Fix: Prevent column names containing 'from' from breaking function names in views

lib/active_record/connection_adapters/clickhouse/schema_creation.rb

@@ -79,7 +79,9 @@ def assign_database_to_subquery!(subquery)
           # If you do not specify a database explicitly, ClickHouse will use the "default" database.
           return unless subquery
 
-          match = subquery.match(/(?<=from)[^.\w]+(?<database>\w+(?=\.))?(?<table_name>[.\w]+)/i)
+          # Match FROM as a keyword (with word boundary), not as part of a column name
+          # \b ensures we only match 'from' as a whole word
+          match = subquery.match(/\bfrom\s+(?<database>\w+(?=\.))?(?<table_name>[.\w]+)/i)
           return unless match
           return if match[:database]
 

spec/single/schema_loading_spec.rb

@@ -0,0 +1,80 @@
+require 'spec_helper'
+
+RSpec.describe 'Schema Loading', :migrations do
+  let(:model) { ActiveRecord::Base }
+  let(:connection) { model.connection }
+  let(:database) { connection.instance_variable_get(:@config)[:database] }
+
+  describe 'assign_database_to_subquery' do
+    after do
+      connection.execute('DROP VIEW IF EXISTS test_view')
+      connection.execute('DROP TABLE IF EXISTS test_target')
+      connection.execute('DROP TABLE IF EXISTS test_source')
+    end
+
+    context 'when column name contains "from"' do
+      it 'does not mistake column name for FROM keyword' do
+        # Bug: The regex /(?<=from)/ matches "from" anywhere in the query,
+        # including in column names like "sourced_from". This causes the next
+        # identifier (often a function name) to be incorrectly prefixed with
+        # the database name.
+        #
+        # Example bug:
+        #   SELECT sourced_from, now() FROM table
+        # Would incorrectly become:
+        #   SELECT sourced_from, default.now() FROM default.table
+        # Causing: Function with name 'default.now' does not exist
+
+        connection.execute(<<~SQL)
+          CREATE TABLE test_source (
+            id UInt64,
+            sourced_from String
+          ) ENGINE = MergeTree ORDER BY id
+        SQL
+
+        connection.execute(<<~SQL)
+          CREATE TABLE test_target (
+            id UInt64,
+            sourced_from String,
+            created_at DateTime
+          ) ENGINE = MergeTree ORDER BY id
+        SQL
+
+        # This query should work: column ending in "from" followed by a function
+        expect {
+          connection.create_table :test_view, view: true, materialized: true, to: 'test_target',
+                                 as: 'SELECT id, sourced_from, now() AS created_at FROM test_source' do |t|
+          end
+        }.not_to raise_error
+      end
+
+      it 'correctly adds database prefix only to table name' do
+        connection.execute(<<~SQL)
+          CREATE TABLE test_source (
+            id UInt64
+          ) ENGINE = MergeTree ORDER BY id
+        SQL
+
+        connection.execute(<<~SQL)
+          CREATE TABLE test_target (
+            id UInt64,
+            created_at DateTime
+          ) ENGINE = MergeTree ORDER BY id
+        SQL
+
+        connection.create_table :test_view, view: true, materialized: true, to: 'test_target',
+                               as: 'SELECT id, now() AS created_at FROM test_source' do |t|
+        end
+
+        # Verify the view was created successfully
+        # Note: show_create_table strips database prefixes, so we just verify
+        # that the view exists and the function wasn't broken by incorrect prefixing
+        show_create = connection.show_create_table('test_view')
+
+        # Should NOT have database prefix on function name
+        expect(show_create).not_to include("#{database}.now")
+        expect(show_create).to include('now()')
+      end
+    end
+  end
+end