On Thu, 2008-12-04 at 09:41 +0100, Lars-Erik Bjørk wrote:
> Just to make sure that I understand things correctly :)
> The trailing spaces (and minimum sort characters) I have been removing
> so far, I have converted to the given character set, so for f.eks UCS2,
> I have been looking to remove 0x0020, etc
> Should it be done like this for trailing zeros as well, so that I in the
> case of ucs2 should be looking to remove 0x0000?
> Also, if I try to remove only spaces (pad chars) and zeros, but not the
> minimum sort character, some queries using LIKE comparison in the Falcon
> suite starts failing.
Actually there is only one test that fails:
CREATE TABLE t1 (a char(4) CHARACTER SET utf8 COLLATE utf8_turkish_ci,
INSERT INTO t1 VALUES ('abc'), ('abc '), ('abcd');
SELECT count(*) FROM t1 WHERE a like 'abc%';
> Kevin is correct btw, that bug#23692 has not been prioritized because it
> does not have a BETA tag.
> On Wed, 2008-12-03 at 14:36 -0500, Ann W. Harrison wrote:
> > Alexander Barkov wrote:
> > >
> > > That's true , <tab> in many collations is the smallest character
> > > (i.e. minSortChar=<tab>), and space is greater than <tab>.
> > >
> > > So does not seem correct to remove both spaces and tabs, and
> > > then execute strnncoll() for the results in the patch for
> > > bug#34479.
> > >
> > You're absolutely right.
> > We should not remove tabs, but only spaces. I don't know what
> > the code does currently, but the original design (not necessarily
> > its implementation) removed only spaces. (And, of course, trailing
> > zeros for numerics, but that's another topic.)
> > Best regards,
> > Ann