On 08/26/2013 11:10 PM, Jan Slodicka wrote:
Thanks, Simon.
As a risk, I see overriding an existing collation name as risky. If any
application makes changes to the database without the extension loaded,
you're going to get corrupt indexes. Can you instead make up a new
collation name ?
I did not think about this, but in our case it should not be that hot. It is
an embedded application and the users are not supposed to manipulate the DB
directly. (In some cases the OS efficiently hides the DB file.)
Sounds like an accident waiting to happen though. A developer could
update a file using the sqlite shell. If the built-in NOCASE is slightly
different from yours, such an update could subtly corrupt an index.
Which could easily lead to SQLite returning incorrect query results
later on.
If you name your collation sequence something different, any attempt
to use the stock shell to update an index will fail with the "no such
collation" message.
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
