《编写高质量代码 改善Python程序的91个建议》
《编写高质量代码 改善Python程序的91个建议》读后程序学习小结 - BigDeng_2014的专栏 - CSDN博客
# coding=utf-8 # Language Reference ''' 参考书:《编写高质量代码 改善Python程序的91个建议》张颖,赖勇浩 著 2014.6 ''' from __future__ import with_statement # assert x, y = 1, 1 assert x == y, "not equals" # time计时的两种方式 import timeit t = timeit.Timer('x,y=y,x','x=1;y=2') print(t.timeit()) # 0.110494892863 import time t = time.time() sum = 0 while True: sum += 1 if sum > 100000: break time.sleep(1) print(time.time()-t) # 1.02999997139 # itertools 结合 yield from itertools import islice def fib(): a, b = 0, 1 while True: yield a a, b = b, a+b print(list(islice(fib(),10))) # [0, 1, 1, 2, 3, 5, 8, 13, 21, 34] # class 定义变量 和 nametuple 定义变量 class Seasons: Sprint, Summer, Autumn, Winter = range(4) print(Seasons.Winter) # 3 from collections import namedtuple Seasons0 = namedtuple('Seasons0','Spring Summer Autumn Winter')._make(range(4)) print(Seasons0.Winter) # 3 # isintance 可以设置多种类型的判断 print(isinstance((2,3),(str, list, tuple))) # True # eval的漏洞,能够处理的范围太大,导致系统文件被读取,谨慎使用 str0 = '__import__("os").system("dir")' eval(str0) # 2017/09/28 09:31 <DIR> 。。。 # 生成器 yield 与 迭代器 iteritems def myenumerate(seq): n = -1 for elem in reversed(seq): yield len(seq) + n, elem n = n -1 e = myenumerate([1,2,3,4,5]) print(e.next()) # (4, 5) dict0 = {'1':1, '2':2} d = dict0.iteritems() print(d.next()) # ('1', 1) # 字符串驻留机制:对于较小的字符串,为了提高系统性能保留其值的一个副本,当创建新的字符串时直接指向该副本。 a = 'hello' b = 'hello' # b 是 a 的引用 print(id(a), id(b), a is b, a == b) # (53383488, 53383488, True, True) # 可变对象list 与 不可变对象str list1 = [1,2,3] list2 = list1 list3 = list1[:] # 浅拷贝 list1.append(4) print(list1, list2, list3) # ([1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3]) - list2可变 str1 = '123' str2 = str1 str3 = str1[:] str1 += '4' print(str1, str2, str3) # ('1234', '123', '123') - str2不可变 # 浅拷贝, 深拷贝 # 浅拷贝和深拷贝的不同仅仅是对组合对象来说, # 所谓的组合对象就是包含了其它对象的对象,如列表,类实例。 # 而对于数字、字符串以及其它“原子”类型,没有拷贝一说,产生的都是原对象的引用。 # 浅拷贝: 创建一个新的对象,其内容是原对象中元素的引用。(拷贝组合对象,不拷贝子对象) # 浅拷贝有:切片操作、工厂函数、对象的copy()方法、copy模块中的copy函数。 # 深拷贝: 创建一个新的对象,然后递归的拷贝原对象所包含的子对象。深拷贝出来的对象与原对象没有任何关联。 # 深拷贝: 虽然实际上会共享不可变的子对象,但不影响它们的相互独立性。 # 深拷贝只有一种方式:copy模块中的deepcopy函数。 a = [[1,2]] # 组合对象 import copy b = copy.copy(a) print(id(a), id(b), a is b, a == b) # (58740048, 59050864, False, True) for i, j in zip(a, b): print(id(i), id(j)) # (61090752, 61090752)子对象相同 b = copy.deepcopy(a) print(id(a), id(b), a is b, a == b) # (58740048, 58737848, False, True) for i, j in zip(a, b): print(id(i), id(j)) # (61090752, 61071568)子对象不同 # 再举一例 class TestCopy(): def get_list(self, list0): self.list0 = list0 def change_list(self, str): self.list0 += str def print_list(self): print(self.list0) list0 = [1, 2, 3] a = TestCopy() a.get_list(list0) a.print_list() # [1, 2, 3] b = copy.copy(a) # 浅拷贝,子对象共享 b.change_list('4') b.print_list() # [1, 2, 3, '4'] a.print_list() # [1, 2, 3, '4'] c = copy.deepcopy(a) # 深拷贝,子对象独立 c.change_list('5') c.print_list() # [1, 2, 3, '4', '5'] b.print_list() # [1, 2, 3, '4'] a.print_list() # [1, 2, 3, '4'] # 赋值操作 a = [1, 2, 3] b = copy.copy(a) # 