قصة تجربة واحدة مع متجه Cython و C ++

cython_experiments.h

 #include "Python.h" static PyObject * convert_to_pyobject(PyObject *obj) { return obj; } 

cython_experiments.pyx

 # -*- coding: utf-8 -*- # distutils: language = c++ # distutils: include=['./'] # distutils: extra_compile_args=["-O1"] from __future__ import unicode_literals import time from libc.stdlib cimport free from cpython.dict cimport PyDict_New, PyDict_SetItemString from cpython.ref cimport PyObject from libcpp.string cimport string from libcpp.vector cimport vector cdef extern from "cython_experiments.h": PyObject* convert_to_pyobject(object obj) ctypedef vector[PyObject*] dict_vec range_attempts = 10 ** 6 # Insert time cdef test_list(): t_start = time.time() data_list = list() for i from 0 <= i < range_attempts: data_list.append(dict( name = 'test_{}'.format(i), test_data = i, test_data2 = str(i), test_data3 = range(10), )) del data_list return time.time() - t_start cdef test_vector(): t_start = time.time() cdef dict_vec *data_list data_list = new dict_vec() data_list.resize(range_attempts) for i from 0 <= i < range_attempts: data = PyDict_New() PyDict_SetItemString(data, 'name', 'test_{}'.format(i)) PyDict_SetItemString(data, 'test_data', i) PyDict_SetItemString(data, 'test_data2', str(i)) PyDict_SetItemString(data, 'test_data3', range(10)) data_list.push_back(convert_to_pyobject(data)) free(data_list) return time.time() - t_start # Get statistic times = dict(list=[], vector=[]) attempts = 100 for i from 0 <= i < attempts: times['list'].append(test_list()) times['vector'].append(test_vector()) print(''' Attempt: {} List time: {} Vector time: {} '''.format(i, times['list'][-1], times['vector'][-1])) avg_list = sum(times['list']) / attempts avg_vector = sum(times['vector']) / attempts print(''' Statistics: attempts: {} list avg time: {} vector avg time: {} '''.format(attempts, avg_list, avg_vector)) 

cython_experiments.h

 #include "Python.h" #include <vector> #include <algorithm> #ifndef PyString_AsString #define PyString_AsString PyUnicode_AsUTF8 #define PyString_FromString PyUnicode_FromString #endif typedef struct { char* name; bool reverse; } sortFiled; class cmpclass { public: cmpclass(std::vector<char*> fields) { for (std::vector<char*>::iterator it = fields.begin() ; it < fields.end(); it++){ bool is_reverse = false; char* name; if (it[0] == "-"){ is_reverse = true; for(int i=1; i<strlen(*it); ++i) name[i] = *it[i]; } else { name = *it; } sortFiled field = {name, is_reverse}; this->fields_to_cmp.push_back(field); } } ~cmpclass() { this->fields_to_cmp.clear(); this->fields_to_cmp.shrink_to_fit(); } bool operator() (PyObject* left, PyObject* right) { // bool result = false; for (std::vector<sortFiled>::iterator it = this->fields_to_cmp.begin() ; it < this->fields_to_cmp.end(); it++){ // PyObject* str_name = PyString_FromString(it->name); PyObject* right_value = PyDict_GetItem(right, str_name); PyObject* left_value = PyDict_GetItem(left, str_name); if(!it->reverse){ result = left_value < right_value; } else { result = (left_value > right_value); } PyObject_Free(str_name); if(!result) return false; } return true; } private: std::vector<sortFiled> fields_to_cmp; }; void vector_multikeysort(std::vector<PyObject *> items, PyObject* columns, bool reverse) { std::vector<char *> _columns; for (int i=0; i<PyList_GET_SIZE(columns); ++i) { PyObject* item = PyList_GetItem(columns, i); char* item_str = PyString_AsString(item); _columns.push_back(item_str); } cmpclass cmp_obj(_columns); std::sort(items.begin(), items.end(), cmp_obj); if(reverse) std::reverse(items.begin(), items.end()); } std::vector<PyObject *> _test_vector(PyObject* store_data_list, PyObject* columns, bool reverse = false) { int range_attempts = PyList_GET_SIZE(store_data_list); std::vector<PyObject *> data_list; for (int i=0; i<range_attempts; ++i) { data_list.push_back(PyList_GetItem(store_data_list, i)); } vector_multikeysort(data_list, columns, reverse); return data_list; } 

cython_experiments.pyx

 # -*- coding: utf-8 -*- # distutils: language = c++ # distutils: include=['./'] # distutils: extra_compile_args=["-O2", "-ftree-vectorize"] from __future__ import unicode_literals import time from libc.stdlib cimport free from cpython.dict cimport PyDict_New, PyDict_SetItemString from cpython.ref cimport PyObject from libcpp.string cimport string from libcpp.vector cimport vector import gc cdef extern from "cython_experiments.h": vector[PyObject*] _test_vector(object store_data_list, object columns, int reverse) range_attempts = 10 ** 6 store_data_list = list() for i from 0 <= i < range_attempts: store_data_list.append(dict( name = 'test_{}'.format(i), test_data = i, test_data2 = str(i), test_data3 = range(10), )) # Insert time def multikeysort(items, columns, reverse=False): items = list(items) columns = list(columns) columns.reverse() for column in columns: # pylint: disable=cell-var-from-loop is_reverse = column.startswith('-') if is_reverse: column = column[1:] items.sort(key=lambda row: row[column], reverse=is_reverse) if reverse: items.reverse() return items cdef test_list(): t_start = time.time() data_list = list() for i in store_data_list: data_list.append(i) data_list = multikeysort(data_list, ('name', '-test_data'), True) for i in data_list: i del data_list return time.time() - t_start cdef test_vector(): t_start = time.time() data_list = _test_vector(store_data_list, ['name', '-test_data'], 1) for i in data_list: i return time.time() - t_start # Get statistic times = dict(list=[], vector=[]) attempts = 10 gc.disable() for i from 0 <= i < attempts: times['list'].append(test_list()) times['vector'].append(test_vector()) gc.collect() print(''' Attempt: {} List time: {} Vector time: {} '''.format(i, times['list'][-1], times['vector'][-1])) del store_data_list avg_list = sum(times['list']) / attempts avg_vector = sum(times['vector']) / attempts print(''' Statistics: attempts: {} list avg time: {} vector avg time: {} '''.format(attempts, avg_list, avg_vector))