理解Python中的GC

Question

我正在阅读 https://devguide.python.org/garbage_collector/ （和 https://docs.python.org/3.8/library/gc.html）来了解Python中GC背后的机制。世代阈值的概念对我来说有点不清楚。提到的文章（以及许多其他文章）说

当世代包含的对象数量达到某个预定义阈值时，就会收集世代

。默认阈值为 (700, 10, 10)。美好的。

我正在运行以下代码以更好地理解该机制，并且输出与我所理解的阈值定义相矛盾：

class A():
    pass

lst = []
gc_lst = []

def strt(p, inf):
    s = "phase = %s\ninf = %s\ngc.get_count() = %s\ngc.get_stats() = %s\nlen(gc.get_objects()) = %s"\
        %(p, inf, gc.get_count(), gc.get_stats(), [len(gc.get_objects(0)),len(gc.get_objects(1)),len(gc.get_objects(2))])
    gc_lst.append(s)

gc.callbacks.append(strt)
s = "gc.get_count() = %s\ngc.get_stats() = %s\nlen(gc.get_objects() for each gen) = %s"\
%(gc.get_count(), gc.get_stats(), [len(gc.get_objects(0)),len(gc.get_objects(1)),len(gc.get_objects(2))])
print(s)

for k in range(8000):
    lst.append(A())
    gc_lst.append(k)

for k in gc_lst:
    print(k)

输出：

0
1
2
...
...
...
5413
5414
phase = start
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (701, 8, 1)
gc.get_stats() = [{'collections': 19, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [703, 5475, 5069]
phase = stop
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (0, 9, 1)
gc.get_stats() = [{'collections': 20, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [2, 6176, 5069]
5415
5416
...
6114
6115
phase = start
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (701, 9, 1)
gc.get_stats() = [{'collections': 20, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [703, 6176, 5069]
phase = stop
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (0, 10, 1)
gc.get_stats() = [{'collections': 21, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [2, 6877, 5069]
6116
6117
...
6815
6816
phase = start
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (701, 10, 1)
gc.get_stats() = [{'collections': 21, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [703, 6877, 5069]
phase = stop
inf = {'generation': 0, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (0, 11, 1)
gc.get_stats() = [{'collections': 22, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [2, 7578, 5069]
6817
6818
...
7516
7517
phase = start
inf = {'generation': 1, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (701, 11, 1)
gc.get_stats() = [{'collections': 22, 'collected': 65, 'uncollectable': 0}, {'collections': 1, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [703, 7578, 5069]
phase = stop
inf = {'generation': 1, 'collected': 0, 'uncollectable': 0}
gc.get_count() = (0, 0, 2)
gc.get_stats() = [{'collections': 22, 'collected': 65, 'uncollectable': 0}, {'collections': 2, 'collected': 33, 'uncollectable': 0}, {'collections': 0, 'collected': 0, 'uncollectable': 0}]
len(gc.get_objects()) = [2, 0, 13341]
7518
7519
...
7998
7999

我看到提到的阈值适用于集合，而不是生成中的对象。晚辈每人收集700件物品。下一代是每 10 个年轻一代的集合（每 7000 个对象）收集一次，按照这个逻辑，最老的一代应该每 70000 个对象收集一次（忽略二次时间问题的修复https://bugs.python.org/issue4074）。我在定义中缺少什么？