{"id":37,"date":"2022-02-18T15:22:51","date_gmt":"2022-02-18T15:22:51","guid":{"rendered":"http:\/\/www.pulsars.info\/wordpress\/?p=37"},"modified":"2022-02-18T15:41:31","modified_gmt":"2022-02-18T15:41:31","slug":"combining-c-and-python-dynamic-library","status":"publish","type":"post","link":"http:\/\/www.pulsars.info\/wordpress\/uncategorized\/combining-c-and-python-dynamic-library\/","title":{"rendered":"Combining C++ and Python: dynamic library"},"content":{"rendered":"\n
The simplest way to combine the unprecedented graphical potential of Python (matplotlib library) and a speed of numerical calculations of pure C\/C++ languages is to compile the C-code as a dynamic library.<\/p>\n\n\n\n
Here I give an example.<\/p>\n\n\n\n
#include <cmath> \/\/ for modf and cos \nusing<\/strong> namespace<\/strong> std;\nextern<\/strong> \"C\" double<\/strong> value (double<\/strong>); \ndouble<\/strong> value (double<\/strong> step) {\ndouble<\/strong> res, pos, fractpart, intpart;\nint<\/strong> n;\nfractpart = modf<\/strong>(M_PI \/ 2.0 \/ step, &intpart); \/\/ Compute a number of steps for numerical integration\nn = intpart;\nres = 0.0;\nfor<\/strong> ( int<\/strong> i=0; i < n; i++) {\n pos = i * step; \/\/ Eulear integration method\n res += step * cos<\/strong>(pos);\n}\nres += fractpart*step * cos<\/strong>(M_PI \/ 2.0);\nreturn<\/strong> res; \n}\n<\/pre>\n\n\n\n
extern “C” is necessary to clarify which functions are visible outside the dynamical library. The python package ‘ctype’ was written for pure C-language. Therefore, the class structure is not inherited. The class structure should be duplicated in the Python code.<\/p>\n\n\n\n
To compile the dynamical library we use following command (iOS):<\/p>\n\n\n\n
The Python script has following structure:<\/p>\n\n\n\n
import<\/strong> numpy as<\/strong> np ## library to work with arrays\nimport<\/strong> matplotlib.pyplot as<\/strong> plt ## plotting library\nimport<\/strong> ctypes\nfrom<\/strong> ctypes import<\/strong> cdll, c_double\nlib = cdll.LoadLibrary('.\/test_func.so') ## use our library written in C\ngrid = np.linspace (0.01, 1.0, 50) ## prepare a grid with stepsizes\nres_list=[]\nlib.value.restype = ctypes.c_double ## Declare a type of return argument (important)\nfor<\/strong> i in<\/strong> range<\/strong> (0, len<\/strong>(grid)): ## for each step size we call our function\n step = c_double(grid[i])\n res = lib.value(step)\n res_list.append(res)\nplt.plot(grid, res_list) ## plot the resylt of numerical integration\nplt.xlabel('Numerical integration step size')\nplt.ylabel('Result of integration')\nplt.savefig('plot.pdf')\nplt.show()\n<\/pre>\n\n\n\n
The only complicated part of the code is how to return values.<\/p>\n\n\n\n
If the code works correctly after we call Python scripts, it prepares following figure:<\/p>\n\n\n\n<\/figure>\n\n\n\n
When the integration step is tiny (~0.1) we get the right result. If the integration step is large, the result fluctuates a lot.<\/p>\n","protected":false},"excerpt":{"rendered":"
The simplest way to combine the unprecedented graphical potential of Python (matplotlib library) and a speed of numerical calculations of pure C\/C++ languages is to compile the…<\/p>\n
<\/figure>\n\n\n\n