Using TFITSHDU to read a Table with arrays as column elements

[cosimoNigro]

Dear ROOTers,

I am trying to use the TFITSHDU class to read the Table in the Header Data Unit (HDU) of a FITS file.
The TFITSHDU class works fine for images and plain table, but it appears unable to support a table
whose columns contain arrays rather than simple floats.

A bit of background: the data I am trying to read (a picture follows) represent a 2-dimensional PDF of an energy estimator. Since it is mostly 0, a sparse matrix is stored, with the columns F_CHAN and N_CHAN declaring the non empty bins and the MATRIX column stating the PDF values in them.

If you want to repeat the test, you can find the file here

open-gamma-ray-astro/joint-crab/blob/master/results/spectra/magic/rmf_obs5029747.fits

Already when I open it I get a warning

root [0] TFITSHDU *hdu = new TFITSHDU("rmf_obs5029747.fits[1]")
Warning in <TFITSHDU::LoadHDU>: error opening FITS file. Column type -21 is currently not supported
Warning in <TFITSHDU::LoadHDU>: error opening FITS file. Column type -21 is currently not supported
Warning in <TFITSHDU::LoadHDU>: error opening FITS file. Column type -42 is currently not supported
(TFITSHDU *) 0x7fe222c607b0

There are no problems in reading plain columns, e.g. the ENERG_LO one

root [1] hdu->GetTabRealVectorColumn("ENERG_LO")->Print()

Vector (80)  is as follows

     |        1  |
------------------
   0 |1e+07 
   1 |1.12202e+07 
   2 |1.25893e+07 
   3 |1.41254e+07 
   4 |1.58489e+07 
   5 |1.77828e+07 

The problem is the TVectorD returned by GetTabRealVectorColumn cannot handle the arrays stored in the entries of the, for example, F_CHAN column

root [2] hdu->GetTabRealVectorColumn("F_CHAN")->Print()

Vector (80)  is as follows

     |        1  |
------------------
   0 |0 
   1 |0 
   2 |0 
   3 |0 
   4 |0 
   5 |0 

as you can see the array nature of each value of the F_CHAN column is completely ignored.

I would like to use TFITSHDU as much as possible, but I am afraid that for this problem I need to go back to the CFITSIO libraries.

Is it correct to say that reading such files with the current TFITS implementation is not possible?

NOTE: I cannot change the input file as this is an established data format used by X-ray and Gamma-ray astronomical instruments.

Thank you,

Cosimo

---

ROOT Version: any version > 6
Platform: platform-independent
Compiler: compiler-independent

---

[Axel]

We really rely on external contributions with FITS. Your post tells me that there is a limitation - we can certainly accept a patch to fix that limitation! Would you be interested in giving it a try?

Cheers, Axel.

[cosimoNigro]

Hello @Axel,

Yes I can give it a try, I never contributed to ROOT, though.
From your contributors guideline it seems that you follow the normal git workflow. So can I just try a patch and submit a pull request, or does it have to be discussed someplace else first?

Thank you,

Cosimo

[Axel]

So can I just try a patch and submit a pull request,

Yes! Looking forward to that!

It would be great if you could add a test, too - we have a tutorial which serves a dual purpose as a test, but you’re welcome to create a Google test under graph2d/fitsio/test, see e.g. what’s in hist/hist/test/.

Cheers, Axel.