CODEWITHSUNDEEP
mysqlutf-8case-insensitivenon-ascii-characters
Lightworker
Lightworker

Reputation: 603

How to MySQL work "case insensitive" and "accent insensitive" in UTF-8

I have a schema in "utf8 -- UTF-8 Unicode" as charset and a collation of "utf8_spanish_ci".

All the inside tables are InnoDB with same charset and collation as mentioned.

Here comes the problem:

with a query like

SELECT *
FROM people p
WHERE p.NAME LIKE '%jose%';

I get 83 result rows. I should have 84 results, because I know it.

Changing where for:

WHERE p.NAME LIKE '%JOSE%';

I get the exact same 83 rows. With combinations like JoSe, Jose, JOSe, etc. All the same 83 rows are reported.

The problem comes when accents play in game. If do:

WHERE p.NAME LIKE '%josé%';

I get no results. 0 rows.

But if I do:

WHERE p.NAME LIKE '%JOSÉ%';

I get just one resulting row, so 1 row. This is the only row which has accented "jose" and capitalized.

I've tried with josÉ, or JoSÉ or whatever combination I do, as long as the accented letter stays capitalized or not, as it really is stored in the database and it stills returning the only row. If I suddenly change "É" for "é" in whatever combination I do with the capitalization in JOSE, it returns no rows.

So conclusions:

What I want?

Solutions like COLLATION on LIKE doesn't work for me, don't know why...

What can I do?

EDIT:

If I do something like:

WHERE p.NAME LIKE '%jose%' COLLATE utf8_general_ci;

I get the error:

COLLATION 'utf8_general_ci' is not valid for CHARACTER SET 'latin1'

And I've changed all the possible collations on the columns too!

And if I do something like:

WHERE p.NAME LIKE _utf8 '%jose%' COLLATE utf8_general_ci;

The same 83 rows are reported, as if I've made nothing...

Upvotes: 22

Views: 28094

Answers (2)

Miguel Mesquita Alfaiate
Miguel Mesquita Alfaiate

Reputation: 2968

Just in case someone else stumbles upon this issue, I have found a way that solves the problem, at least for me.

I am using PHP to insert and retrieve records from the database. Even though my Database, tables and columns are utf8, as well as the encoding of the PHP files, the truth is that the encoding used in the connection between PHP and MySQL is being made using latin1. I managed to find this using

$mysqli->character_set_name();

where $mysqli is your object.

For the searches to start working as expected, returning accent insensitive and case insentive records for characters with accents or not, I have to explicitly set the character set of the connection.

To do this, you just have to do the following:

$mysqli->set_charset('utf8');

where $mysqli is your mysqli object. If you have a database management class that wraps your database functionality, this is easy to apply to a complete app. If not, you have to set this explicitly everywhere you open a connection.

I hope this helps someone out, as I was already freaking out about this!

Upvotes: 8

O. Jones
O. Jones

Reputation: 108806

You have already tried to use an accent-insensitive collation for your search and ordering.

http://dev.mysql.com/doc/refman/5.0/en/charset-collation-implementations.html

The thing is, your NAME column seems to be stored in the latin1 (8-bit) character set. That's why mySQL is grumbling at you like this:

  COLLATION 'utf8_general_ci' is not valid for CHARACTER SET 'latin1'

You may get the results you want if you try

 WHERE CONVERT(p.NAME USING utf8) LIKE _utf8 '%jose%' COLLATE utf8_general_ci;

But, be careful!

When you use any kind of function (in this example, CONVERT) on the column in a WHERE statement, you defeat MySQL's attempts to optimize your search with indexes. If this project is going to get large (that is, if you will have lots of rows in your tables) you need to store your data in utf8 format, not latin1. (You probably already know that your LIKE '%whatever%' search term also defeats MySQL's indexing.)

Upvotes: 11

Related Questions