return null for wrong array

Author: morningman

be/src/vec/data_types/data_type_array.cpp

@@ -21,10 +21,10 @@
 #include "vec/data_types/data_type_array.h"
 
 #include "gen_cpp/data.pb.h"
+#include "util/stack_util.h"
 #include "vec/columns/column_array.h"
 #include "vec/columns/column_nullable.h"
 #include "vec/data_types/data_type_nullable.h"
-#include "util/stack_util.h"
 
 namespace doris::vectorized {
 
@@ -185,10 +185,13 @@ Status DataTypeArray::from_string(ReadBuffer& rb, IColumn* column) const {
         return Status::InvalidArgument("Array does not start with '[' character, found '{}'",
                                        *rb.position());
     }
+    if (*(rb.end() - 1) != ']') {
+        return Status::InvalidArgument("Array does not end with ']' character, found '{}'",
+                                       *(rb.end() - 1));
+    }
     ++rb.position();
     bool first = true;
     size_t size = 0;
-    bool find_end = false;
     while (!rb.eof() && *rb.position() != ']') {
         if (!first) {
             if (*rb.position() == ',') {
@@ -201,7 +204,6 @@ Status DataTypeArray::from_string(ReadBuffer& rb, IColumn* column) const {
         }
         first = false;
         if (*rb.position() == ']') {
-            find_end = true;
             break;
         }
         size_t nested_str_len = 0;
@@ -257,9 +259,6 @@ Status DataTypeArray::from_string(ReadBuffer& rb, IColumn* column) const {
         DCHECK_LE(rb.position(), rb.end());
         ++size;
     }
-    if (!(!rb.eof() && *rb.position() != ']') && !find_end) {
-        return Status::InvalidArgument("Array does not start with ']' character");
-    }
     offsets.push_back(offsets.back() + size);
     return Status::OK();
 }

be/src/vec/exec/format/parquet/schema_desc.cpp

@@ -164,7 +164,7 @@ TypeDescriptor FieldDescriptor::get_doris_type(const tparquet::SchemaElement& ph
         type = convert_to_doris_type(physical_schema.converted_type);
     }
     // use physical type instead
-    if (type.type ==INVALID_TYPE) {
+    if (type.type == INVALID_TYPE) {
         switch (physical_schema.type) {
         case tparquet::Type::BOOLEAN:
             type.type = TYPE_BOOLEAN;

be/src/vec/functions/function_cast.h

@@ -320,10 +320,12 @@ struct ConvertImplGenericFromString {
                     check_and_get_column<StringColumnType>(&col_from)) {
             auto col_to = data_type_to->create_column();
 
-            //IColumn & col_to = *res;
             size_t size = col_from.size();
             col_to->reserve(size);
 
+            ColumnUInt8::MutablePtr col_null_map_to = ColumnUInt8::create(size);
+            ColumnUInt8::Container* vec_null_map_to = &col_null_map_to->get_data();
+
             for (size_t i = 0; i < size; ++i) {
                 const auto& val = col_from_string->get_data_at(i);
                 // Note: here we should handle the null element
@@ -332,9 +334,16 @@ struct ConvertImplGenericFromString {
                     continue;
                 }
                 ReadBuffer read_buffer((char*)(val.data), val.size);
-                RETURN_IF_ERROR(data_type_to->from_string(read_buffer, col_to));
+                Status st = data_type_to->from_string(read_buffer, col_to);
+                // if parsing failed, will return null
+                (*vec_null_map_to)[i] = !st.ok();
+                if (!st.ok()) {
+                    col_to->insert_default();
+                }
             }
-            block.replace_by_position(result, std::move(col_to));
+            // block.replace_by_position(result, std::move(col_to));
+            block.get_by_position(result).column =
+                    ColumnNullable::create(std::move(col_to), std::move(col_null_map_to));
         } else {
             return Status::RuntimeError(
                     "Illegal column {} of first argument of conversion function from string",
@@ -870,8 +879,9 @@ struct ConvertThroughParsing {
         if constexpr (IsDataTypeDecimal<ToDataType>) {
             UInt32 scale = additions;
             col_to = ColVecTo::create(size, scale);
-        } else
+        } else {
             col_to = ColVecTo::create(size);
+        }
 
         typename ColVecTo::Container& vec_to = col_to->get_data();