Merge pull request #35917 from kitaisreal/executable-user-defined-function-fix

Executable user defined functions prevent executing during analysis

src/Common/ProfileEvents.cpp

@@ -112,6 +112,8 @@
     M(CompileExpressionsMicroseconds, "Total time spent for compilation of expressions to LLVM code.") \
     M(CompileExpressionsBytes, "Number of bytes used for expressions compilation.") \
     \
+    M(ExecuteShellCommand, "Number of shell command executions.") \
+    \
     M(ExternalSortWritePart, "") \
     M(ExternalSortMerge, "") \
     M(ExternalAggregationWritePart, "") \

src/Common/ShellCommand.cpp

@@ -29,6 +29,11 @@ namespace
     };
 }
 
+namespace ProfileEvents
+{
+    extern const Event ExecuteShellCommand;
+}
+
 namespace DB
 {
 
@@ -158,6 +163,7 @@ std::unique_ptr<ShellCommand> ShellCommand::executeImpl(
     const Config & config)
 {
     logCommand(filename, argv);
+    ProfileEvents::increment(ProfileEvents::ExecuteShellCommand);
 
 #if !defined(USE_MUSL)
     /** Here it is written that with a normal call `vfork`, there is a chance of deadlock in multithreaded programs,

src/Interpreters/UserDefinedExecutableFunctionFactory.cpp

@@ -57,6 +57,10 @@ public:
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
     {
+        /// Do not start user defined script during query analysis. Because user script startup could be heavy.
+        if (input_rows_count == 0)
+            return result_type->createColumn();
+
         auto coordinator = executable_function->getCoordinator();
         const auto & coordinator_configuration = coordinator->getConfiguration();
         const auto & configuration = executable_function->getConfiguration();

tests/queries/0_stateless/02252_executable_user_defined_function_short_circuit.reference

@@ -0,0 +1 @@
+0

tests/queries/0_stateless/02252_executable_user_defined_function_short_circuit.sql

@@ -0,0 +1,10 @@
+SELECT number FROM numbers(10) WHERE number > 15 and test_function(number, number) == 4;
+
+SYSTEM FLUSH LOGS;
+
+SELECT ProfileEvents['ExecuteShellCommand'] FROM system.query_log WHERE
+    current_database = currentDatabase()
+    AND type = 'QueryFinish'
+    AND query == 'SELECT number FROM numbers(10) WHERE number > 15 and test_function(number, number) == 4;'
+    AND event_date >= yesterday() AND event_time > now() - interval 10 minute
+    LIMIT 1;