row) + +-- generation test for v7 with same unix_ts_ms +TRUNCATE guid1; +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7(now())); +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7(now())); +SELECT count(DISTINCT guid_field) FROM guid1; + count +------- + 2 +(1 row) + +-- check that timestamp is extracted correctly +SELECT uuid_v7_time(gen_uuid_v7(TIMESTAMP '2024-01-16 13:37:00')) - TIMESTAMP '2024-01-16 13:37:00'; + ?column? +---------- + @ 0 +(1 row) + +-- support functions for UUID versions and variants +SELECT uuid_ver(gen_uuid_v7()); + uuid_ver +---------- + 7 +(1 row) + +SELECT uuid_var(gen_uuid_v7()); + uuid_var +---------- + 2 +(1 row) + +-- uuid_v7_time() must refuse to accept non-UUIDv7 +select uuid_v7_time(gen_random_uuid()); +ERROR: uuid_v7_time() can only extract timestamp from UUID v7 -- clean up DROP TABLE guid1, guid2 CASCADE; diff --git a/src/test/regress/sql/uuid.sql b/src/test/regress/sql/uuid.sql index 9a8f437c7d..fb28766ece 100644 --- a/src/test/regress/sql/uuid.sql +++ b/src/test/regress/sql/uuid.sql @@ -85,5 +85,27 @@ INSERT INTO guid1 (guid_field) VALUES (gen_random_uuid()); INSERT INTO guid1 (guid_field) VALUES (gen_random_uuid()); SELECT count(DISTINCT guid_field) FROM guid1; +-- generation test for v7 +TRUNCATE guid1; +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7()); +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7()); +SELECT count(DISTINCT guid_field) FROM guid1; + +-- generation test for v7 with same unix_ts_ms +TRUNCATE guid1; +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7(now())); +INSERT INTO guid1 (guid_field) VALUES (gen_uuid_v7(now())); +SELECT count(DISTINCT guid_field) FROM guid1; + +-- check that timestamp is extracted correctly +SELECT uuid_v7_time(gen_uuid_v7(TIMESTAMP '2024-01-16 13:37:00')) - TIMESTAMP '2024-01-16 13:37:00'; + +-- support functions for UUID versions and variants +SELECT uuid_ver(gen_uuid_v7()); +SELECT uuid_var(gen_uuid_v7()); + +-- uuid_v7_time() must refuse to accept non-UUIDv7 +select uuid_v7_time(gen_random_uuid()); + -- clean up DROP TABLE guid1, guid2 CASCADE; -- 2.37.1 (Apple Git-137.1)