Re: Problems with infinity

On Thursday 13 January 2005 15:18, Kris Jurka wrote:
> On Thu, 13 Jan 2005, Oliver Siegmar wrote:
> > On Thursday 13 January 2005 14:42, Kris Jurka wrote:
> > > This is indeed a bug in the 7.4 drivers.
> >
> > Will you fix it? ;-)
>
> Sure, but you should be aware that the driver only does this conversion on
> output (rs.getTimestamp()). No conversion is done for setTimestamp, so
> it's a one way solution. This was part of my reasoning for removing it.

Of course it should work in both directions.

> > Couldn't you make it configurable (like 'convert_infinite_to_min_max =
> > true')?
>
> That could be done, but how do you define min/max. There are at least
> three possible definitions I can think of: Java's min/max value for
> Timestamp and pg's min/max, but pg's depends on how it was compiled
> (--enable-integer-datetimes). So how do you pick which one? You've got
> to pick the one with the smallest range, so it fits inside the other's
> range, which kind of sucks because then you have a special magic value
> that isn't even at the extremes of the type range. The user level
> code to manipulate such a thing will certainly be ugly.

Well, I'd set Long.MIN_VALUE / Long.MAX_VALUE to the Timestamp - the pgsql
driver could simply convert this to '-infinity' / 'infinity' strings, not?
This could be done for both directions.

I agree, that these are special magic values but its better than someone can
break down my java applications by inserting a infinity value to a timestamp
field.

What do you think?

Cheers,
Oliver