其实是赋值操作,不属于浅拷贝,赋值对象相互独立 b.append(4) print(a, b) # ([1, 2, 3], [1, 2, 3, 4]) # encode, decode, gbk, utf-8 with open("test.txt", 'w') as f: f.write('python' + '中文测试') with open("test.txt", 'r') as f: # print(f.read()) # python中文测试 print((f.read().decode('utf-8')).encode('gbk')) # python���IJ��� # --1 = 1 = ++1 = (1) print(+++1, ---1) # (1, -1) print (1), (1,) # 1 (1,) print(''.split(), ''.split(' ')) # ([], ['']) # with 调用 class, 初始化调用__enter__,退出调用__exit__ class MyContextManager(): def __enter__(self): print('entering...') def __exit__(self, exc_type, exc_val, exc_tb): print('leaving...') if exc_type is None: print('no exceptions') return False elif exc_type is ValueError: print('value error') return True else: print('other error') return True with MyContextManager(): print('Testing...') raise(ValueError) # entering... Testing... leaving... value error # else 结合 for 和 try for i in range(5): print(i), else: print('for_else') # 0 1 2 3 4 for_else try: print('try'), except: pass else: print('try_else') # try try_else # None a = None b = None print(id(a), id(b), a is b, a==b) # (505354444, 505354444, True, True) a = 'a' b = u'b' print(isinstance(a,str), isinstance(b,unicode), isinstance(a, basestring), isinstance(b, basestring)) # (True, True, True, True) # 对 class 操作,先调用 __init__() class A: def __nonzero__(self): print('A.__nonzero__()') return True def __len__(self): print('A.__len__()') return False def __init__(self): print('A.__init__()') self.name = 'I am A' def __str__(self): print('A.__str__()') return 'A.__str__{self.name}'.format(self=self) if A(): print('not empty') # A.__init__() A.__nonzero__() not empty else: print('empty') print(str(A())) # A.__init__() A.__str__() A.__str__I am A # 尽量采用''.join()(效率更高),而不是 str + str s1, s2 ,s3 = 'a', 'b', 'c' print(s1+s2+s3, ''.join([s1, s2, s3])) # ('abc', 'abc') # map 与 list 结合 list0 = [('a','b'), ('c','d')] formatter = "choose {0[0]} and {0[1]}".format for item in map(formatter, list0): print(item) # choose a and b choose c and d # map 结合 type product_info = '1-2-3' a, b, c = map(int, product_info.split('-')) print(a, b, c) # (1, 2, 3) # 格式化输出,尽量采用 format,采用 %s 输出元组时需要加逗号 itemname = list0[0] + list0[1] print(itemname) # ('a', 'b', 'c', 'd') print('itemname is %s' % (itemname,)) # 必须有个逗号 itemname is ('a', 'b', 'c', 'd') print('itemname is {}'.format(itemname)) # itemname is ('a', 'b', 'c', 'd') # 格式化输出 %2.5f,小数点后5位优先级高 print('data: %6.3f' % 123.456789123) # data: 123.457 print('data: %2.5f' % 123.456789123) # data: 123.45679 # class传参:__init__ 中传入可变对象 - 会在子类中继承该可变对象的值 class ChangeA(): def __init__(self, list0 = []): # mutable 可变 self.list0 = list0 def addChange(self, content): self.list0.append(content) a = ChangeA() a.addChange('add change') b = ChangeA() print(a.list0, b.list0) # (['add change'], ['add change']) # 函数传参:传对象或对象的引用。