GammaLib

Here is the current error message:

>>> import gammalib
>>> b=gammalib.GCTABackground3D('hess_bkg_example.fits')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/gamma/lib/python2.7/site-packages/gammalib/cta.py", line 4238, in __init__
    this = _cta.new_GCTABackground3D(*args)
ValueError: *** ERROR in GCTAResponseTable::read_tables(GFitsTable&): Invalid value. Table column "NEVENTS" contains 20 elements while 50000 elements were expected from the axis definitions. Please check the response table for consistency.
>>> 

This was a tricky one since the exception is thrown at the level of the GCTAResponseTable::read() method which does not know anything about the file structure since this is a generic method. Removing the exception at that level remove the check for the file structure consistency. If for some reason the background matrix would be incompatible with the axis definitions, the class would simply ignore the matrix, and GCTABackground3D would pretend that the matrix was not found, although it exists in the file.

I therefore implemented a kludge at the level of the GCTABackground3D::read() where the expected file structure is known:

// Copy response table and skip all columns that are not mandatory. This
// kludge is needed to get rid of additional columns in the HESS
// background model.
GFitsTable *ptr = table.clone();  // Make sure that table is of same type
for (int i = 0; i < table.ncols(); ++i) {
    std::string colname = table[i]->name();
    if ((colname != "DETX_LO")  && (colname != "DETX_HI")  &&
        (colname != "DETY_LO")  && (colname != "DETY_HI")  &&
        (colname != "ENERG_LO") && (colname != "ENERG_HI") &&
        (colname != "BKG")      && (colname != "BGD")) {
        ptr->remove(colname);
    }
}

// Read background table from reduced FITS table
m_background.read(*ptr);

// Delete reduced FITS table
delete ptr;

I also added a protected GFitsTable::update_header() method that makes sure that the header is up to date after any row or column operations. Before there was a risk for a discrepancy. Hence FITS column manipulation is now much cleaner.

Code merged into devel.