FW: BUG #17258: Unexpected results in CHAR(1) data type

-----Original Message-----
From: David M. Calascibetta <david@calascibetta.com>
Sent: Friday, October 29, 2021 2:04 PM
To: 'Mark Dilger' <mark.dilger@enterprisedb.com>
Subject: RE: BUG #17258: Unexpected results in CHAR(1) data type

Ok, but my example was just a simplified version of what is going on.
The actual problem stems from a CHAR(1) column data type that is behaving the same way.
I was just trying to create a super-simple example of the problem.
It still seems to me that a CHAR(1) should never be zero length, regardless of how it's implemented.


-----Original Message-----
From: Mark Dilger <mark.dilger@enterprisedb.com>
Sent: Friday, October 29, 2021 1:58 PM
To: david@calascibetta.com; PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Subject: Re: BUG #17258: Unexpected results in CHAR(1) data type



> On Oct 29, 2021, at 11:14 AM, PG Bug reporting form <noreply@postgresql.org> wrote:
>
> select length(substr('   '::varchar,1,1)), ascii(substr('
> '::varchar,1,1));
> produces==> 1    32   (as expected)
>
> select length(substr('   '::char,1,1)), ascii(substr('   '::char,1,1));
> produces==> 0     0     (not as expected)

There are four substr() functions defined in the standard distribution.  They are:

  substr(text, int4, int4)
  substr(text, int4)
  substr(bytea, int4, int4)
  substr(bytea, int4)

There is none defined directly for char(n).  Since "text" is the preferred type (not "bytea"), I expect a cast of your
''::char to text before entry to the function.  So the question is why casting ' '::char to text is different than '
'::text. The answer is that cast is implemented using rtrim1, which trims trailing space. 

—
Mark Dilger
EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company