The design of the SANS models is not easily scalable. Although a more efficient
way (in terms of design and speed) as been implemented on the C/C++ side, the python
design should be revisited. Areas of improvements include:

1- [C side] Incorporate the C functions that use libigor directly into the new C++ class.
  	That would mean rewritting the script generator to eliminate all the .c files.
  
2- Re-design the sans.models.BaseComponent class to store the parameters as objects,
	like it's done on the C++ side. A better design may allow us to avoid the awful
	hard-coding of the translation between the python dictionary holding the 
	parameters and the C++ Parameter class.
	 
 Length = 400
I(Q=0.001) = 450.355
I(Q=0.001) = 472.332
I(Q=0.001) = 4723.32
I(Qx=0.001, Qy=0.001) = 4743.56
I(Q=0.00447,  Phi=0.7854) = 4733
pi=3.14159

I(Qx=0.01, Qy=0.01) = 2299
I(Q=0.0141421)         = 2326.75
ratio 0.988074 1.01207

Cyl (python):  450.35506578
Disp(pyhton):  472.332213339

--------------
Hello cylinder
450.35506578
449.358988524
446.344929175

{'npts': 1, 'type': 'gaussian', 'nsigmas': 2, 'width': 0.0}
{'npts': 1, 'type': 'flat', 'width': 0.0}

 q= 0.001
New Cyl :  450.35506578
New Disp:  472.332213339

cyl(0.01, 0.01) =96.8691
cyl(0.0141421) =232.675
{'type': 'array'}
cyl(0.01, 0.01) =229.9
Hello cylinder

Old Cyl :  450.35506578
Old Disp:  472.332213339