On Oct 28, 2005, at 9:23 , Trent Shipley wrote:
On Thursday 2005-10-27 16:22, Andrew Dunstan wrote:
Jim C. Nasby wrote:
Like I said, if we're going to support a concept of ordering of
items in
an enum then we need to support it fully. For starters that means
having
the ability to re-order things in an enum seamlessly.
I do not see this at all. An enumeration defines an ordering and a
set
of labels. Why should you be able to change it? If you want a
different
ordering, create a new enumeration. Let's do this right because
it's a
feature worth having, not just mimic the competition's idiocy
The symbols in the set have no _per se_ order.
A collation rule is necessary to sort the symbols consistently.
ASCII is an enumeration
Unicode is a large enumeration with a simple naive collation and a
complex
default collation.
Defining a set results in an unordered specification of symbols.
Defining a collation produces an ordering for the set.
There can be many collations for a set.
An enumeration is just a computer science short-hand way to define
a set and a
"native" collation for the set.
An enumeration's native collation need not be the only, or even the
most
common, collation for the enumerated set of symbols.
Relational databases already have a type for unordered sets: tables.
IMO, if there's going to be a separate enumerated type, it should be
more than just an alternative way of defining a set of key-value pairs.
Michael Glaesemann
grzm myrealbox com
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match