Make GCTAPsfKing save against invalid values (#2133)

Jürgen Knödlseder
1 parent 32017030
Showing 10 changed files with 63 additions and 196 deletions
ChangeLog
NEWS
README.md
configure.ac
doc/Doxyfile
doc/source/conf.py
gammalib.pc.in
inst/cta/src/GCTAPsfKing.cpp
inst/cta/src/GCTAResponseIrf.cpp
sonar-project.properties
+2017-06-26
+
+        * Bug fix version 1.3.1 released
+        ================================
+
+        Make GCTAPsfKing save against invalid values (#2133)
+
+
 2017-06-06
  
         * Version 1.3.0 released
-New Features and Important Changes in GammaLib 1.3.0
+New Features and Important Changes in GammaLib 1.3.1
  
-6 June 2017
+26 June 2017
  
  
 1. Introduction
@@ -12,15 +12,7 @@ made since the last release of GammaLib.
 2. Public interface modifications
 ---------------------------------
 The following classes have been added:
-- GPhases
-- GWcsGLS
-- GWcsSFL
-- GModelSpectralSmoothBrokenPlaw
-- GModelSpatialRadialProfile
-- GModelSpatialRadialProfileGauss
-- GModelSpatialRadialProfileDMBurkert
-- GModelSpatialRadialProfileDMEinasto
-- GModelSpatialRadialProfileDMZhao
+- n.a.
  
 The following classes have been removed:
 - n.a.
@@ -29,48 +21,13 @@ The following classes have been renamed:
 - n.a.
  
 The following methods have been added:
-- GEbounds::set(GEnergies&)
-- GEbounds::emin(int&, GEnergy&)
-- GEbounds::emax(int&, GEnergy&)
-- GEnergies::set(GEbounds&)
-- GFilename::type()
-- GMatrix::operator+(double&)
-- GMatrix::operator-(double&)
-- GMatrix::operator+=(double&)
-- GMatrix::operator-=(double&)
-- GMatrixSymmetric::operator+(double&)
-- GMatrixSymmetric::operator-(double&)
-- GMatrixSymmetric::operator+=(double&)
-- GMatrixSymmetric::operator-=(double&)
-- GMatrixSparse::operator+(double&)
-- GMatrixSparse::operator-(double&)
-- GMatrixSparse::operator+=(double&)
-- GMatrixSparse::operator-=(double&)
-- GCTAResponseIrf::rspname()
-- GCTARoi::contains(GCTAInstDir&)
-- GCTAEdisp2D::fetch()
-- GCTAAeff::ebounds()
-- GCTAAeff2D::ebounds()
-- GCTAAeffArf::ebounds()
-- GCTAAeffPerfTable::ebounds()
-- GCTAEventList::append_column(GFitsTableCol&)
-- GCTAEventList::has_phase(bool&)
-- GCTAEventList::has_phase()
-- GCTAEventList::has_detxy(bool&)
-- GCTAEventList::has_detxy()
-- GCTAEventList::has_mc_id(bool&)
-- GCTAEventList::has_mc_id()
-- GCTAEventList::set_mc_id_names(std::vector<int>&, std::vector<std::string>&)
-- GCTAEventAtom::mc_id(int&)
-- GCTAEventAtom::mc_id()
+- n.a.
  
 The following methods have been removed:
-- GFitsTable::size()
+- n.a.
  
 The following methods have been renamed:
-- GFitsImage::size() => GFitsImage::npix()
-- GFitsHDU::size() => GFitsHDU::cards()
-- GFitsTableCol::length() => GFitsTableCol::nrows()
+- n.a.
  
 The arguments for the following methods have been changed:
 - n.a.
@@ -81,19 +38,7 @@ The return value of the following methods has been changed:
  
 3. Configuration
 -----------------
-The Python unit tests now raise exceptions instead of exiting with a non-
-zero return code so that the user is not thrown out of Python after calling
-gammalib.test() (#1983).
-
-If the code originates from the Git repository the configuration now switches
-to developer mode which implies that the Python bindings are always built.
-This solves the problem that occured for developers that could not get rid
-of existing Python wrappers, even if they modified the Python interface
-(#1985).
-
-The shared libraries needed by the Python wrappers are now copied into a
-build/gammalib directory for unit testing instead of creating a symbolic
-link since on shared VM file systems symbolic links may not work (#1987).
+None
  
