ctools

There is the possibility to pass INDEF, NONE, UNDEF or UNDEFINED as a real parameter and gammalib will set the status of these parameters to ST_UNDEFINED (can be checked using GApplicationPar::is_undefined()). We should use this functionality instead of setting RA and DEC to -1, which formally is not really an invalid coordinate.

I agree, it would be better to pass UNDEFINED to the parameter values. How would that look work with the python interface, where ST_UNDEFINED is not existent?
What does the the following line pass to the double in case the parameter is undefined?

double xref = (*this)["ra"].real()

You don’t have to worry about ST_UNDEFINED, just use the GApplicationPar::is_undefined() method, e.g.

if sim["ra"].is_undefined():
    print("This parameter is undefined.")

$ ctobssim ra=UNDEFINED dec=UNDEFINED
*** ERROR encounterted in the execution of ctobssim. Run aborted ...
*** ERROR in GApplicationPar::real(): Invalid value.
Parameter "ra" is undefined. Please specify a valid parameter value.

For parameters that may be undefined you should use GApplicationPar::is_undefined() (or better GApplicationPar::is_valid() ) before reading a parameter:

if sim["ra"].is_valid():
    ra = sim["ra"].real()
else:
    ra = -1 # Although I won't use -1 to signal a bad value

I see. My basic idea was to omit the RoI and time selection in csspec (#1364)

So the required changes to ctselect would be to add three protected booleans: m_select_energy, m_select_roi, m_select_time.
Then, in ctselect::get_parameters(), we incorporate the following checks:

if ((*this)["ra"].is_valid()) {
    m_ra = (*this)["ra"].real();
}
else {
    m_select_roi = false;
}

The same has then to be done for “dec”, “rad”, “emin”, “emax”, “tmin”, and “tmax”. The new protected booleans can then be used to define the selections in ctselect::select_events().
The drawback would be that this option could not be used from the command line, because the parameter will be checked directly after query. Accordingly, an exception would be thrown if the user wants to pass an undefined value.

An alternative would be to just add three hidden boolean parameters to ctselect that steer if a selection of RoI, Energy or time is performed respectively.
What do you think?

Mayer Michael wrote:

I see. My basic idea was to omit the RoI and time selection in csspec (#1364)

So the required changes to ctselect would be to add three protected booleans: m_select_energy, m_select_roi, m_select_time.
Then, in ctselect::get_parameters(), we incorporate the following checks:
[...]

The same has then to be done for “dec”, “rad”, “emin”, “emax”, “tmin”, and “tmax”. The new protected booleans can then be used to define the selections in ctselect::select_events().
The drawback would be that this option could not be used from the command line, because the parameter will be checked directly after query. Accordingly, an exception would be thrown if the user wants to pass an undefined value.

No, the exception is only thrown if you try to access the value of a parameter with the real() method. Hence if a value is undefined, but you check whether the parameter is_valid() before reading, there should be no exception thrown as you never would call the real() method.

An alternative would be to just add three hidden boolean parameters to ctselect that steer if a selection of RoI, Energy or time is performed respectively.
What do you think?

This is not needed (see above)