The short answer
sys.getsizeof(obj) returns the size of the object itself in bytes (including CPython object overhead for that container), not the size of all objects it references.
Examples of what it misses:
- Elements inside a list/dict
- Attributes on an instance
- Nested structures
- Allocator arena overhead / fragmentation
- Interpreter and module baseline RSS
So a list of 1 million strings can show a modest getsizeof(list) while process RSS is huge because of the string objects.
import sys
s = "x" * 1000
lst = [s] * 1000 # 1000 references to the SAME string
print(sys.getsizeof(s)) # size of one str object
print(sys.getsizeof(lst)) # size of list object / pointer array
# NOT 1000 * sizeof(s), and sharing means unique payload is one string
nested = {"a": [1, 2, 3], "b": [1, 2, 3]}
print(sys.getsizeof(nested)) # dict shell only
# Rough deep size sketch (not perfect; watch for shared/cyclic refs)
def deep_getsizeof(obj, seen=None):
if seen is None:
seen = set()
obj_id = id(obj)
if obj_id in seen:
return 0
seen.add(obj_id)
size = sys.getsizeof(obj)
if isinstance(obj, dict):
size += sum(deep_getsizeof(k, seen) + deep_getsizeof(v, seen) for k, v in obj.items())
elif isinstance(obj, (list, tuple, set, frozenset)):
size += sum(deep_getsizeof(i, seen) for i in obj)
return sizeFor process-level usage, inspect RSS via OS tools (ps, resource, tracemalloc, memory profilers), not getsizeof alone.