若可变对象 - 共享, 若不可变对象 - 生成新对象后赋值 def inc(n, list0): n = n + 1 list0.append('a') n = 3 list0 = [1] inc(n, list0) print(n, list0) # (3, [1, 'a']) # 子类继承父类,传参举例 class Father(): def print_fa(self): print(self.total) def set(self, total): self.total = total class SonA(Father): pass class SonB(Father): pass a = SonA() a.set([1]) a.print_fa() # [1] b = SonB() b.set([2]) b.print_fa() # [2] # 在需要生成列表的时候使用列表解析 # 对于大数据处理不建议用列表解析,过多的内存消耗会导致MemoryError print([(a,b) for a in [1,2,3] for b in [2,3,4] if a != b]) # [(1, 2), (1, 3), (1, 4), (2, 3), (2, 4), (3, 2), (3, 4)] # class 中的 装饰符 :类方法,静态方法,实例方法 class CS(): def instance_method(self,x): print(x) @classmethod def class_method(cls,x): print('class',x) @staticmethod def static_method(x): print('static',x) CS().instance_method(1) # 1 CS().class_method(2) # ('class', 2) CS().static_method(3) # ('static', 3) # itemgetter 字典排序,输出为元组 dict0 = {'a':1, 'c':3, 'b':2} from operator import itemgetter print(sorted(dict0.viewitems(), key = itemgetter(1))) # [('a', 1), ('b', 2), ('c', 3)] # Counter 计数 from collections import Counter print(Counter('success')) # Counter({'s': 3, 'c': 2, 'e': 1, 'u': 1}) # 配置文件:优点:不需修改代码,改变程序行为,继承[DEFAULT]属性 with open('format.conf', 'w') as f: f.write('[DEFAULT]' + ' ') f.write('conn_str = %(dbn)s://%(user)s:%(pw)s@%(host)s:%(port)s/%(db)s' + ' ') f.write('dbn = mysql' + ' ') f.write('user = root' + ' ') f.write('host = localhost' + ' ') f.write('port = 3306' + ' ') f.write('[db1]' + ' ') f.write('user = aaa' + ' ') f.write('pw = ppp' + ' ') f.write('db = example1' + ' ') f.write('[db2]' + ' ') f.write('host = 192.168.0.110' + ' ') f.write('pw = www' + ' ') f.write('db = example2' + ' ') from ConfigParser import ConfigParser conf = ConfigParser() conf.read('format.conf') print(conf.get('db1', 'conn_str')) # mysql://aaa:ppp@localhost:3306/example1 print(conf.get('db2', 'conn_str')) # mysql://root:www@192.168.0.110:3306/example2 # pandas - 大文件(1G)读取操作 - 需要安装 pandas f = open('large.csv', 'wb') f.seek(1073741824 - 1) f.write(' ') f.close() import os print(os.stat('large.csv').st_size) # 1073741824 import csv with open('large.csv', 'rb') as csvfile: mycsv = csv.reader(csvfile, delimiter = ';') # for row in mycsv: # MemoryError # print(row) # import pandas as pd # reader = pd.read_table('large.csv', chunksize = 10, iterator = True) # iter(reader).next() # 序列化:把内存中的数据结构在不丢失其身份和类型信息的情况下,转成对象的文本或二进制表示。 # pickle, json, marshal, shelve import cPickle as pickle my_data = {'a':1, 'b':2, 'c':3} fp = open('picklefile.dat', 'wb') pickle.dump(my_data, fp) # class - __getstate__(self) fp.close() fp = open('picklefile.dat', 'rb') out = pickle.load(fp) # class - __setstate__(self, state) fp.close() print(out) # {'a': 1, 'c': 3, 'b': 2} pickle.loads("cos system (S'dir' tR.") # 列出当前目录下所有文件,不安全 - 解决:继承类并定制化内容 # 编码器 json.JSONEncoder try: import simplejson as json except ImportError: import json import datetime d = datetime.datetime.now() d1 = d.strftime('%Y-%m-%d %H:%M:%S') print(d1, json.dumps(d1, cls = json.JSONEncoder)) # 也可以继承修改指定编码器json.JSONEncoder # ('2017-09-28 11:00:46', '"2017-09-28 11:00:46"') # traceback:出错时查看 调用栈 import sys print(sys.getrecursionlimit()) # 最大递归深度:1000 import traceback try: a = [1] print(a[1]) except IndexError as ex: print(ex) # list index