  
 4. Application module
@@ -108,57 +53,17 @@ None
  
 6. FITS I/O Interface
 ---------------------
-FITS image type conversion constructors have been added to convert a FITS
-image of any type into a specific storage class (#1986).
-
-A map between header card keywords and header card pointers was added to
-GFitsHeader to speed up the access of FITS header cards (#2120).
-
-The GFitsImage::size() method has been renamed to GFitsImage::npix() since
-the size() method overloaded the GFitsHDU::size() method (#1858).
-
-The GFitsTable::size() method has been removed since the meaning of the
-method was ambiguous. It returned the number of table columns, but the 
-GFitsTable::ncols() does exactly the same thing, hence the method was
-redundant (#1857).
-
-The GFitsHDU::size() method has been renamed to GFitsHDU::cards() since the
-meaning of the size() method was not clear (#1859).
-
-The GFitsTableCol::length() method was renamed to GFitsTableCol::nrows()
-to clarify the meaning of thie method (#1860).
+None
  
  
 7. Linear algebra module
 ------------------------
-Add operators for addition and subtraction of scalars to all matrix classes
-(#1990).
+None
  
  
 8. Model module
 ---------------
-The GModelSpectralNodes class now throws an exception if the energy or
-intensity of a node is set to a non-positive value (#990).
-
-The GModelSpectralSmoothBrokenPlaw class that implements a smoothly broken
-power law spectrum has been added (#1948).
-
-The scale of the composite spatial models can now be fitted by adding the
-free_scale="1" attribute to the XML file (#1992).
-
-Add GModelSpatialRadialProfile class to handle arbitrary radial profiles.
-The method deals with pre-computation and normalization of the profiles,
-and implements the eval() and mc() methods. The following derived classes
-have been implemented:
-- GModelSpatialRadialProfileGauss: radial Gaussian profile
-- GModelSpatialRadialProfileDMBurkert: Burkert Dark Matter profile
-- GModelSpatialRadialProfileDMEinasto: Einasto Dark Matter profile
-- GModelSpatialRadialProfileDMZhao: Zhao Dark matter profile
-Please note that these models are not deeply tested so far and the code
-and interface is still subject to change (#1520).
-
-The names of spatial model parameters are now always set to RA/DEC even if
-the input parameters were GLON/GLAT (#2113).
+None
  
  
 9. Numerics module
@@ -168,39 +73,22 @@ None
  
 10. Observation module
 ----------------------
-A set() method has been added to GEbounds and GEnergies to allow converting
-energy boundaries into an energy container and vice versa. In addition,
-conversion constructors have been added (#1977).
-
-The GEbounds::set_lin() and GEbounds::set_log() method now throw exceptions
-if the minimum energy is larger than the maximum energy. In addition, the
-GEbounds::set_log() method throws an exception if either the minimum or
-maximum energy are zero. Furthermore, the methods clear the energy boundaries
-if the requested number of intervals is non-zero (#1989). 
-
-The GPhases class has been added to handle phase selections of events (#1996).
+None
  
  
 11. Optimization module
 -----------------------
-The Levenberg-Marquardt optimiser convergence criterion is now scaled with the
-step size so that a small step size does no longer lead to a stop of the
-iterative process (#2022).
+None
  
  
 12. Sky module
 --------------
-The GWcsGLS and GWcsSFL classes have been added to support the GLS and SFL
-World Coordinate System projections (#1851).
-
-The GWcs base class now also supports CD keywords instead of CDELT keywords
-(#1951).
+None
  
  
 13. Support module
 ------------------
-A GFilename::type() method has been added that returns the type of the file
-based on the file type extension (e.g. ".fits") (#1863).
+None
  
