Item_func_between::fix_length_and_dec has the following code:
if (m_comparator.cmp_type() == STRING_RESULT &&
|
agg_arg_charsets_for_comparison(cmp_collation, args, 3))
|
return;
|
|
/* See the comment about the similar block in Item_bool_func2 */
|
if (args[0]->real_item()->type() == FIELD_ITEM &&
|
!thd->lex->is_ps_or_view_context_analysis())
|
{
|
Item_field *field_item= (Item_field*) (args[0]->real_item());
|
if (field_item->field_type() == MYSQL_TYPE_LONGLONG ||
|
field_item->field_type() == MYSQL_TYPE_YEAR)
|
{
|
const bool cvt_arg1= convert_const_to_int(thd, field_item, &args[1]);
|
const bool cvt_arg2= convert_const_to_int(thd, field_item, &args[2]);
|
if (cvt_arg1 && cvt_arg2)
|
{
|
// Works for all types
|
m_comparator.set_handler(&type_handler_longlong);
|
}
|
}
|
}
|
It's not friendly for pluggable data type. For example, the INET6 data type will most likely have STRING_RESULT, but it won't need character set and collation aggregation.
We'll introduce Type_handler::Item_func_between_fix_length_and_dec() and split this code.