We have at least two implementations of string-to-decimal conversion:

• dataconvert::number_int_value(), which is used either directly or inside the wrapper SystemCatalog::TypeAttributesStd::decimal128FromString().
• Func_cast_decimal::getDecimalVal()

These two implementations have a lot of duplicate code and have problems:

• MCOL-4462 - This problem existed for a long time
• MCOL-4500 - This problem was introduced by MCOL-4464
• MCOL-4532 - Introduced by MCOL-641
• They use a lot of std::string objects copying and padding, which is not performance friendly.

Under terms of this task we'll do the following:

• Introduce a new implementation of a string-to-decimal conversion function which won't have the mentioned bugs
• The new function should not do any std::string copying/padding and should only use the source string (and therefore should hopefully be faster). It should only iterate on the source string.
• The new function should get a const char * pointer and its length instead of a '\0' terminated C-style string. This is to avoid calling string::c_ptr() which is also not performance friendly.
• Add a new constructor:

  Decimal::Decimal(const char *str, size_t length, DataCondition & convError,

  int8_t s, uint8_t p)

  which will use the new conversion function internally.

• Replace the code in GenericToBitOperand() in func_bit.cc to the new function call
• Replace the code in Func_cast_decimal::getDecimalVal() to thew new constructor call

Further steps

Note, after these changes, the functions dataconvert::number_int_value() and SystemCatalog::TypeAttributesStd::decimal128FromString() will still exist.

Later, under terms of a separate tasks (or even multiple tasks), we'll:

• Replace all remaining calls for dataconvert::number_int_value() to the new functionality.
• Replace all remaining calls for SystemCatalog::TypeAttributesStd::decimal128FromString() to the new functionality.
• Remove dataconvert::number_int_value() and SystemCatalog::TypeAttributesStd::decimal128FromString() completely.