PyConfigure

C++ class that uses a python dictionary to hold an arbitrary amount of configuration data.

TPython already does almost all that I want to do :(
The rest is not worth the hassle.

My thoughts:

• make a class that hold configuration data.
• one can pass a string with python code which is then parsed into an internal dictionary.
• a python dictionary can be passed directly if it is available either from python or from another C api program. The dictionary replaced the internal one.
• The calib object has move semantics like python objects. It is only needed to read from. Deep copying is not possible.
• The object can be queried with a get(string) function. The result object has cast operators to a number of standard calibration type objects like:
  • ints
  • double
  • bools
  • strings
  • vector
  • vector
  • vector
  • vector
  • another calib dict object
• The last kind would be the result of the key being found to be a dict.
• The result object would not be possible to be copied, stored, or assigned. This way one is forced to cast it immediately.
• in all the class would be well suited to configure code that may be called both from C++ code or from a python api, without having to pass down a lot of individual config options.

Usage will be through a single PyConfigure.h. Then:

PyConfigure::Dict mydict;
mydict =    "#Python yay"
            "a=5\n"
            "s='hello'\n"
            "v=[1,2,3]\n"
            "delegated=dict("
                "a=1,"
                "b='meh'"
                ")\n";
int a = mydict["a"];
string s = mydict["s"];
vector<int> = mydict["v"];
PyConfigure::Dict otherdict = mydict["delegated"];

all python exceptions become C++ exceptions.

Copyright Simon Heisterkamp 2012