Thread

Hi. I have an odd search problem on a site I’m managing, http://www.cehis.se

When site-searching some words doesnt show up depending on case, but others do.

Almost all words work fine whether written in upper or lowercase, it finds the pages. We can rule out that the search is case sensitive as searching for “ExEmPeL” returns hits, and it’s certanly not written like than anywhere. Word censoring is also turned of.

However, some words doesnt work. Searching for “npö” returns no hits, but “NPÖ” does. And “EHÄLSA” doens’t work, but “ehälsa” does.

I can’t really find a pattern here? Any ideas where to start looking?

Running ExpressionEngine 1.6.9, Revision:  20100805

Thanks

Which version of MySQL are you using?

5.0.77

So it’s only happening with umlauts, is that correct? Do both “NPÖ” and “npÖ” work? (Uppercase Ö in both cases).

Yes, both “NPÖ” and “npÖ” works, but lowercase “ö” doesn’t. But it works (both lowercase and uppercase) in all other words we’ve tried. The problem only occurs when typing the word “npö”.

It looks like an issue with your database collation. Can you run a few custom SQL queries for me?

SELECT 'a' = 'A', 'a' = 'À', 'a' = 'á', 'ö' = 'Ö', 'u' = 'Ü';

You should be getting 5 x 1, but I bet there’ll be a few zeros in your case.

SELECT CHARSET(USER()), COLLATION(USER())

What do you get?

Thanks, take a look at the attached images for the query results.

Thanks for posting the screen shots, Fredrik. It looks like the problem is with the collation/version of MySQL you’re using. I’ll ask Ingmar if there is a better collation to use. Thanks in advance for your patience.

One more, please:

SELECT COLLATION("test")

Hi! Missed following up on this!

Here’s the screenshot of “SELECT COLLATION(“test”)”

Hi Fredrik,

I’ll make sure Ingmar sees your response. Thanks in advance for your patience.

Cheers

Greg

There are two collations in use here, utf8_generel_ci and utf8_unicode_ci. The former is a very simple one, basically stripping all accents, converting to upper case and then use that. The latter (utf8_unicode_ci) is clearly the better choice. Can you make sure your default settings and db tables are using that?