Aryeh Gregor wrote:
On Fri, May 22, 2009 at 6:09 PM, O. O.
<olson_ot(a)yahoo.com> wrote:
Thanks for your reply Platonides. I am trying
your suggestion right now.
It would take a few hours to crash – if it does. (I hope sed handles
UTF-8 correctly.) I would try yesterdays pagelinks.sql later.
sed treats UTF-8 as
a stream of bytes. Since the pattern won't match
UTF-8 (UTF-8 only contains ASCII bytes if they represent ASCII code
points), it will just ignore those bytes.
(That sed pattern is pretty horrifying and fragile, though. I'd
recommend something more like: sed -i 's/^) TYPE=InnoDB;$/)
TYPE=InnoDB DEFAULT CHARSET=binary;/' )
I wanted to be sure that the command provided to olson only modified
that line.
The content below that may be everything. Not that I think your sed will
fail in this case, given the allowed titles, but it *can* fail on some
of the other sql dumps.
And yes, the syntax is horrible. There's a way to do it providing a
number instead of repeating the "n;", but it's a gnu extension.
Thanks Platonides for the clarification on your syntax. Even if your
syntax is bad, I can get another script (which does not modify the SQL
file) to generate it.
Also, I think that your suggestion is working, because while my import
has not yet completed, from the mysql> show processlist; it seems that
it has crossed the point where it used to crash.
Keeping my fingers crossed till it completes,
O. O.