Skip site navigation (1) Skip section navigation (2)

Re: How well does PostgreSQL 9.6.1 support unicode?

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Vick Khera <vivek(at)khera(dot)org>
Cc: pgsql-general <pgsql-general(at)postgresql(dot)org>, Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com>
Subject: Re: How well does PostgreSQL 9.6.1 support unicode?
Date: 2016-12-21 16:31:38
Message-ID: 19113.1482337898@sss.pgh.pa.us (view raw, whole thread or download thread mbox)
Thread:
Lists: pgsql-general
Vick Khera <vivek(at)khera(dot)org> writes:
> On Wed, Dec 21, 2016 at 2:56 AM, Kyotaro HORIGUCHI <
> horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
>> A PostgreSQL database with encoding=UTF8 just accepts the whole
>> range of Unicode, regardless that a character is defined for the
>> code or not.

> Interesting... when I converted my application and database to utf8
> encoding, I discovered that Postgres is picky about UTF-8. Specifically the
> UTF-8 code point 0xed 0xa0 0x8d which maps to UNICODE code point 0xd80d.
> This looks like a proper character but in fact is not a defined character
> code point.

Well, we're picky to the extent that RFC 3629 tells us to be picky:
http://www.faqs.org/rfcs/rfc3629.html

The case you mention is rejected because it would be half of a UTF16
"surrogate pair", which should not be used in any Unicode representation
other than UTF16; if we allowed it then there would be more than one way
to represent the same Unicode code point, which is undesirable for a lot
of reasons.

> So I think when you present an actual string of UTF8 encoded characters,
> Postgres does refuse characters unknown. However, as you observe, inserting
> the unicode code point directly does not produce an error:

> insert into unicode(id, string) values(1, U&'\d80d');
> INSERT 0 1

Hm.  I think that's a bug.  The lexer does know that \d80d is half of a
surrogate pair, and it expects the second half to come next.  If you
supply something that isn't the second half of a surrogate pair, you
get an error as expected:

u8=# insert into unicode(id, string) values(1, U&'\d80dfoo');
ERROR:  invalid Unicode surrogate pair at or near "foo'"
LINE 1: insert into unicode(id, string) values(1, U&'\d80dfoo');
                                                          ^

But it looks like if you just end the string after the first half of a
surrogate, it just drops the character without complaint.  Notice that
what got inserted was a zero-length string, not U+D08D:

u8=# select *, length(string) from unicode;
 id | string | length 
----+--------+--------
  1 |        |      0
(1 row)

I'd have expected a syntax error along the line of "incomplete Unicode
surrogate pair".  Peter, I think this was your code to begin with ---
was it intentional to not raise error here, or is that an oversight?

			regards, tom lane


In response to

Responses

pgsql-general by date

Next:From: Adrian KlaverDate: 2016-12-21 16:58:58
Subject: Re: error updating a tuple after promoting a standby
Previous:From: Tom DalPozzoDate: 2016-12-21 16:17:03
Subject: Re: error updating a tuple after promoting a standby

Privacy Policy | About PostgreSQL
Copyright © 1996-2017 The PostgreSQL Global Development Group