On 05/16/2014 06:05 PM, Tom Lane wrote:
> Quite some time ago, we made the chr() function accept Unicode code points
> up to U+1FFFFF, which is the largest value that will fit in a 4-byte UTF8
> string. It was pointed out to me though that RFC3629 restricted the
> original definition of UTF8 to only allow code points up to U+10FFFF (for
> compatibility with UTF16). While that might not be something we feel we
> need to follow exactly, pg_utf8_islegal implements the checking algorithm
> specified by RFC3629, and will therefore reject points above U+10FFFF.
>
> This means you can use chr() to create values that will be rejected on
> dump and reload:
>
> u8=# create table tt (f1 text);
> CREATE TABLE
> u8=# insert into tt values(chr('x001fffff'::bit(32)::int));
> INSERT 0 1
> u8=# select * from tt;
> f1
> ----
>
> (1 row)
>
> u8=# \copy tt to 'junk'
> COPY 1
> u8=# \copy tt from 'junk'
> ERROR: 22021: invalid byte sequence for encoding "UTF8": 0xf7 0xbf 0xbf 0xbf
> CONTEXT: COPY tt, line 1
> LOCATION: report_invalid_encoding, wchar.c:2011
>
> I think this probably means we need to change chr() to reject code points
> above 10ffff. Should we back-patch that, or just do it in HEAD?
+1 for back-patching. A value that cannot be restored is bad, and I
can't imagine any legitimate use case for producing a Unicode character
larger than U+10FFFF with chr(x), when the rest of the system doesn't
handle it. Fully supporting such values might be useful, but that's a
different story.
- Heikki