DM-21355: Try to prevent NaN values turning up in FITS headers #488
Conversation
timj
force-pushed
the
tickets/DM-21355
branch
from
133f5af
to
d3321da
Compare
| boost::format("In %s, found NaN in metadata item '%s'") % | ||
| BOOST_CURRENT_FUNCTION % name); | ||
| // Convert it to FITS undefined | ||
| out += " "; |
There was a problem hiding this comment.
Is this FITS undefined, or is it "BLANK "? https://heasarc.gsfc.nasa.gov/docs/fcg/standard_dict.html
It could be that there's another translation layer somewhere else in the bowels of afw, or that I'm misreading something.
There was a problem hiding this comment.
FITS undefined (there is a keyword and an = sign)
| BOOST_CURRENT_FUNCTION % name); | ||
| // Convert it to FITS undefined | ||
| out += " "; | ||
| } | ||
| } else if (type == typeid(std::nullptr_t)) { | ||
| out += " "; |
There was a problem hiding this comment.
Hmmm, I guess this is what has always been used in afw, so I clearly don't have the full picture.
There was a problem hiding this comment.
I added support for FITS undefined a few months back.
| # This code causes a SEGV without DM-21355 implemented | ||
| frameSet1 = readFitsWcs(metadata, strip=False) | ||
| self.assertEqual(type(frameSet1), ast.FrameSet) | ||
|
|
There was a problem hiding this comment.
It makes sense that this test SIGABRTs on master, this was an untested edge case that could lead to an SIGABRT and it should now be tested (and fixed).
NaN is not allowed in a FITS header so we convert it to an undefined value.
timj
force-pushed
the
tickets/DM-21355
branch
from
d3321da
to
75ba42d
Compare
The FITS standard doesn't allow NaN as a value. NaNs turning up in a PropertyList can confuse things so this change traps this and either drops the entire item or replaces with FITS undefined value.