now() vs transaction_timestamp()

Konstantin Knizhnik <k.knizhnik@postgrespro.ru>

From: Konstantin Knizhnik <k.knizhnik@postgrespro.ru>
To: PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2018-10-05T13:29:28Z
Lists: pgsql-hackers
Postgres documentation says that |"now()| is a traditional PostgreSQL 
equivalent to |transaction_timestamp()|".
Also both use the same implementation.
But them have different parallel safety property:

postgres=# \df+ now
List of functions
    Schema   | Name |     Result data type     | Argument data types | 
Type | Volatility |  Parallel  |  Owner   | Security | Access privileges 
| Language |
  Source code |       Description
------------+------+--------------------------+---------------------+------+------------+------------+----------+----------+-------------------+----------+
-------------+--------------------------
  pg_catalog | now  | timestamp with time zone | | func | stable     | 
restricted | knizhnik | invoker |                   | internal |
  now         | current transaction time
(1 row)

postgres=# \df+ transaction_timestamp
List of functions
    Schema   |         Name          |     Result data type     | 
Argument data types | Type | Volatility | Parallel |  Owner   | Security 
| Access privileg
es | Language | Source code |       Description
------------+-----------------------+--------------------------+---------------------+------+------------+----------+----------+----------+----------------
---+----------+-------------+--------------------------
  pg_catalog | transaction_timestamp | timestamp with time zone 
|                     | func | stable     | safe     | knizhnik | 
invoker  |
    | internal | now         | current transaction time
(1 row)

As a result using now() in query disable parallel execution while 
transaction_timestamp allows it.
Was it done intentionally or it is just a bug?

-- 
Konstantin Knizhnik
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company

Commits

  1. Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers.