BUG #17573: timestamptz casting precision goes down as the year goes up

Started by PG Bug reporting formover 3 years ago2 messagesbugs
Jump to latest
#1PG Bug reporting form
noreply@postgresql.org

The following bug has been logged on the website:

Bug reference: 17573
Logged by: Luc Lavoie
Email address: luc.lavoie@usherbrooke.ca
PostgreSQL version: 14.4
Operating system: macOS 12.5
Description:

timestamptz casting on year 2022 values returns a precision of one
microsecond (as it is expected).
As the year goes up, the precision goes down as the script below will show
for years 2022, 23621, 2713, 4022, and 8022.

select
  'Exact p = 1 microsecond' as "message",
  to_timestamp (extract (epoch from timestamptz '2022-08-01 00:00:00.000000
+00:00')) as "2022-08-01 00:00:00.000000",
  to_timestamp (extract (epoch from timestamptz '2022-08-01 00:00:00.000001
+00:00')) as "2022-08-01 00:00:00.000001",
  to_timestamp (extract (epoch from timestamptz '2022-08-01 00:00:00.000002
+00:00')) as "2022-08-01 00:00:00.000002",
  to_timestamp (extract (epoch from timestamptz '2022-08-01 00:00:00.000003
+00:00')) as "2022-08-01 00:00:00.000003" ;
select
  'Error p ~ 2 microseconds' as "message",
  to_timestamp (extract (epoch from timestamptz '2361-08-01 00:00:00.000000
+00:00')) as "2361-08-01 00:00:00.000000",
  to_timestamp (extract (epoch from timestamptz '2361-08-01 00:00:00.000001
+00:00')) as "2361-08-01 00:00:00.000001",
  to_timestamp (extract (epoch from timestamptz '2361-08-01 00:00:00.000002
+00:00')) as "2361-08-01 00:00:00.000002",
  to_timestamp (extract (epoch from timestamptz '2361-08-01 00:00:00.000003
+00:00')) as "2361-08-01 00:00:00.000003";
select
  'Error p ~ 4 microseconds' as "message",
  to_timestamp (extract (epoch from timestamptz '2713-08-01 00:00:00.000000
+00:00')) as "2713-08-01 00:00:00.000000",
  to_timestamp (extract (epoch from timestamptz '2713-08-01 00:00:00.000001
+00:00')) as "2713-08-01 00:00:00.000001",
  to_timestamp (extract (epoch from timestamptz '2713-08-01 00:00:00.000002
+00:00')) as "2713-08-01 00:00:00.000002",
  to_timestamp (extract (epoch from timestamptz '2713-08-01 00:00:00.000003
+00:00')) as "2713-08-01 00:00:00.000003";
select
  'Error p ~ 8 microseconds' as "message",
  to_timestamp (extract (epoch from timestamptz '4022-08-01 00:00:00.000000
+00:00')) as "4022-08-01 00:00:00.000000",
  to_timestamp (extract (epoch from timestamptz '4022-08-01 00:00:00.000010
+00:00')) as "4022-08-01 00:00:00.000010",
  to_timestamp (extract (epoch from timestamptz '4022-08-01 00:00:00.000020
+00:00')) as "4022-08-01 00:00:00.000020",
  to_timestamp (extract (epoch from timestamptz '4022-08-01 00:00:00.000030
+00:00')) as "4022-08-01 00:00:00.000030";
select
  'Error p ~ 32 microseconds' as "message",
  to_timestamp (extract (epoch from timestamptz '8022-08-01
00:00:00.000000')) as "8022-08-01 00:00:00.000000",
  to_timestamp (extract (epoch from timestamptz '8022-08-01
00:00:00.000010')) as "8022-08-01 00:00:00.000010",
  to_timestamp (extract (epoch from timestamptz '8022-08-01
00:00:00.000020')) as "8022-08-01 00:00:00.000020",
  to_timestamp (extract (epoch from timestamptz '8022-08-01
00:00:00.000030')) as "8022-08-01 00:00:00.000030";
Results obtained
+-----------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|message                |2022-08-01 00:00:00.000000       |2022-08-01
00:00:00.000001       |2022-08-01 00:00:00.000002       |2022-08-01
00:00:00.000003       |
+-----------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|Exact p = 1 microsecond|2022-08-01 00:00:00.000000 +00:00|2022-08-01
00:00:00.000001 +00:00|2022-08-01 00:00:00.000002 +00:00|2022-08-01
00:00:00.000003 +00:00|
+-----------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|message                 |2361-08-01 00:00:00.000000       |2361-08-01
00:00:00.000001       |2361-08-01 00:00:00.000002       |2361-08-01
00:00:00.000003       |
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|Error p ~ 2 microseconds|2361-08-01 00:00:00.000000 +00:00|2361-08-01
00:00:00.000002 +00:00|2361-08-01 00:00:00.000002 +00:00|2361-08-01
00:00:00.000004 +00:00|
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|message                 |2713-08-01 00:00:00.000000       |2713-08-01
00:00:00.000001       |2713-08-01 00:00:00.000002       |2713-08-01
00:00:00.000003       |
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|Error p ~ 4 microseconds|2713-08-01 00:00:00.000000 +00:00|2713-08-01
00:00:00.000000 +00:00|2713-08-01 00:00:00.000004 +00:00|2713-08-01
00:00:00.000004 +00:00|
+------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|message                  |4022-08-01 00:00:00.000000       |4022-08-01
00:00:00.000010       |4022-08-01 00:00:00.000020       |4022-08-01
00:00:00.000030       |
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|Error p ~ 8 microseconds|4022-08-01 00:00:00.000000 +00:00|4022-08-01
00:00:00.000008 +00:00|4022-08-01 00:00:00.000024 +00:00|4022-08-01
00:00:00.000032 +00:00|
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|message                  |8022-08-01 00:00:00.000000       |8022-08-01
00:00:00.000010       |8022-08-01 00:00:00.000020       |8022-08-01
00:00:00.000030       |
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
|Error p ~ 32 microseconds|8022-08-01 00:00:00.000000 +00:00|8022-08-01
00:00:00.000000 +00:00|8022-08-01 00:00:00.000032 +00:00|8022-08-01
00:00:00.000032 +00:00|
+-------------------------+---------------------------------+---------------------------------+---------------------------------+---------------------------------+
#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: PG Bug reporting form (#1)
Re: BUG #17573: timestamptz casting precision goes down as the year goes up

PG Bug reporting form <noreply@postgresql.org> writes:

timestamptz casting on year 2022 values returns a precision of one
microsecond (as it is expected).
As the year goes up, the precision goes down as the script below will show
for years 2022, 23621, 2713, 4022, and 8022.

The issue here is not with the timestamp, nor even with the extract().
It's with your use of to_timestamp(), which takes float8, which has
limited precision. The precision is disappearing in the implicit
cast to float8.

Prior to v14, there was also an issue with extract() returning float8,
but nowadays it returns numeric. If you look directly at the extract()
results you'll see no precision is lost.

Maybe we should invent a variant of to_timestamp that takes numeric, but
the use-case seems pretty niche ...

regards, tom lane