[Issues] ldacconv and floating point vs integer

Hans Vaith hav at mpe.mpg.de
Thu Jul 12 18:53:19 CEST 2007

Dear All,

I ran into a problem when doing Astrometry on LBCBLUE data. The problem is 
caused by a mismatch between how the ldac fits utility function 'fitspick' 
distinguishes between floats and integers, and how floating point numbers are 
(sometimes) specified in the FITS headers of LBCBLUE data.

'fitspick' uses the presence or absence of a dot (.) as a criterion to 
determine if a number in FITS header keyword/value pair is a float or an 
integer. The LBCBLUE FITS headers (and also the catalogs 'sex.cat' created 
from them) unfortunately sometimes contain numbers that are not integers but 
lack the dot. An example:

CD1_2   =               -7e-08 / no comment
CD2_1   =               -7e-08 / no comment

The missing dot causes the numbers to be stored as integers (with a value 
of -7). In the example case preastrom then complains that the values of the 
CDi_j keywords are expected to be floats.

I don't know if the formal floating point specification requires a dot as part 
of the number. I can see two ways out of this situation:

Either make sure (prior to data ingestion) that the LBCBLUE FITS headers 
always use dots with floating point numbers, or modify the fitspick function 
within ldac. I could look into the latter, but will do so only if you think 
this is the way to go. Please let me know your preferred solution. Thanks.


More information about the Issues mailing list