MySQL "incorrect string value" error when save unicode string in Django

Question

I got strange error message when tried to save first_name, last_name to Django's auth_user model.

Failed examples

user = User.object.create_user(username, email, password)
user.first_name = u'Rytis'
user.last_name = u'Slatkevičius'
user.save()
>>> Incorrect string value: '\xC4\x8Dius' for column 'last_name' at row 104

user.first_name = u'Валерий'
user.last_name = u'Богданов'
user.save()
>>> Incorrect string value: '\xD0\x92\xD0\xB0\xD0\xBB...' for column 'first_name' at row 104

user.first_name = u'Krzysztof'
user.last_name = u'Szukiełojć'
user.save()
>>> Incorrect string value: '\xC5\x82oj\xC4\x87' for column 'last_name' at row 104

Succeed examples

user.first_name = u'Marcin'
user.last_name = u'Król'
user.save()
>>> SUCCEED

MySQL settings

mysql> show variables like 'char%';
+--------------------------+----------------------------+
| Variable_name            | Value                      |
+--------------------------+----------------------------+
| character_set_client     | utf8                       | 
| character_set_connection | utf8                       | 
| character_set_database   | utf8                       | 
| character_set_filesystem | binary                     | 
| character_set_results    | utf8                       | 
| character_set_server     | utf8                       | 
| character_set_system     | utf8                       | 
| character_sets_dir       | /usr/share/mysql/charsets/ | 
+--------------------------+----------------------------+
8 rows in set (0.00 sec)

Table charset and collation

Table auth_user has utf-8 charset with utf8_general_ci collation.

Results of UPDATE command

It didn't raise any error when updating above values to auth_user table by using UPDATE command.

mysql> update auth_user set last_name='Slatkevičiusa' where id=1;
Query OK, 1 row affected, 1 warning (0.00 sec)
Rows matched: 1  Changed: 1  Warnings: 0

mysql> select last_name from auth_user where id=100;
+---------------+
| last_name     |
+---------------+
| Slatkevi?iusa | 
+---------------+
1 row in set (0.00 sec)

PostgreSQL

The failed values listed above can be updated into PostgreSQL table when I switched the database backend in Django. It's strange.

mysql> SHOW CHARACTER SET;
+----------+-----------------------------+---------------------+--------+
| Charset  | Description                 | Default collation   | Maxlen |
+----------+-----------------------------+---------------------+--------+
...
| utf8     | UTF-8 Unicode               | utf8_general_ci     |      3 | 
...

But from http://www.postgresql.org/docs/8.1/interactive/multibyte.html, I found the following:

Name Bytes/Char
UTF8 1-4

Is it means unicode char has maxlen of 4 bytes in PostgreSQL but 3 bytes in MySQL which caused above error?

Answer 1

None of these answers solved the problem for me. The root cause being:

You cannot store 4-byte characters in MySQL with the utf-8 character set.

MySQL has a 3 byte limit on utf-8 characters (yes, it's wack, nicely summed up by a Django developer here)

To solve this you need to:

1. Change your MySQL database, table and columns to use the utf8mb4 character set (only available from MySQL 5.5 onwards)
2. Specify the charset in your Django settings file as below:

settings.py

DATABASES = {
    'default': {
        'ENGINE':'django.db.backends.mysql',
        ...
        'OPTIONS': {'charset': 'utf8mb4'},
    }
}

Note: When recreating your database you may run into the 'Specified key was too long' issue.

The most likely cause is a CharField which has a max_length of 255 and some kind of index on it (e.g. unique). Because utf8mb4 uses 33% more space than utf-8 you'll need to make these fields 33% smaller.

In this case, change the max_length from 255 to 191.

Alternatively you can edit your MySQL configuration to remove this restriction but not without some django hackery

UPDATE: I just ran into this issue again and ended up switching to PostgreSQL because I was unable to reduce my VARCHAR to 191 characters.

Answer 2

I had the same problem and resolved it by changing the character set of the column. Even though your database has a default character set of utf-8 I think it's possible for database columns to have a different character set in MySQL. Here's the SQL QUERY I used:

    ALTER TABLE database.table MODIFY COLUMN col VARCHAR(255)
    CHARACTER SET utf8 COLLATE utf8_general_ci NOT NULL;

Answer 3

If you have this problem here's a python script to change all the columns of your mysql database automatically.

#! /usr/bin/env python
import MySQLdb

host = "localhost"
passwd = "passwd"
user = "youruser"
dbname = "yourdbname"

db = MySQLdb.connect(host=host, user=user, passwd=passwd, db=dbname)
cursor = db.cursor()

cursor.execute("ALTER DATABASE `%s` CHARACTER SET 'utf8' COLLATE 'utf8_unicode_ci'" % dbname)

sql = "SELECT DISTINCT(table_name) FROM information_schema.columns WHERE table_schema = '%s'" % dbname
cursor.execute(sql)

results = cursor.fetchall()
for row in results:
  sql = "ALTER TABLE `%s` convert to character set DEFAULT COLLATE DEFAULT" % (row[0])
  cursor.execute(sql)
db.close()

Answer 4

If it's a new project, I'd just drop the database, and create a new one with a proper charset:

CREATE DATABASE <dbname> CHARACTER SET utf8;

Answer 5

I just figured out one method to avoid above errors.

Save to database

user.first_name = u'Rytis'.encode('unicode_escape')
user.last_name = u'Slatkevičius'.encode('unicode_escape')
user.save()
>>> SUCCEED

print user.last_name
>>> Slatkevi\u010dius
print user.last_name.decode('unicode_escape')
>>> Slatkevičius

Is this the only method to save strings like that into a MySQL table and decode it before rendering to templates for display?

Answer 6

You can change the collation of your text field to UTF8_general_ci and the problem will be solved.

Notice, this cannot be done in Django.

Answer 7

Improvement to @madprops answer - solution as a django management command:

import MySQLdb
from django.conf import settings

from django.core.management.base import BaseCommand


class Command(BaseCommand):

    def handle(self, *args, **options):
        host = settings.DATABASES['default']['HOST']
        password = settings.DATABASES['default']['PASSWORD']
        user = settings.DATABASES['default']['USER']
        dbname = settings.DATABASES['default']['NAME']

        db = MySQLdb.connect(host=host, user=user, passwd=password, db=dbname)
        cursor = db.cursor()

        cursor.execute("ALTER DATABASE `%s` CHARACTER SET 'utf8' COLLATE 'utf8_unicode_ci'" % dbname)

        sql = "SELECT DISTINCT(table_name) FROM information_schema.columns WHERE table_schema = '%s'" % dbname
        cursor.execute(sql)

        results = cursor.fetchall()
        for row in results:
            print(f'Changing table "{row[0]}"...')
            sql = "ALTER TABLE `%s` convert to character set DEFAULT COLLATE DEFAULT" % (row[0])
            cursor.execute(sql)
        db.close()

Hope this helps anybody but me :)

Answer 8

You aren't trying to save unicode strings, you're trying to save bytestrings in the UTF-8 encoding. Make them actual unicode string literals:

user.last_name = u'Slatkevičius'

or (when you don't have string literals) decode them using the utf-8 encoding:

user.last_name = lastname.decode('utf-8')