From: Georg Richter <georg@mariadb.com>
Date: Thu, 30 Sep 2021 22:13:49 +0000 (-0400)
Subject: correct charset for mariadb connector with mdb 10.6
X-Git-Tag: rel_1_4_26~50
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=2e6285d8d1d293c841f18313c94e493b2804055c;p=thirdparty%2Fsqlalchemy%2Fsqlalchemy.git

correct charset for mariadb connector with mdb 10.6

Updated test suite to pass correctly for MariaDB 10.6 when using the
mariadb-connector driver, which made some adjustments to default encoding.
Pull request courtesy Georg Richter.

Use character set utf8mb4 instead of utf8:

1) Character set for mariadb connector is utf8mb4 and
cannot be changed.

2) Since MariaDB 10.6.1 utf8 is mapped to utf8mb3, and
will be mapped to utf8mb4 in 10.7, see [MDEV-8334](https://jira.mariadb.org/browse/MDEV-8334) (rename
utf8 to utf8mb3)

utf8mb4 character set support was added 12 years ago and
will work with all non eoled MariaDB server versions.

Fixes: #7115

Closes: #7116
Pull-request: https://github.com/sqlalchemy/sqlalchemy/pull/7116
Pull-request-sha: 79516ccce2dead11e6658596c9ad400b964d24c9

Change-Id: I86c30fb636c46307cb551080641e324d2bbc1cb3
---

diff --git a/doc/build/changelog/unreleased_14/7115.rst b/doc/build/changelog/unreleased_14/7115.rst
new file mode 100644
index 0000000000..561e152315
--- /dev/null
+++ b/doc/build/changelog/unreleased_14/7115.rst
@@ -0,0 +1,8 @@
+.. change::
+    :tags: bug, mysql, tests
+    :tickets: 7115
+
+    Updated test suite to pass correctly for MariaDB 10.6 when using the
+    mariadb-connector driver, which made some adjustments to default encoding.
+    Pull request courtesy Georg Richter.
+
diff --git a/lib/sqlalchemy/dialects/mysql/base.py b/lib/sqlalchemy/dialects/mysql/base.py
index 0c24ebf5e7..530e7c9ccb 100644
--- a/lib/sqlalchemy/dialects/mysql/base.py
+++ b/lib/sqlalchemy/dialects/mysql/base.py
@@ -3537,6 +3537,8 @@ class _DecodingRow(object):
         "koi8u": "koi8_u",
         "utf16": "utf-16-be",  # MySQL's uft16 is always bigendian
         "utf8mb4": "utf8",  # real utf8
+        "utf8mb3": "utf8",  # real utf8; saw this happen on CI but I cannot
+        # reproduce, possibly mariadb10.6 related
         "eucjpms": "ujis",
     }
 
diff --git a/test/dialect/mysql/test_reflection.py b/test/dialect/mysql/test_reflection.py
index 3091a6a7da..60d7e3a5dd 100644
--- a/test/dialect/mysql/test_reflection.py
+++ b/test/dialect/mysql/test_reflection.py
@@ -307,7 +307,7 @@ class ReflectionTest(fixtures.TestBase, AssertsCompiledSQL):
         if testing.against("mariadb"):
             kwargs = dict(
                 mariadb_engine="MEMORY",
-                mariadb_default_charset="utf8",
+                mariadb_default_charset="utf8mb4",
                 mariadb_auto_increment="5",
                 mariadb_avg_row_length="3",
                 mariadb_password="secret",
@@ -338,7 +338,7 @@ class ReflectionTest(fixtures.TestBase, AssertsCompiledSQL):
         if testing.against("mariadb"):
             assert def_table.kwargs["mariadb_engine"] == "MEMORY"
             assert def_table.comment == comment
-            assert def_table.kwargs["mariadb_default_charset"] == "utf8"
+            assert def_table.kwargs["mariadb_default_charset"] == "utf8mb4"
             assert def_table.kwargs["mariadb_auto_increment"] == "5"
             assert def_table.kwargs["mariadb_avg_row_length"] == "3"
             assert def_table.kwargs["mariadb_password"] == "secret"
@@ -348,7 +348,7 @@ class ReflectionTest(fixtures.TestBase, AssertsCompiledSQL):
 
             assert reflected.comment == comment
             assert reflected.kwargs["mariadb_comment"] == comment
-            assert reflected.kwargs["mariadb_default charset"] == "utf8"
+            assert reflected.kwargs["mariadb_default charset"] == "utf8mb4"
             assert reflected.kwargs["mariadb_avg_row_length"] == "3"
             assert reflected.kwargs["mariadb_connection"] == "fish"