On Tue, Nov 13, 2012 at 12:16 AM, Willy-Bas Loos <willybas@xxxxxxxxx> wrote: > On Fri, Sep 9, 2011 at 5:09 PM, Guillaume Lelarge <guillaume@xxxxxxxxxxxx> > wrote: >> >> You divide an integer with an integer, that should give you an integer. > > Can you tell me the reasoning behind that idea? > Is it a rule that the output type of an operator must equal the input type? > In this case that doesn't seem locigal. I think that the "/" operator should > return something that allows fractions, since the operator creates fractions > so frequently. This is an argument that comes up regularly on the Python list, partly because version 2 had int/int -> int, but version 3 declared that int/int -> float made more sense. One of the problems of going to floating point is that it's not a superset of integers - especially not when your integer type supports arbitrary precision. It might seem obvious that 7/2 should yield 3.5 and not 3, but what about when the numbers are so large that you lose precision by going float? Or are there to be some cases where int/int makes float and some where it makes int? That would be nicely confusing. I'm generally happy with either behaviour, as long as its consistent, and as long as it can be overridden with an explicit type cast when the other is needed. ChrisA -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
