On Mon, Oct 24, 2005 at 05:07:40AM -0000, Andrew - Supernews wrote:
>
> I'm inclined to suspect that the whole sequence c1 f9 d4 c2 d0 c7 d2 b9
> was never actually a valid utf-8 string, and that the d2 b9 is only valid
> by coincidence (it's a Cyrillic letter from Azerbaijani). I know the 8.0
> utf-8 check was broken, but I didn't realize it was quite so bad.
Looking at the data it appears that it is a sequence of latin1
characters. They all have the eighth bit set and all seem to pass the
check.
In a million rows I found 2 examples of this.
However I'm running into another problem now. The command:
iconv -c -f UTF8 -t UTF8
does strip out the invalid characters. However, iconv reads the
entire file into memory before it writes out any data. This is not so
good for multi-gigabyte dump files and doesn't allow for it to be used
in a pipe between pg_dump and psql.
Anyone have any other recommendations? GNU recode might do it, but
I'm a bit stymied by the syntax. A quick perl script using
Text::Iconv didn't work either. I'm off to look at some other perl
modules and will try to create a script so I can strip out the invalid
characters.
--
Paul Lindner ||||| | | | | | | | | |
lindner@inuus.com