Re: problem with float8 input format

Louis-David Mitterrand <cunctator(at)apartia(dot)ch> writes:
> where "date_part" comes from "date_part('epoch', stopdate)" in a
> previous query. The problem is the value of stop_date is not the number
> of seconds since the epoch but some internal representation of the data.
> So I can't compare stop_date with the output of
> GetCurrentAbsoluteTime().

GetCurrentAbsoluteTime yields an "abstime", so you should coerce the
"timestamp" result of date_part() to abstime and then you will get a
value you can compare directly.

> What function should I use to convert the Datum to a C int?
> DatumGetInt32 doesn't seem to work here.

No, because timestamps are really floats. (abstime is an int though.)

> And what is the method for float8 Datum conversion to C double?

double x = * DatumGetFloat64(datum);

This is pretty grotty because it exposes the fact that float8 datums
are pass-by-reference (ie, pointers). 7.1 will let you write

double x = DatumGetFloat8(datum);

which is much cleaner. (I am planning that on 64-bit machines it will
someday be possible for float8 and int64 to be pass-by-value, so it's
important to phase out explicit knowledge of the representation in user
functions.)

regards, tom lane