out of range # traceback.print_exc() # 会导致程序中断 tb_type, tb_val, exc_tb = sys.exc_info() for filename, linenum, funcname, source in traceback.extract_tb(exc_tb): print("%-33s:%s '%s' in %s()" % (filename, linenum, source, funcname)) # H:/python/suggest0928.py :353 'print(a[1])' in <module>() # LOG的五个等级:DEBUG, INFO, WARNING(默认), ERROR, CRITICAL # Logger, Handler, Formatter, Filter import logging logging.basicConfig( level = logging.DEBUG, filename = 'log.txt', filemode = 'w', format = '%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s', ) logger = logging.getLogger() logger.info('[INFO]:I am a tester') logger.debug('test logging module') logger.error('this is error') logger.critical('this is critical') # thread: 多线程底层支持,以低级原始的方式处理和控制线程,较复杂 # threading: 基于thread,操作对象化,提供丰富特性 import threading import time def myfunc(a, delay): print('calculate %s after %s' % (a, delay)) time.sleep(delay) print('begin') res = a*a print('result:', res) return res t1 = threading.Thread(target = myfunc, args = (2, 5)) t2 = threading.Thread(target = myfunc, args = (6, 8)) print(t1.isDaemon()) # False 守护线程,默认False print(t2.isDaemon()) # t2.setDaemon(True) # True 表示 线程全部执行完成后,主程序才会退出 t1.start() t2.start() # lock, mutex, condition, event, with lock, put,get # 生产者消费者模型 import Queue import threading import random write_lock = threading.Lock() class Producer(threading.Thread): def __init__(self, q, con, name): super(Producer, self).__init__() self.q = q self.name = name self.con = con print('Producer ', self.name, ' started') def run(self): while(1): global write_lock # self.con.acquire() if self.q.full(): with write_lock: print('Queue is full, producer wait') # self.con.wait() else: value = random.randint(0,10) with write_lock: print(self.name, 'put value:', self.name+':'+str(value), 'into queue') self.q.put(self.name+':'+str(value)) # self.con.notify() # self.con.release() class Consumer(threading.Thread): def __init__(self, q, con, name): super(Consumer, self).__init__() self.q = q self.name = name self.con = con print('Consumer ', self.name, ' started') def run(self): while(1): global write_lock # self.con.acquire() if self.q.empty(): with write_lock: print('Queue is empty, consumer wait') # self.con.wait() else: value = self.q.get() with write_lock: print(self.name, 'get value:', value, 'from queue') # self.con.notify() # self.con.release() q = Queue.Queue(10) # 先进先出,循环队列大小10 con = threading.Condition() p1 = Producer(q, con, 'P1') p2 = Producer(q, con, 'P2') c1 = Consumer(q, con, 'C1') p1.setDaemon(False) p2.setDaemon(False) c1.setDaemon(False) # p1.setDaemon(True) # p2.setDaemon(True) # c1.setDaemon(True) # p1.start() # p2.start() # c1.start() ''' 设计模式,静态语言风格 单例模式,保证系统中一个类只有一个实例而且该实例易于被外界访 模板方法:在一个方法中定义一个算法的骨架,并将一些事先步骤延迟到子类中。 子类在不改变算法结构的情况下,重新定义算法中的某些步骤。 混入mixins模式:基类在运行中可以动态改变(动态性)。 ''' # 发布 publish 订阅 subscribe 松散耦合 - 中间代理人 Broker # blinker - python-message # 库函数:关注日志产生,不关注日志输出; # 应用:关注日志统一放置,不关注谁产生日志。 