Enhance minmax index to support more data types.

[hongyunyan]

Enhancement

Our minmax index now is only support for the handle column, and the columns whose data type satisfies isInteger() or isDateorDateTime()).

For lots of other data types, such as float, double, decimal, string, etc, minmax index is unsupported currently.

In order to enhance our index ability, we first can enhance our minmax index for more data types, to utilize our minmax index better.

[hongyunyan]

/assign

[JaySon-Huang]

For float, double - Comparing floating-point values is not an accurate operation, can we ensure it won't filter the wrong result? How does other database systems handle that? I think when comparing floating-point column to a value, the executor usually cast the floating-point value to a decimal value for comparing, which could make the min-max index of floating-point invalid.

For string

• The "min" "max" of random long string is usually useless for filtering. Unless we can ensure the strings are low-cardinality
• Generating the min max value of string column could slow down the write speed a lot
• We need to pay attention for the collation of string content when using it, which makes it difficult for comparing

There are more kinds of rough set indexes (or some systems call them zone-maps) other than min-max that could help us better filter the irrelevant data when executing queries.