  
 14. XML module
@@ -235,62 +123,10 @@ None
  
 20. CTA interface
 -----------------
-Correct the containment radius computation in GCTAPsf2D (#1876).
-
-The GCTAResponseIrf::rspname() method has been added that returns the name
-of the response function (attribute m_rspname).
-
-Do no longer interpret a CTA calibration database as a file path in the
-GCTAResponseIrf::read() method (#1969).
-
-Add GCTARoi::contains() method that checks whether an instrument direction
-is contained within the region of interest.
-
-Lazy loading of the energy dispersion in GCTAEdisp2D has been implemented by
-adding a fetch() method to the class. Now the energy dispersion is only loaded
-when needed, for example for function evaluation or Monte Carlo sampling (#1940).
-
-The GCTAEdisp2D::compute_ebounds_src() method now loops over the migration to
-enhance the precision in the true energy boundaries that are returned by this
-method. In addition, the validity range of the table is now stored in members
-of the class, and a non-zero energy dispersion is only returned for valid
-elements of the response table.
-
-An GCTAAeff::ebounds() method has been added to the abstract interface for
-effective area responses to return the valid energy range of the effective
-area (#1989).
-
-The GCTAModelAeffBackground::mc() could get stuck in the rejection loop if the
-effective area returned is always zero. A counter has been added that leads to
-the exit of the loop in case that 1000 subsequent zero values are returned
-(#1988).
-
-A GCTAEventList::append_column() method has been added to allow appending of
-FITS columns to an event list. The method checks whether the length of the
-FITS column is compatible with the number of events that are present in the
-event list (#1991).
-
-Add GCTAEventList::has_phase() and GCTAEventList::has_detxy() getter and setter
-methods to manipulate the event list information flags (#1982).
-
-Add GCTAEventList::has_mc_id() and GCTAEventAtom::mc_id() getter and setter methods
-to set the Monte Carlo identifier of a given event. This allows to attribute a
-Monte Carlo simulated event to a specific model (#1483).
-
-Add GCTAEventList::set_mc_id_names() method to store Monte Carlo metadata into an
-event list and to write this information into the FITS header. Furthermore, the
-information is read back when loading an event list (#1483).
-
-The loading and saving of CTA response cubes was put into an OpenMP critial zone to
-avoid FITS file pointer exceptions (#2021).
-
-The CTA background template HDUCLAS2 was changed from "BGD" to "BKG". Both strings
-are supported by GammaLib.
-
-The maximum radius of the King PSF has been increased from 0.7 to 2 degrees to cope
-with some extreme cases. Now also the maximum PSF radius is computed for a given
-energy so that if the PSF is smaller than the maximum radius of 2 degrees a smaller
-region of interest will be used for the computation (#2112).
+The GCTAPsfKing::r_max() method returned an infinite radius for a too small
+gamma parameter. If the gamma parameter is too small the method now returns
+a maximum radius of zero, and the GCTAPsfKing::update() method will set the
+normalization to zero (#2133).
  
  
 21. COMPTEL interface
 GammaLib information
 ====================
-* Version: 1.3.0 (12 August 2016)
+* Version: 1.3.1 (12 August 2016)
 * Author:  Juergen Knoedlseder (jurgen.knodlseder@irap.omp.eu)
  
 [![Build Status](https://cta-jenkins.irap.omp.eu/buildStatus/icon?job=gammalib-integrate-os)](https://cta-jenkins.irap.omp.eu/job/gammalib-integrate-os/)
@@ -184,7 +184,7 @@ If everything works successfully you should see
 or
  
     ============================================================================
-    Testsuite summary for gammalib 1.3.0
+    Testsuite summary for gammalib 1.3.1
     ============================================================================
     # TOTAL: 21
     # PASS:  21
@@ -29,7 +29,7 @@
 #############################################################################
 # Initialisation                                                            #
 #############################################################################
-AC_INIT([gammalib], [1.3.0], [jurgen.knodlseder@irap.omp.eu], [gammalib])
+AC_INIT([gammalib], [1.3.1], [jurgen.knodlseder@irap.omp.eu], [gammalib])
 AC_CONFIG_SRCDIR([src/support/GTools.cpp])
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_AUX_DIR([.])
@@ -32,7 +32,7 @@ PROJECT_NAME           = GammaLib
 # This could be handy for archiving the generated documentation or
 # if some version control system is used.
  
