Python Forum
Cheat against recursion limit
Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Cheat against recursion limit
#1
My question was: Is it possible to bypass the recursion limit with generators and yield from?
Answer: You can prevent the RecursionError, but you are still limited in recursion.


Here a function to generate a nested list:
def construct_deep_list(n): 
    x = y = [] 
    for _ in range(n): 
        x.append([]) 
        x = x[0] 
    x.append('Secret') 
    return y
construct_deep_list(10)
Output:
[[[[[[[[[[['Secret']]]]]]]]]]]
Now getting the recursionlimit, generating the deep_list and printing it to console.
import sys


deep_list = construct_deep_list(sys.getrecursionlimit())
print(deep_list)
Then you get this error:
Error:
RecursionError Traceback (most recent call last) <ipython-input-158-1a3df04d5171> in <module> 1 deep_list = construct_deep_list(sys.getrecursionlimit()) ----> 2 print(deep_list) RecursionError: maximum recursion depth exceeded while getting the repr of an object


Now a normal function to traverse the deep list.
str and bytes are not allowed to prevent more
complicated code.
def test_recursion(iterable):
    try:
        iterator = iter(iterable)
    except TypeError:
        return iterable
    else:
        while True:
            try:
                element = next(iterator)
                if isinstance(element, (str, bytes)):
                    raise ValueError('str or bytes are not allowed')
                return test_recursion(element)
            except StopIteration:
                break
Testing the functions:
result = test_recursion([[True]])
print(result)
It works:
Output:
True
Now let's try it with the deep_list:
result2 = test_recursion(deep_list)
Error:
RecursionError Traceback (most recent call last) <ipython-input-197-311c4b04144f> in <module> ----> 1 result2 = test_recursion(deep_list) <ipython-input-190-48a43684de23> in test_recursion(iterable) 13 if isinstance(element, (str, bytes)): 14 raise ValueError('str or bytes are not allowed') ---> 15 return test_recursion(element) 16 except StopIteration: 17 break ... last 1 frames repeated, from the frame below ... <ipython-input-190-48a43684de23> in test_recursion(iterable) 13 if isinstance(element, (str, bytes)): 14 raise ValueError('str or bytes are not allowed') ---> 15 return test_recursion(element) 16 except StopIteration: 17 break RecursionError: maximum recursion depth exceeded in __instancecheck__
The deep_list has one str inside. But we got the RecursionError and not the ValueError.
This means, the interpreter was not able to get finally the str element. The recursion limit was hit before.


Now let's try it with the generator version:
def deep_iter(iterable):
    for element in iterable:
        try:
            yield from deep_iter(element)
        except:
            yield element
result3 = list(deep_iter(deep_list))
print(result3)
Output:
[[[[[[[[[[[[[[[[[[['Secret']]]]]]]]]]]]]]]]]]]
Conclusion: Generators can simplify your code. But I still do not know why the yield from does not hit the recursion limit.
I guess the recursion limit is still thrown, but goes in silence away.


Apply the generator twice gives:
result4 = list(deep_iter(deep_iter(deep_list)))
print(result4)
Output:
['Secret']
Extended generator, which does not iterate over str and bytes:
def deep_iter2(iterable, ignore=(str, bytes)):
    for element in iterable:
        if isinstance(element, ignore):
            yield element
            continue
        try:
            yield from deep_iter2(element)
        except:
            yield element
print(list(deep_iter2(deep_iter2(deep_list))))                                                
Output:
['Secret']
I'm not sure if it's usable in production code. But usually you should not have deep nested data structures in production code (hopefully).

Instead of using yield from, I use the deque in the final version.
This also faster.
from collections import deque


def deep_iter3(iterable, ignore=(str, bytes)):
    heap = deque([iterable])
    while heap:
        element = heap.popleft()
        if isinstance(element, ignore):
            yield element
            continue
        try:
            sub_heap = []
            for sub_element in element:
                sub_heap.append(sub_element)
        except TypeError:
            yield element
        else:
            heap.extendleft(reversed(sub_heap))
list(deep_iter3([1,2, ['test'], (3,1),4, deep_list]))
Output:
[1, 2, 'test', 3, 1, 4, 'Secret']
Some benchmarks:
Output:
In [83]: %timeit list(deep_iter3([1,2, ['test'], (3,1),4, deep_list])) 1.26 ms ± 8.92 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) In [84]: %timeit list(deep_iter2([1,2, ['test'], (3,1),4, deep_list])) 17.9 ms ± 33.8 µs per loop (mean ± std. dev. of 7 runs, 100 loops each) In [85]: %timeit list(deep_iter([1,2, ['test'], (3,1),4, deep_list])) 85.6 ms ± 167 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
Almost dead, but too lazy to die: https://sourceserver.info
All humans together. We don't need politicians!
Reply
#2
On topic of Python recursion: A thing I learned about Python recursion

Quote:In Python 3, all comprehensions (and in Python 2 all except list comprehensions) are actually compiled as nested functions. Executing the generator comprehension calls that hidden nested function, using up an extra stack frame.
I'm not 'in'-sane. Indeed, I am so far 'out' of sane that you appear a tiny blip on the distant coast of sanity. Bucky Katt, Get Fuzzy

Da Bishop: There's a dead bishop on the landing. I don't know who keeps bringing them in here. ....but society is to blame.
Reply
#3
Interesting article. I like the pragmatic talks from Ned Batchelder.
I added a comment to the article with a link to this thread.
Almost dead, but too lazy to die: https://sourceserver.info
All humans together. We don't need politicians!
Reply


Forum Jump:

User Panel Messages

Announcements
Announcement #1 8/1/2020
Announcement #2 8/2/2020
Announcement #3 8/6/2020