Обсуждение: convert string function and built-in conversions

It seems to me that these values should be the same:

select 'lydia eugenia treviño', convert('lydia eugenia treviño' using
ascii_to_utf_8);

but they seem to be different.  What am I missing?

culley

On Sun, 19 Oct 2003, culley harrelson wrote:

> It seems to me that these values should be the same:
>
> select 'lydia eugenia treviño', convert('lydia eugenia treviño' using
> ascii_to_utf_8);
>
> but they seem to be different.  What am I missing?

I don't think the marked n is a valid ascii character (it might be
extended ascii, but that's different and not really standard afaik).
You're probably getting the character associated with the lower 7 bits.

It is one of the extended characters in iso-8859-1.  This data was taken
from a text field in a SQL_ASCII database.  Basically what I am trying to
do is migrate data from a SQL_ASCII database to a UNICODE database by
running all the data through an external script that does something like:

select convert(my_field using ascii_to_utf_8) from my_table;

then inserts the selected text into an identical table in the unicode
database.  All the data goes across, but extended characters such as ñ
are getting munged.  The docs indicate that ascii_to_utf_8 is for
SQL_ASCII -> UNICODE...  Are you saying that ñ isn't really an ASCII
character even though it is valid in a SQL_ASCII database?  I have found
that all extended characters of the various LATIN encodings will work
just fine in my SQL_ASCII database.

This project is a big can of worms...  Every 6 months I open the can,
stir the worms around a bit, wrinkle my nose then promptly close the can
again and stuff it away for another 6 months. :)  Wish I could figure it
out.



On Sun, 19 Oct 2003 00:31:43 -0700 (PDT), "Stephan Szabo"
<sszabo@megazone.bigpanda.com> said:
> On Sun, 19 Oct 2003, culley harrelson wrote:
>
> > It seems to me that these values should be the same:
> >
> > select 'lydia eugenia treviño', convert('lydia eugenia treviño' using
> > ascii_to_utf_8);
> >
> > but they seem to be different.  What am I missing?
>
> I don't think the marked n is a valid ascii character (it might be
> extended ascii, but that's different and not really standard afaik).
> You're probably getting the character associated with the lower 7 bits.

On Sun, 19 Oct 2003, culley harrelson wrote:

> It is one of the extended characters in iso-8859-1.  This data was taken
> from a text field in a SQL_ASCII database.  Basically what I am trying to
> do is migrate data from a SQL_ASCII database to a UNICODE database by
> running all the data through an external script that does something like:
>
> select convert(my_field using ascii_to_utf_8) from my_table;
>
> then inserts the selected text into an identical table in the unicode
> database.  All the data goes across, but extended characters such as �
> are getting munged.  The docs indicate that ascii_to_utf_8 is for
> SQL_ASCII -> UNICODE...  Are you saying that � isn't really an ASCII
> character even though it is valid in a SQL_ASCII database?  I have found
> that all extended characters of the various LATIN encodings will work
> just fine in my SQL_ASCII database.

I would guess that it's not actually forcing/checking the characters for 7
bitness in SQL_ASCII, but that the conversions are treating them as if you
had actually only put in valid 7 bit values (as they appear to be doing
an & 0x7F in at least the routines I looked at).

If you're actually putting iso-8859-1 (latin1) in there, try the
conversion from iso-8859-1 to utf8. It doesn't appear to display properly
in my iso-8859-1 terminal, but taking that string and inserting it into a
unicode database and then setting my client_encoding to iso-8859-1 gives
me the original string back when I select it.