from collections import defaultdict route_table = defaultdict(list) def sub(topic, callback): if callback in route_table[topic]: return route_table[topic].append(callback) def pub(topic, *a, **kw): for func in route_table[topic]: func(*a, **kw) def greeting(name): print('hello, %s' % name) sub('greet', greeting) # 订阅的时候将待调用的greeting放入dict中 pub('greet', 'tester') # hello, tester 发布的时候调用greeting函数 # 类的状态转移,例,当telnet注册成功后,就不再需要登录注册了。 def workday(): print('work hard') def weekend(): print('play harder') class People(): pass people = People() while True: for i in range(1,8,1): if i == 6: people.day = weekend if i == 1: people.day = workday people.day() break # 工厂模式 # __init__(): 在类对象创建好后,进行变量的初始化 # __new__(): 创建实例,类的构造方法,需要返回object.__new__() class TestMode(object): def __init__(self): print('i am father') def test(self): print('test is father') class A(TestMode): def __init__(self): print('i am A') def test(self): print('test is A') class B(TestMode): def __init__(self): print('i am B') def test(self): print('test is B') class FactoryTest(object): content = {'a':A, 'b':B} def __new__(cls, name): if name in FactoryTest.content.keys(): print('create old %s' % name) return FactoryTest.content[name]() else: print('create new %s' % name) return TestMode() FactoryTest('a').test() # create old a - i am A - test is A FactoryTest('A').test() # create new A - i am father - test is father # 局部作用域 > 嵌套作用域 > 全局作用域 > 内置作用域 a = 1 def foo(x): global a a = a * x def bar(): global a b = a * 2 a = b + 1 print(a) return bar() foo(1) # 3 # self 隐式传递 -- 显式 优于 隐式 # 当子类覆盖了父类的方法,但仍然想调用父类的方法 class SelfTest(): def test(self): print('self test') SelfTest.test(SelfTest()) # self test assert id(SelfTest.__dict__['test']) == id(SelfTest.test.__func__) # 古典类 classic class class A: pass # 新式类 new style class class B(object): pass class D(dict): pass # 元类 metaclass class C(type): pass a = A b = B() c = C(str) d = D() print(type(a)) # <type 'classobj'> print(b.__class__, type(b)) # (<class '__main__.B'>, <class '__main__.B'>) print(c.__class__, type(c)) # (<type 'type'>, <type 'type'>) print(d.__class__, type(d)) # (<class '__main__.D'>, <class '__main__.D'>) # 菱形继承 - 应避免出现 try: class A(object): pass class B(object): pass class C(A, B): pass class D(B, A): pass class E(C, D): pass except: print('菱形继承 - '+'order (MRO) for bases B, A') # __dict__[] 描述符,实例调用方法为bound,类调用方法为unbound class MyClass(object): def my_method(self): print('my method') print(MyClass.__dict__['my_method'], MyClass.my_method) # (<function my_method at 0x03B62630>, <unbound method MyClass.my_method>) print(MyClass.__dict__['my_method'](MyClass()), MyClass.my_method(MyClass())) a = MyClass() print(a.my_method, MyClass.my_method) # (<bound method MyClass.my_method of <__main__.MyClass object at 0x038D3650>>, <unbound method MyClass.my_method>) print(a.my_method.im_self, MyClass.my_method.im_self) # (<__main__.MyClass object at 0x0391D650>, None) # __getattribute__()总会被调用,而__getattr__()只有在__getattribute__()中引发异常的情况下才会被调用 class AA(object): def __init__(self, name): self.name = name self.x = 20 def __getattr__(self, name): print('call __getattr__:', name) if name == 'z': return self.x ** 2 elif name == 'y': return self.x ** 3 def __getattribute__(self, attr): print('call __getattribute__:', attr) try: return super(AA, self).