-PROJECT_NUMBER         = 1.3.0
+PROJECT_NUMBER         = 1.3.1
  
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer
@@ -58,7 +58,7 @@ copyright = u&#39;2006-2016, GammaLib developers&#39;
 version = '1.3'
  
 # The full version, including alpha/beta/rc tags.
-release = '1.3.0'
+release = '1.3.1'
  
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -6,7 +6,7 @@ includedir=${prefix}/include
 Name: GammaLib
 Description: Versatile toolbox for analysis of astronomical gamma-ray data
 URL: http://cta.irap.omp.eu/gammalib/
-Version: 1.3.0
+Version: 1.3.1
 Requires: 
 Requires.private: @CFITSIO@
 Conflicts: 
@@ -518,7 +518,21 @@ double GCTAPsfKing::delta_max(const double&amp; logE,
     if (m_par_norm > 0.0) {
         radius = r_max(logE, theta);
     }
-    
+if (gammalib::is_notanumber(radius) || gammalib::is_infinite(radius)) {
+    const double F = 0.999;
+    double u_max  = (std::pow((1.0 - F), (1.0/(1.0-m_par_gamma))) - 1.0) *
+                     m_par_gamma;
+    radius = m_par_sigma * std::sqrt(2.0 * u_max);
+std::cout << "GCTAPsfKing::delta_max";
+std::cout << " base=" << (1.0 - F);
+std::cout << " exponent=" << (1.0/(1.0-m_par_gamma));
+std::cout << " u_max=" << u_max;
+std::cout << " radius=" << radius;
+std::cout << " m_par_norm=" << m_par_norm;
+std::cout << " m_par_gamma=" << m_par_gamma;
+std::cout << " m_par_sigma=" << m_par_sigma << std::endl;
+}
+
     // Return maximum PSF radius
     return radius;
 }
@@ -760,6 +774,12 @@ void GCTAPsfKing::update(const double&amp; logE, const double&amp; theta) const
  
             } // endif: restricted PSF radius
  
+            // If maximum PSF radius is zero then also set the normalization
+            // to zero
+            else if (m_par_rmax == 0.0) {
+                m_par_norm = 0.0;
+            }
+
         } // endif: King profile parameters were valid
  
     } // endif: PSF parameters have changed
@@ -780,7 +800,8 @@ void GCTAPsfKing::update(const double&amp; logE, const double&amp; theta) const
  * is set by this method to where the containment fraction become 99.9%.
  *
  * This method requires the m_par_gamma and m_par_sigma to be set to valid
- * values.
+ * values. In case that the parameters lead to an infinite result the method
+ * will return a maximum size of zero.
  ***************************************************************************/
 double GCTAPsfKing::r_max(const double& logE,
                           const double& theta) const
@@ -791,7 +812,8 @@ double GCTAPsfKing::r_max(const double&amp; logE,
     // Compute maximum PSF radius for containment fraction
     double u_max  = (std::pow((1.0 - F), (1.0/(1.0-m_par_gamma))) - 1.0) *
                      m_par_gamma;
-    double radius = m_par_sigma * std::sqrt(2.0 * u_max);
+    double radius = (gammalib::is_infinite(u_max)) ? 0.0 :
+                     m_par_sigma * std::sqrt(2.0 * u_max);
  
     // Return maximum PSF radius
     return radius;
@@ -2075,10 +2075,11 @@ double GCTAResponseIrf::npsf(const GSkyDir&amp;      srcDir,
                 std::cout << "*** ERROR: GCTAResponseIrf::npsf:";
                 std::cout << " NaN/Inf encountered";
                 std::cout << " (value=" << value;
-                std::cout << ", roi_radius=" << roi_radius;
-                std::cout << ", roi_psf_distance=" << roi_psf_distance;
-                //std::cout << ", sigma=" << sigma;
-                std::cout << ", r=[" << rmin << "," << rmax << "])";
+                std::cout << ", roi_radius=" << roi_radius * gammalib::rad2deg;
+                std::cout << ", roi_psf_distance=";
+                std::cout << roi_psf_distance * gammalib::rad2deg;
+                std::cout << ", r=[" << rmin * gammalib::rad2deg;
+                std::cout << "," << rmax * gammalib::rad2deg << "])";
                 std::cout << std::endl;
             }
             #endif
 # Required metadata
 sonar.projectKey=gammalib
 sonar.projectName=GammaLib
-sonar.projectVersion=1.3.0
+sonar.projectVersion=1.3.1
  
 # Description
 sonar.projectDescription="Versatile toolbox for scientific analysis of astronomical gamma-ray data"