讲解Python中的递归函数
本文的最重要的收获在于:尾递归是指,在函数返回的时候,调用自身本身,并且,return语句不能包含表达式。
在函数内部,可以调用其他函数。如果一个函数在内部调用自身本身,这个函数就是递归函数。
举个例子,我们来计算阶乘n! = 1 x 2 x 3 x ... x n,用函数fact(n)表示,可以看出:
1
|
fact(n) = n! = 1 x 2 x 3 x ... x (n - 1 ) x n = (n - 1 )! x n = fact(n - 1 ) x n |
所以,fact(n)可以表示为n x fact(n-1),只有n=1时需要特殊处理。
于是,fact(n)用递归的方式写出来就是:
1
2
3
4
|
def fact(n): if n = = 1 : return 1 return n * fact(n - 1 ) |
上面就是一个递归函数。可以试试:
1
2
3
4
5
6
|
>>> fact( 1 ) 1 >>> fact( 5 ) 120 >>> fact( 100 ) 93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000L |
如果我们计算fact(5),可以根据函数定义看到计算过程如下:
1
2
3
4
5
6
7
8
9
10
|
= = = > fact( 5 ) = = = > 5 * fact( 4 ) = = = > 5 * ( 4 * fact( 3 )) = = = > 5 * ( 4 * ( 3 * fact( 2 ))) = = = > 5 * ( 4 * ( 3 * ( 2 * fact( 1 )))) = = = > 5 * ( 4 * ( 3 * ( 2 * 1 ))) = = = > 5 * ( 4 * ( 3 * 2 )) = = = > 5 * ( 4 * 6 ) = = = > 5 * 24 = = = > 120 |
递归函数的优点是定义简单,逻辑清晰。理论上,所有的递归函数都可以写成循环的方式,但循环的逻辑不如递归清晰。
使用递归函数需要注意防止栈溢出。在计算机中,函数调用是通过栈(stack)这种数据结构实现的,每当进入一个函数调用,栈就会加一层栈帧,每当函数返回,栈就会减一层栈帧。由于栈的大小不是无限的,所以,递归调用的次数过多,会导致栈溢出。可以试试fact(1000):
1
2
3
4
5
6
7
|
>>> fact( 1000 ) Traceback (most recent call last): File "<stdin>" , line 1 , in <module> File "<stdin>" , line 4 , in fact ... File "<stdin>" , line 4 , in fact RuntimeError: maximum recursion depth exceeded |
解决递归调用栈溢出的方法是通过尾递归优化,事实上尾递归和循环的效果是一样的,所以,把循环看成是一种特殊的尾递归函数也是可以的。
尾递归是指,在函数返回的时候,调用自身本身,并且,return语句不能包含表达式。这样,编译器或者解释器就可以把尾递归做优化,使递归本身无论调用多少次,都只占用一个栈帧,不会出现栈溢出的情况。
上面的fact(n)函数由于return n * fact(n - 1)引入了乘法表达式,所以就不是尾递归了。要改成尾递归方式,需要多一点代码,主要是要把每一步的乘积传入到递归函数中:
1
2
3
4
5
6
7
|
def fact(n): return fact_iter( 1 , 1 , n) def fact_iter(product, count, max ): if count > max : return product return fact_iter(product * count, count + 1 , max ) |
可以看到,return fact_iter(product * count, count + 1, max)仅返回递归函数本身,product * count和count + 1在函数调用前就会被计算,不影响函数调用。
fact(5)对应的fact_iter(1, 1, 5)的调用如下:
1
2
3
4
5
6
7
|
= = = > fact_iter( 1 , 1 , 5 ) = = = > fact_iter( 1 , 2 , 5 ) = = = > fact_iter( 2 , 3 , 5 ) = = = > fact_iter( 6 , 4 , 5 ) = = = > fact_iter( 24 , 5 , 5 ) = = = > fact_iter( 120 , 6 , 5 ) = = = > 120 |
尾递归调用时,如果做了优化,栈不会增长,因此,无论多少次调用也不会导致栈溢出。
遗憾的是,大多数编程语言没有针对尾递归做优化,Python解释器也没有做优化,所以,即使把上面的fact(n)函数改成尾递归方式,也会导致栈溢出。
有一个针对尾递归优化的decorator,可以参考源码:
http://code.activestate.com/recipes/474088-tail-call-optimization-decorator/
我们后面会讲到如何编写decorator。现在,只需要使用这个@tail_call_optimized,就可以顺利计算出fact(1000):
1
2
|
>>> fact( 1000 ) 402387260077093773543702433923003985719374864210714632543799910429938512398629020592044208486969404800479988610197196058631666872994808558901323829669944590997424504087073759918823627727188732519779505950995276120874975462497043601418278094646496291056393887437886487337119181045825783647849977012476632889835955735432513185323958463075557409114262417474349347553428646576611667797396668820291207379143853719588249808126867838374559731746136085379534524221586593201928090878297308431392844403281231558611036976801357304216168747609675871348312025478589320767169132448426236131412508780208000261683151027341827977704784635868170164365024153691398281264810213092761244896359928705114964975419909342221566832572080821333186116811553615836546984046708975602900950537616475847728421889679646244945160765353408198901385442487984959953319101723355556602139450399736280750137837615307127761926849034352625200015888535147331611702103968175921510907788019393178114194545257223865541461062892187960223838971476088506276862967146674697562911234082439208160153780889893964518263243671616762179168909779911903754031274622289988005195444414282012187361745992642956581746628302955570299024324153181617210465832036786906117260158783520751516284225540265170483304226143974286933061690897968482590125458327168226458066526769958652682272807075781391858178889652208164348344825993266043367660176999612831860788386150279465955131156552036093988180612138558600301435694527224206344631797460594682573103790084024432438465657245014402821885252470935190620929023136493273497565513958720559654228749774011413346962715422845862377387538230483865688976461927383814900140767310446640259899490222221765904339901886018566526485061799702356193897017860040811889729918311021171229845901641921068884387121855646124960798722908519296819372388642614839657382291123125024186649353143970137428531926649875337218940694281434118520158014123344828015051399694290153483077644569099073152433278288269864602789864321139083506217095002597389863554277196742822248757586765752344220207573630569498825087968928162753848863396909959826280956121450994871701244516461260379029309120889086942028510640182154399457156805941872748998094254742173582401063677404595741785160829230135358081840096996372524230560855903700624271243416909004153690105933983835777939410970027753472000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 |
小结
使用递归函数的优点是逻辑简单清晰,缺点是过深的调用会导致栈溢出。
针对尾递归优化的语言可以通过尾递归防止栈溢出。尾递归事实上和循环是等价的,没有循环语句的编程语言只能通过尾递归实现循环。
Python标准的解释器没有针对尾递归做优化,任何递归函数都存在栈溢出的问题。