__getattribute__(attr) except KeyError: return 'default' a = AA("attribute") print(a.name) # attribute print(a.z) # 400 if hasattr(a, 'test'): # 动态添加了 test 属性,但不会在 __dict__ 中显示 c = a.test print(c) # None else: print('instance a has no attribute t') print(a.__dict__) # {'x': 20, 'name': 'attribute'} 没有‘test’ # 数据描述符:一个对象同时定义了__get__()和__set__()方法,高级 - property装饰符 # 普通描述符:一种较为低级的控制属性访问机制 class Some_Class(object): _x = None def __init__(self): self._x = None @property def x(self): return self._x @x.setter def x(self, value): self._x = value @x.getter def x(self): return self._x @x.deleter def x(self): del self._x obj = Some_Class() obj.x = 10 print(obj.x + 2) # 12 print(obj.__dict__) # {'_x': 10} del obj.x print(obj.x) # None print(obj.__dict__) # {} # metaclass元类是类的模板,元类的实例为类 # 当你面临一个问题还在纠结要不要使用元类时,往往会有其他更为简单的解决方案 # 元方法可以从元类或者类中调用,不能从类的实例中调用。 # 类方法可以从类中调用,也可以从类的实例中调用 class TypeSetter(object): def __init__(self, fieldtype): print('TYpeSetter __init__', fieldtype) self.fieldtype = fieldtype def is_valid(self, value): return isinstance(value, self.fieldtype) class TypeCheckMeta(type): # type为父类,是对type的重写,作为一个元类 def __new__(cls, name, bases, dict): print('TypeCheckMeta __new__', name, bases, dict) return super(TypeCheckMeta, cls).__new__(cls, name, bases, dict) def __init__(self, name, bases, dict): self._fields = {} for key,value in dict.items(): if isinstance(value, TypeSetter): self._fields[key] = value def sayHi(cls): print('HI') class TypeCheck(object): __metaclass__ = TypeCheckMeta # 所有继承该类的子类都将使用元类来指导类的生成 # 若未设置__metaclass__,使用默认的type元类来生成类 def __setattr__(self, key, value): print('TypeCheck __setattr__') if key in self._fields: if not self._fields[key].is_valid(value): raise TypeError('Invalid type for field') super(TypeCheck, self).__setattr__(key, value) class MetaTest(TypeCheck): # 由元类 TypeCheckMeta 指导生成 name = TypeSetter(str) num = TypeSetter(int) mt = MetaTest() mt.name = 'apple' mt.num = 100 MetaTest.sayHi() # 元方法可以从元类或者类中调用,不能从类的实例中调用。 # ('TypeCheckMeta __new__', 'TypeCheck', (<type 'object'>,), {'__module__': '__main__', '__metaclass__': <class '__main__.TypeCheckMeta'>, '__setattr__': <function __setattr__ at 0x0393DA70>}) # ('TYpeSetter __init__', <type 'str'>) # ('TYpeSetter __init__', <type 'int'>) # ('TypeCheckMeta __new__', 'MetaTest', (<class '__main__.TypeCheck'>,), {'__module__': '__main__', 'num': <__main__.TypeSetter object at 0x03951D50>, 'name': <__main__.TypeSetter object at 0x03951CF0>}) # TypeCheck __setattr__ # TypeCheck __setattr__ # HI # 协议:一种松散的约定,没有相应的接口定义。 # 迭代器:统一的访问容器或集合 + 惰性求值 + 多多使用,itertools from itertools import * # print(''.join(i) for i in product('AB', repeat = 2)) for i in product('ABCD', repeat = 2): print(''.join(i)), # AA AB AC AD BA BB BC BD CA CB CC CD DA DB DC DD print for i in combinations('ABCD', 2): # AB AC AD BC BD CD print(''.join(i)), print # 生成器:按一定的算法生成一个序列。 # 生成器函数:使用了 yield,返回一个迭代器,以生成器的对象放回。 def fib(n): a, b = 1, 1 while a < n: test = (yield a) print('test:', test) a, b = b, a+b for i, f in enumerate(fib(10)): print(f), # 1 1 2 3 5 8 # 调用生成器函数时,函数体并不执行,当第一次调用next()方法时才开始执行,并执行到yield表达式后中止。 generator = fib(10) print(generator, generator.next(), generator.next()) # (<generator object fib at 0x03A39EB8>, 1, 1) print(generator.send(3)) # ('test:', 3) # yield 与 上下文管理器 结合 from contextlib import contextmanager @contextmanager def tag(name): print('<%s>' % name) yield print('<%s>' % name) with tag('hi'): print('hello') # <hi> # hello # <hi> # GIL : Global Interpreter Lock 全局解释器锁 # sys.setcheckinterval 自动线程间切换,默认每隔100个时钟 # 单核上的多线程本质上是顺序执行的 # 多核的效率比较低,考虑 multiprocessing ''' 无论使用何种语言开发,无论开发的是何种类型,何种规模的程序,都存在这样一点相同之处。 即:一定比例的内存块的生存周期都比较短,通常是几百万条机器指令的时间, 而剩下的内存块,起生存周期比较长,甚至会从程序开始一直持续到程序结束。 ''' # 引用计数算法 - 无法解决循环引用问题 - 设置threshold阈值 gc 模块 import gc print(gc.isenabled()) # True print(gc.get_threshold()) # (700, 10, 10) print(gc.garbage) # [] # 循环引用可以使一组对象的引用计数不为0,然而这些对象实际上并没有被任何外部对象所引用, # 它们之间只是相互引用。这意味着不会再有人使用这组对象,应该回收这组对象所占用的内存空间, # 然后由于相互引用的存在,每一个对象的引用计数都不为0,因此这些对象所占用的内存永远不会被释放。 a = [] b = [] a.append(b) b.append(a) print(a, b) # ([[[...]]], [[[...]]]) # python解决方案:当某些内存块M经过了3次垃圾收集的清洗之后还存活时,我们就将内存块M划到一个集合A中去, # 而新分配的内存都划分到集合B中去。当垃圾收集开始工作时,大多数情况都只对集合B进行垃圾回收, # 而对集合A进行垃圾回收要隔相当长一段时间后才进行,这就使得垃圾收集机制需要处理的内存少了,效率自然就提高了。 # 在这个过程中,集合B中的某些内存块由于存活时间长而会被转移到集合A中, # 当然,集合A中实际上也存在一些垃圾,这些垃圾的回收会因为这种分代的机制而被延迟。 # 在Python中,总共有3“代”,也就是Python实际上维护了3条链表 # PyPI : Python Package Index - Python包索引 # https://pypi.python.org/pypi/{package} # python setup.py install # PyUnit unittest模块 - 测试代码先于被测试的代码,更有利于明确需求。 # import unittest # unittest.main() # 使用 Pylint 检查代码风格 # 代码审查工具:review board # 将包发布到PyPI,供下载使用 - 这个流程需要走一遍 # 代码优化: # 优先保证代码是可工作的 # 权衡优化的代价 # 定义性能指标,集中力量解决首要问题 # 不要忽略可读性 # 定位性能瓶颈问题 - CProfile import cProfile def foo(): sum = 0 for i in range(100): sum += i return sum cProfile.run('foo()') # 针对 foo() 函数的运行时间分布统计 # 算法的评价 = 时间复杂度(重点) + 空间复杂度(硬件),一般采用以空间换时间的方法 # O(1)<O(logn)<O(n)<O(nlogn)<O(n2)<O(cn)<O(n!)<O(nn) # 循环优化:减少循环过程中的计算量,将内层计算提到上一层 # 使用不同的数据结构优化性能 # 列表 list # 栈和队列 deque # heapify()将序列容器转化为堆 heapq import heapq import random list0 = [random.randint(0,100) for i in range(10)] print(list0) #[55, 62, 17, 56, 82, 45, 87, 48, 65, 32] heapq.heapify(list0) print(list0) # [17, 32, 45, 48, 62, 55, 87, 56, 65, 82] import array a = array.array('c', 'string') print(a.tostring()) # string import sys print(sys.getsizeof(a)) # 28 print(sys.getsizeof(list('string'))) # 72 import timeit t = timeit.Timer("''.join(list('string'))") print(t.timeit()) # 0.688437359057 t = timeit.Timer("a.tostring()", "import array; a = array.array('c', 'string')") print(t.timeit()) # 0.163860804606 # set 集合的使用 list0 = [i for i in range(10)] list1 = [i for i in range(20)] print(set(list0)&set(list1)) # set([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]) # 进程同步:multiprocessing - Pipe, Queue - 解决多核下的GIL效率问题 # 线程同步:threading - Lock, Event, Condition, Semaphore # 线程的生命周期:创建,就绪,运行,阻塞,终止 # 避免多次创建线程 - 线程池 threadpool