> > 3. SQL steps
> > create table test (name text);
> > insert into test values ('á'); # the first char is E1
> from LATIN 2
> > coding
> > insert into test values ('áb');
> > create index test_index on test (name);
> > set cpu_tuple_cost=1; # force backend to use index
> > scanning
> > select * from test where name like 'á%';
>
> > BUG: Only 1 line is selected with 'á' only instead of both lines.
>
> The problem here is that given the search pattern '\341%', the planner
> generates index limit conditions
> name >= '\341' AND name < '\342';
I see that you are addressing a real problem (in german 'o' sorts same as
'ö',
upper case sorts same as lower case)
but ist that related in this case ?
Seems this example has exactly the same first character in both rows,
so the bug seems to be of another class, no ?
Andreas