MathGroup Archive 2013

[Date Index] [Thread Index] [Author Index]

Search the Archive

FITS fits and workarounds

Mathematica 9.0.1 now provides extended FITS file import and export functionality, including arbitrary header fields.  However, there is a bug and a hiccup.  I've submitted the bug report to Wolfram support.  Meanwhile, here are the issues and workarounds:


To demonstrate the bug, we'll generate a FITS file as follows:

Here's a simple 3D image cube as the Data Unit.

img = Table[RandomReal[{0, 10^9}], {2}, {32}, {32}];

Here's metadata that will become additional Header Unit content.  We will generate n additional keys, where depending on the value of n, the Data Unit will be stepped upon by a faultily written Header.  Mathematica  automatically generates 13 header keys upon Export.  Each key/value/comment triple occupies an 80-byte "card".  The header must be in multiples of 2880 byte "blocks" (equivalent to 36 cards), so padding (ascii char 20) must conditionally be added after the card entries to make this so. 

Evidently Mathematica's padding logic is broken for the case where exactly 36 cards (n=23 user + 13 automatic) go into the header.  This can be verified by executing the code below, and examining the resulting 3 files (say with DS9 FITS viewer). 

One can vary the value of n below and see what happens for n<23, n=23 (busted) and n>23. 

n = 23;

metadata = Table["KEY" <> ToString[i] -> {i, "comment"}, {i, 1, n}];

fname = "test.fits";

Export[fname, {img, metadata}, {{"Data", "Metadata"}}];

So for the moment, the workaround is to simply avoid having a multiple of 36 header keys. 


Certain float values generate expressions rejected by FITS Export.  So pre-filter your key/value/comment triples as follows:

formatKeyValue[x_Real] := N[FromDigits[RealDigits[x, 10, 6]]]; 
formatKeyValue[x_] := x;

metadata = (Rule[#[[1]], {formatKeyValue[#[[2, 1]]], #[[2, 2]]}]) & /@ metadata;

  • Prev by Date: Re: ComplexExpand off by a factor of (-1)?
  • Next by Date: Re: Mathematica and Lisp
  • Previous by thread: Re: Compiling numerical iterations
  • Next by thread: Hold & Evaluate