[GLUTEN-4898][CH]Bug fix to date diff (apache#4900)

What changes were proposed in this pull request?
(Please fill in changes proposed in this fix)

(Fixes: apache#4898)

Fix diff problem of to_date function;
Fix exception throws from parseDateTimeBestEffort of SparkFunctionToDateTime;
simplify substring function code
How was this patch tested?
TEST BY UT

backends-clickhouse/src/test/scala/io/glutenproject/execution/GlutenClickHouseTPCHParquetSuite.scala

@@ -2103,7 +2103,10 @@ class GlutenClickHouseTPCHParquetSuite extends GlutenClickHouseTPCHAbstractSuite
         |(4, '2023-09-02 00:00:01.333-11'),
         |(5, '  2023-09-02 agdfegfew'),
         |(6, 'afe2023-09-02 11:22:33'),
-        |(7, '1970-01-01 00:00:00')
+        |(7, '1970-01-01 00:00:00'),
+        |(8, '2024-3-2'),
+        |(9, '2024-03-2'),
+        |(10, '2024-03')
         |""".stripMargin
     spark.sql(create_table_sql)
     spark.sql(insert_data_sql)

cpp-ch/local-engine/Functions/SparkFunctionToDate.cpp

@@ -22,6 +22,7 @@
 #include <Functions/FunctionFactory.h>
 #include <IO/ReadBufferFromMemory.h>
 #include <IO/ReadHelpers.h>
+#include <IO/parseDateTimeBestEffort.h>
 
 namespace DB
 {
@@ -50,7 +51,9 @@ class SparkFunctionConvertToDate : public DB::FunctionToDate32OrNull
             for (size_t i = start; i < start + length; ++i)
             {
                 if (!isNumericASCII(*(rb.position() + i)))
+                {
                     return false;
+                }
             }
             return true;
         };
@@ -63,7 +66,7 @@ class SparkFunctionConvertToDate : public DB::FunctionToDate32OrNull
         };
         if (!checkNumbericASCII(buf, 0, 4) 
             || !checkDelimiter(buf, 4) 
-            || !checkNumbericASCII(buf, 5, 2) 
+            || !checkNumbericASCII(buf, 5, 2)
             || !checkDelimiter(buf, 7) 
             || !checkNumbericASCII(buf, 8, 2))
             return false;
@@ -112,19 +115,19 @@ class SparkFunctionConvertToDate : public DB::FunctionToDate32OrNull
             throw DB::Exception(DB::ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Function {}'s return type must be date32.", name);
 
         using ColVecTo = DB::DataTypeDate32::ColumnType;
-        typename ColVecTo::MutablePtr result_column = ColVecTo::create(size);
+        typename ColVecTo::MutablePtr result_column = ColVecTo::create(size, 0);
         typename ColVecTo::Container & result_container = result_column->getData();
-        DB::ColumnUInt8::MutablePtr null_map = DB::ColumnUInt8::create(size);
+        DB::ColumnUInt8::MutablePtr null_map = DB::ColumnUInt8::create(size, 0);
         typename DB::ColumnUInt8::Container & null_container = null_map->getData();
-        const DateLUTImpl * time_zone = &DateLUT::instance();
+        const DateLUTImpl * local_time_zone = &DateLUT::instance();
+        const DateLUTImpl * utc_time_zone = &DateLUT::instance("UTC");
 
         for (size_t i = 0; i < size; ++i)
         {
             auto str = src_col->getDataAt(i);
-            if (str.size < 10)
+            if (str.size < 4)
             {
                 null_container[i] = true;
-                result_container[i] = 0;
                 continue;
             }
             else
@@ -134,20 +137,17 @@ class SparkFunctionConvertToDate : public DB::FunctionToDate32OrNull
                 {
                     buf.position() ++;
                 }
-                if(buf.buffer().end() - buf.position() < 10)
+                if(buf.buffer().end() - buf.position() < 4)
                 {
                     null_container[i] = true;
-                    result_container[i] = 0;
                     continue;
                 }
-                if (!checkAndGetDate32(buf, result_container[i], *time_zone))
-                {
-                    null_container[i] = true;
-                    result_container[i] = 0;
-                }
-                else
+                if (!checkAndGetDate32(buf, result_container[i], *local_time_zone))
                 {
-                    null_container[i] = false;
+                    time_t tmp = 0;
+                    bool parsed = tryParseDateTimeBestEffort(tmp, buf, *local_time_zone, *utc_time_zone);
+                    result_container[i] = local_time_zone->toDayNum<time_t>(tmp);
+                    null_container[i] = !parsed;
                 }
             }
         }

cpp-ch/local-engine/Functions/SparkFunctionToDateTime.h

@@ -163,8 +163,8 @@ class SparkFunctionConvertToDateTime : public DB::FunctionToDateTime64OrNull
             }
             else
             {
-                parseDateTime64BestEffort(dst_data[i], scale, buf, *local_time_zone, *utc_time_zone);
-                null_map_data[i] = 0;
+                bool parsed = tryParseDateTime64BestEffort(dst_data[i], scale, buf, *local_time_zone, *utc_time_zone);
+                null_map_data[i] = !parsed;
             }
         }
     }

cpp-ch/local-engine/Parser/scalar_function_parser/substring.cpp

@@ -51,15 +51,14 @@ class FunctionParserSubstring : public FunctionParser
          /**
             parse substring(str, start_index, length) as
             if (start_index == 0)
-                substring(str, start_index+1, length)
+                substring(str, 1, length)
             else
                 substring(str, start_index, length)
         */
         auto * const_zero_node = addColumnToActionsDAG(actions_dag, start_index_data_type, Field(0));
         auto * const_one_node = addColumnToActionsDAG(actions_dag, start_index_data_type, Field(1));
         auto * equals_zero_node = toFunctionNode(actions_dag, "equals", {parsed_args[1], const_zero_node});
-        auto * index_plus_node = toFunctionNode(actions_dag, "plus", {parsed_args[1], const_one_node});
-        auto * if_node = toFunctionNode(actions_dag, "if", {equals_zero_node, index_plus_node, parsed_args[1]});
+        auto * if_node = toFunctionNode(actions_dag, "if", {equals_zero_node, const_one_node, parsed_args[1]});
         const DB::ActionsDAG::Node * substring_func_node;
         if (parsed_args.size() == 2)
             substring_func_node = toFunctionNode(actions_dag, "substringUTF8", {parsed_args[0], if_node});