On Mon, Aug 26, 2013 at 2:03 PM, Roman Fleysher <roman.fleys...@einstein.yu.edu> wrote:
Certainly associating a type conversion with a lexical instance of an operator, applying to all of the operator's operands (or perhaps even just to some, while having to explicitly cast others) would work, at least for operations which can only be accessed via the given syntactic operator. But while some implicit type conversions are useful and handy, I don't think *collations* can be implied, really, at least not if there are no constant literals around (and we have no syntax for associating collations with those!)... Now, perhaps for most operators we can't easily pick one operand to supply type information for implicit casting of the other(s), but for BETWEEN you might argue that the type of the first operand should be used to inform type conversion of the second and third operands. That would work, and it'd be easy to remember even. But conceptually even then the collation is best thought of as an attribute of types. For string values changing their collation is easy, thankfully: it's just an attribute of the string. And most systems don't keep track of collations (or lots of other info) associated with specific strings -- this includes SQLite, IIRC. In such systems one workaround for the lack of compile- and/or run-time sting collation information... would be to specify the collation syntactically. So you have a point, but I think the point really is that an RDBMS should keep track of column and string collations so as to avoid having to require lexical clues for collation. Of course, lacking a syntax for associating collations with string literals there will be times when some, or even all of the operands to an operation that needs collation information, is missing. In such cases the collation information of some operands could be used to cast the others, or lexical clues might be needed, or a default might be provided. So, in short, I think you and James are both right, and maybe we're squaring a circle. But still, notionally I'm with James on this: collation is an attribute of type. Nico -- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
