[fitsbits] 64-bit integer comments

Mike Nolan writes:
I guess that my point is that if you wait for the need, it will come
and go before you can use it. If the solution is clear and much of the
software implements it anyway, it's time to do it; particularly when
there are a few decisions to make (e.g. signed or not) that can be
worked around either way, but not if half choose one and half the
other.

It looks like we are reaching a philosophical disagreement on whether
64-bit integers should be called overengineering or good planning for
the future. We understand each other's point I think, but just don't
give them identical weights. Part of that might be people who use
FITS as their internal format versus those who use for transport
only.

I don't see how we are going to bridge that difference (unless someone
comes up with the killer application that does need I*8), so that particular
issue might have to be resolved by vote rather than consensus.

Mark Calabretta wrote:
According to the standard "A free format integer value follows the same
rules as fixed format integers [i.e. right-justified in cols 11-30]
except that it may occur anywhere within columns 11-80." It doesn't
actually say anything about the number of digits that are or are not
allowed, but as you say it could well be inferred that up to 70 digits
are permitted, and there are plausible examples where such might occur,
e.g. public keys and the like. Presumably no current software can
handle such keywords properly.

However, I was really suggesting that the standard should warn
unsuspecting implementors of this surprising feature, not tell them how
to implement it.

There have been many other suggestions for clarifying various sections of
the FITS Standard since the last update in 1999, so we should probably start
thinking about convening another technical revision panel to address all of
these issues at one time. (Maybe next year, after the current round of WCS
and 64-bit integer proposals have been dealt with?)

more aware of the need to support larger values. FITS interface software
like CFITSIO already supports reading FITS keywords into 64-bit int
variables, so there is no technical reason to prevent generic software from
always reading integer keywords into 64-bit ints. As an example, I recently


I looked for the corresponding functions in v2.500 but only found the
'long int' forms. Does it rely on long int being 64-bit?

No, CFITSIO only requires that the C compiler support a 64-bit integer data
type, but it could be called 'longlong', 'long', '__int64', or something
else, depending on the particular compiler/platform. When calling the
general keyword reading routines in CFITSIO, one needs to specify the data
type of the variable into which the keyword value is to be returned. The
constant 'TLONGLONG' should be used to specify the 64-bit integer data type.
CFITSIO also defines and uses 'LONGLONG' as a machine independent 64-bit
integer data type. It is 'typedefed' to be equivalent to whatever intrinsic
64-bit integer data type is available on each particular platform.

Bill Pence
--
__________________________________________________ __________________
Dr. William Pence
NASA/GSFC Code 662 HEASARC +1-301-286-4599 (voice)
Greenbelt MD 20771 +1-301-286-1684 (fax)