marking subtree as runnable; changes to execution flow to support correct execution of nested&complex aggregation queries

Signed-off-by: gal salomon <gal.salomon@gmail.com>

Author: galsalomon66

include/s3select.h

@@ -400,11 +400,10 @@ struct s3select : public bsc::grammar<s3select>
 
         if (aggregate_expr)
         {
-          //per each column, subtree is mark to skip,
-          //the second code-line marks the aggregation subtree not-to-skip.
+          //per each column, subtree is mark to skip except for the aggregation function subtree. 
           //for an example: substring( ... , sum() , count() ) :: the substring is mark to skip execution, while sum and count not.
           e->set_skip_non_aggregate(true);
-          aggregate_expr->set_skip_non_aggregate(false);
+          e->mark_aggreagtion_subtree_to_execute();
         }
         else
         {
@@ -1490,6 +1489,7 @@ class csv_object : public base_s3object
             for (auto& i : m_projections)
             {
               i->set_last_call();
+              i->set_skip_non_aggregate(false);//projection column is set to be runnable
               result.append( i->eval().to_string() );
               result.append(",");
             }

include/s3select_functions.h

@@ -216,6 +216,11 @@ class __function : public base_statement
   __function(const char* fname, s3select_functions* s3f) : name(fname), m_func_impl(nullptr), m_s3select_functions(s3f) {}
 
   virtual value& eval()
+  {
+    return eval_internal();
+  }
+
+  virtual value& eval_internal()
   {
 
     _resolve_name();
@@ -1704,6 +1709,29 @@ bool base_statement::is_nested_aggregate(bool &aggr_flow)
   return false;
 }
 
+bool base_statement::mark_aggreagtion_subtree_to_execute()
+{//purpase:: set aggregation subtree as runnable.
+ //the function search for aggregation function, and mark its subtree {skip = false}
+  if (is_aggregate())
+    set_skip_non_aggregate(false);
+  
+  if (left())
+    left()->mark_aggreagtion_subtree_to_execute();
+  
+  if(right())
+    right()->mark_aggreagtion_subtree_to_execute();
+
+  if (is_function())
+  {
+      for (auto& i : dynamic_cast<__function*>(this)->get_arguments())
+      {
+          i->mark_aggreagtion_subtree_to_execute();
+      }
+  }
+
+  return true;
+}
+
 } //namespace s3selectEngine
 
 #endif

include/s3select_oper.h

@@ -920,12 +920,37 @@ class base_statement
   base_statement* m_projection_alias;
   int m_eval_stack_depth;
   bool m_skip_non_aggregate_op;
+  value value_na;
 
 public:
   base_statement():m_scratch(nullptr), is_last_call(false), m_is_cache_result(false),
   m_projection_alias(nullptr), m_eval_stack_depth(0), m_skip_non_aggregate_op(false) {}
 
-  virtual value& eval() =0;
+  virtual value& eval()
+  {
+    //purpose: on aggregation flow to run only the correct subtree(aggregation subtree)
+     
+    if (m_skip_non_aggregate_op == false)
+      return eval_internal();//not skipping this node.
+    else
+    {
+    //skipping this node.
+    //in case execution should skip a node, it will traverse (left and right) 
+    //and search for subtree to execute.   
+    //example: sum( ... ) - sum( ... ) ; the minus operand is skipped while sum() operand is not.
+    if(left())
+      left()->eval_internal();
+    
+    if(right())
+      right()->eval_internal();
+    
+    }
+
+    return value_na;
+  }
+
+  virtual value& eval_internal() = 0;
+  
   virtual base_statement* left()
   {
     return 0;
@@ -978,6 +1003,7 @@ class base_statement
   base_statement* get_aggregate();
   bool is_nested_aggregate(bool&);
   bool is_column_reference();
+  bool mark_aggreagtion_subtree_to_execute();
 
   virtual void set_last_call()
   {
@@ -1211,7 +1237,7 @@ class variable : public base_statement
     return var_value;
   }
 
-  virtual value& eval()
+  virtual value& eval_internal()
   {
     if (m_var_type == var_t::COL_VALUE)
     {
@@ -1326,7 +1352,7 @@ class arithmetic_operand : public base_statement
     return std::string("#");//TBD
   }
 
-  virtual value& eval()
+  virtual value& eval_internal()
   {
 
     switch (_cmp)
@@ -1425,7 +1451,7 @@ class logical_operand : public base_statement
     //return out;
     return std::string("#");//TBD
   }
-  virtual value& eval()
+  virtual value& eval_internal()
   {
     bool res;
     if (!l || !r)
@@ -1503,7 +1529,7 @@ class mulldiv_operation : public base_statement
     return std::string("#");//TBD
   }
 
-  virtual value& eval()
+  virtual value& eval_internal()
   {
     switch (_mulldiv)
     {
@@ -1579,7 +1605,7 @@ class addsub_operation : public base_statement
     return std::string("#");//TBD
   }
 
-  virtual value& eval()
+  virtual value& eval_internal()
   {
     if (_op == addsub_op_t::NA) // -num , +num , unary-operation on number
     {
@@ -1633,7 +1659,7 @@ class negate_function_operation : public base_statement
     return function_to_negate;
   }
 
-  virtual value& eval()
+  virtual value& eval_internal()
   {
     res = function_to_negate->eval();
 

test/s3select_test.cpp

@@ -2352,7 +2352,7 @@ TEST(TestS3selectFunctions, nested_call_aggregate_with_non_aggregate)
     //purpose: test projections with aggregation functions, in mix of nested calls. 
     s3select s3select_syntax;
 
-    const std::string input_query = "select substring('abcdefghijklm',(2-1)*3+sum(cast(_1 as int))*0+1,2*count(0)/count(0)) ,count(0),sum(cast(_1 as int))*0 from stdin;";
+    const std::string input_query = "select substring('abcdefghijklm',(2-1)*3+sum(cast(_1 as int))*0+1,(count() + count(0))/count(0)) ,count(0),(sum(cast(_1 as int))*min(cast(_2 as int)))*0 from stdin;";
 
     auto status = s3select_syntax.parse_query(input_query.c_str());
     ASSERT_EQ(status, 0);