Could custom types benefit significantly from custom operators as well? Do custom C functions stand a good chance of introducing speed benefits over their raw SQL or pl/sql counterparts? Or is the field too broad to speculate on the general case? The scenario that inspired this question was about data that had to be stored accurately, but the data itself wasn't usually precise. You could think of an individual datum being a set of ranges. You could definitely define equality on this data type, but the ordering operators would probably be meaningless. On the other hand, some (but not all) of the geometric operators could probably be interpreted to apply to this data set, as long as you ignore the 'above' and 'below' semantics, and replace the concepts of left and right with less than and greater than. So for example, while << (is strictly left of) Wouldn't make sense, using the same operator to mean 'strictly less than' might. Would R-tree indexes be useful for a data type like this? Would it be possible to define the base type such that an R-tree index would always be created? Once again - this is entirely idle curiosity. This isn't anything I have a real need for. -----Original Message----- From: pgsql-general-owner@xxxxxxxxxxxxxx [mailto:pgsql-general-owner@xxxxxxxxxxxxxx] On Behalf Of Tom Lane Sent: Wednesday, June 07, 2006 9:37 AM To: Tim Hart Cc: pgsql-general@xxxxxxxxxxxxxx Subject: Re: [GENERAL] What are the characteristics of a good user-defined data type? "Tim Hart" <tjhart@xxxxxxx> writes: > I've reviewed the PostgreSQL documentation on user defined data types. It > seems that data types that can't be ordered or compared for equality would > be bad candidates. After all, if a data type can't be indexed or used in a > where clause, what value does a custom type bring over a binary or textual > representation? Well, the possibility of error-checking for bad values might alone justify a custom type, depending on what you're doing. A type with no support beyond the required I/O functions could offer that. But it's kinda hard to imagine a datatype in which there is no meaningful way to define equality ... regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 4: Have you searched our list archives? http://archives.postgresql.org
