[Once again, please copy the mailing list on replies so others can participate in and learn from the discussion. Also, pgsql-general might be a more appropriate list than pgsql-admin.] On Mon, Feb 26, 2007 at 10:53:48AM +0100, j n wrote: > 1. At first i tried use to_ascii ... convert > it works well on some letter 'á' converted to 'a' 'é' to 'e' but some of > them like č or š convert as empty string it means 'not working fine' We can't explain why this doesn't work unless you show exactly what you did. Please post a set of SQL statements that somebody could run in their own database to reproduce the problem you're seeing. Did any of this data originate on Windows? If so then it's possible that some accented characters aren't represented by the proper Unicode code points. This can happen, for example, if you load Windows-1250 data into the database with client_encoding set to LATIN2. Depending on how you're viewing the data the "wrong" characters might still display correctly. To give a specific example, š is 0x9a in Windows-1250 but if you load this character with client_encoding set to LATIN2 then the database converts it to U+009A, a control character, instead of to U+0161 latin small letter s with caron (háček). An application that reads the control character might render it as š assuming that that's what character was meant, but functions that operate on the data won't work as expected. This wouldn't fully explain the problems you're seeing but it's something I've seen cause similar problems. > 2. than i tried to use perl func but i didn't have configured postgres to > use perl so i have to : > > ./configure --enable-multibyte=UTF8 prefix=/usr/local/pgsqlProd > --exec-prefix=/usr/local/pgsqlProd --with-perl The --enable-multibyte option was removed in 7.3 so you don't need it; at the end of the configure you should have seen a warning that this option was ignored. And did you mean --prefix instead of prefix? Also, there's no need to set --exec-prefix if it gets the same value as --prefix. > gmake > gmake install > > 3. After this everythink have started to work fine also unaccent perl > function ... Configuring with --with-perl is necessary if you want to use server- side Perl functions, but you didn't say anything about Perl functions in your previous message. You said that to_ascii() and convert() didn't work unless you used --with-perl, which doesn't make sense because those functions have nothing to do with Perl. I tried it both ways and got the same behavior so I'm still skeptical that --with-perl is the relevant difference for to_ascii() and convert() behavior. -- Michael Fuhr
