feat: add CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT django setting

Author: jayvynl

CHANGELOG.md

@@ -1,3 +1,7 @@
+### 1.1.6
+
+- add `CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT` django setting.
+
 ### 1.1.5
 - refactor: refactor uniq aggregate function.
 - feat: add some ClickHouse tuple and hash functions.

clickhouse_backend/__init__.py

@@ -1,5 +1,5 @@
 from clickhouse_backend.utils.version import get_version
 
-VERSION = (1, 1, 5, "final", 0)
+VERSION = (1, 1, 6, "final", 0)
 
 __version__ = get_version(VERSION)

clickhouse_backend/driver/connection.py

@@ -5,6 +5,7 @@
 from clickhouse_driver.dbapi import connection as dbapi_connection
 from clickhouse_driver.dbapi import cursor, errors
 from clickhouse_pool.pool import ChPoolError
+from django.conf import settings
 
 from .escape import escape_params
 from .pool import ClickhousePool
@@ -93,7 +94,9 @@ def __del__(self):
 
     def execute(self, operation, parameters=None):
         """fix https://github.com/jayvynl/django-clickhouse-backend/issues/9"""
-        if update_pattern.match(operation):
+        if getattr(
+            settings, "CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT", True
+        ) and update_pattern.match(operation):
             query = self._client.substitute_params(
                 operation, parameters, self._client.connection.context
             )

docs/Configurations.md

@@ -73,3 +73,20 @@ If data insertions happen on multiple datacenter, server, process or thread, you
 Because work_id and datacenter_id are 5 bits, they should be an integer between 0 and 31. CLICKHOUSE_WORKER_ID default to 0, CLICKHOUSE_DATACENTER_ID will be generated randomly if not provided.
 
 `clickhouse.idworker.snowflake.SnowflakeIDWorker` is not thread safe. You could inherit `clickhouse.idworker.base.BaseIDWorker` and implement one, then set `CLICKHOUSE_ID_WORKER` in `settings.py` to doted import path of your IDWorker instance.
+
+### Other
+
+#### CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT
+
+*New in 1.1.6*
+
+Whether returning rowcount when executing `update` query, default is `True`.
+
+Django rely on [rowcount](https://peps.python.org/pep-0249/#rowcount) of [Python DBAPI 2.0](https://peps.python.org/pep-0249/) to get number of rows deleted or updated.
+However, clickhouse_driver always have rowcount equals -1 when updating and deleting.
+
+Updating of Django builtin model instances will [work abnormally](https://github.com/jayvynl/django-clickhouse-backend/issues/9) when this feature is missing.
+To fix this issue, ClickHouse backend use a modified version of clickhouse_driver which execute count aggregation and set rowcount before updating.
+
+However, this will cause another problem, count aggregation will be extremely slow sometimes, because it scans all affected rows.
+Set `CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT` to `False` when this feature is not needed.

tests/queries/__init__.py

@@ -0,0 +1,6 @@
+from django.db import models
+
+
+class Author(models.Model):
+    name = models.CharField(max_length=10)
+    num = models.IntegerField()

tests/queries/models.py

@@ -0,0 +1,22 @@
+from django.test import TestCase
+
+from . import models
+
+
+class QueriesTests(TestCase):
+    @classmethod
+    def setUpTestData(cls):
+        cls.a1 = models.Author.objects.create(name="a1", num=1001)
+
+    def test_update(self):
+        with self.settings(CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT=True):
+            self.assertEqual(models.Author.objects.update(name="a11"), 1)
+            with self.assertNumQueries(1):
+                self.a1.save()
+            self.a1.refresh_from_db()
+        with self.settings(CLICKHOUSE_ENABLE_UPDATE_ROWCOUNT=False):
+            self.assertEqual(models.Author.objects.update(name="a11"), -1)
+            with self.assertNumQueries(2):
+                self.a1.save()
+            with self.assertRaises(models.Author.MultipleObjectsReturned):
+                self.a1.refresh_from_db()