# HG changeset patch
# User Mads Kiilerich <mads@kiilerich.com>
# Date 2020-06-15 17:27:19
# Node ID 7b0aafc6b7cac0f4ae766287b14d744f3add1bbd
# Parent  210e76d69b62057e04921169b272cc3f8c19fb9c

mysql: create database with explicit UTF-8 character set and collation

A spin-off from Issue #378.

In MySQL, the character sets for server, database, tables, and connection are
set independently. Ideally, they should all use UTF-8, but systems tend to use
latin1 as default encoding, for example:

    character_set_server = latin1
    collation_server = latin1_swedish_ci

Databases would thus by default be created as:

    character_set_database = latin1
    collation_database = latin1_swedish_ci

To make things work consistently anyway, we have so far specified the utf8mb4
charset explicitly when creating tables, but there is no corresponding simple
option for specifying the collation for tables. We need a better solution.

If necessary and possible, the system charset and collation should be set to
UTF-8. Some systems already have these defaults default - see
https://mariadb.com/kb/en/differences-in-mariadb-in-debian-and-ubuntu/ .
The defaults can be changed as described on
https://mariadb.com/kb/en/setting-character-sets-and-collations/#example-changing-the-default-character-set-to-utf-8
to give something like:

    character_set_server = utf8mb4
    collation_server = utf8mb4_unicode_ci

Databases will then by default be created as:

    character_set_database = utf8mb4
    collation_database = utf8mb4_unicode_ci

and there is thus no longer any need for specifying the charset when creating
tables.

To be reasonably resilient across all systems without relying on system
defaults, we will now start specifying the charset and collation when creating
the database, but drop the specification of charset when creating tables.

For existing databases, it is recommended to change encoding (and collation) by
altering the database and each of the tables inside it as described on
https://stackoverflow.com/questions/6115612/how-to-convert-an-entire-mysql-database-characterset-and-collation-to-utf-8 .

Note the use of utf8mb4_unicode_ci instead of utf8mb4_general_ci  - see
https://stackoverflow.com/questions/766809/whats-the-difference-between-utf8-general-ci-and-utf8-unicode-ci .

For investigation of these issues, consider the output from:
    show variables like '%char%';
    show variables like '%collation%';
    show create database `KALLITHEA_DB_NAME`;
    SELECT * FROM information_schema.SCHEMATA WHERE schema_name = "KALLITHEA_DB_NAME";
    SELECT * FROM information_schema.TABLES T, information_schema.COLLATION_CHARACTER_SET_APPLICABILITY CCSA WHERE CCSA.collation_name = T.table_collation AND T.table_schema = "KALLITHEA_DB_NAME";

diff --git a/kallithea/lib/db_manage.py b/kallithea/lib/db_manage.py
--- a/kallithea/lib/db_manage.py
+++ b/kallithea/lib/db_manage.py
@@ -97,7 +97,7 @@ class DbManage(object):
                 engine = sqlalchemy.create_engine(url)
                 with engine.connect() as conn:
                     conn.execute('DROP DATABASE IF EXISTS `%s`' % database)
-                    conn.execute('CREATE DATABASE `%s`' % database)
+                    conn.execute('CREATE DATABASE `%s` CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci' % database)
             elif url.drivername == 'postgresql':
                 from psycopg2.extensions import ISOLATION_LEVEL_AUTOCOMMIT
                 url.database = 'postgres'  # connect to the system database (as the real one might not exist)
diff --git a/kallithea/model/db.py b/kallithea/model/db.py
--- a/kallithea/model/db.py
+++ b/kallithea/model/db.py
@@ -167,7 +167,6 @@ class BaseDbModel(object):
 
 _table_args_default_dict = {'extend_existing': True,
                             'mysql_engine': 'InnoDB',
-                            'mysql_charset': 'utf8mb4',
                             'sqlite_autoincrement': True,
                            }