Iterating over processed paths in a for loop

[gardiens]

Hello I have a question,
I have a pytorch geometric dataset and I need at some point to iterate over all my processed paths like this :
`
class myclass(InMemoryDataset):
def init(self):
# Do the usual init
self.stage="true"
self.pre_transform_hash="non"
self.stage="oui"

    self.cloud_ids=[x for x in range(1000000)]
def my_test(self):
    start=time.time()
    l_path=self.processed_file_names
    for i in range(len(l_path)):
        print("the whole loading took",time.time()-start)
        start=time.time()
        l_path[i]
        print("accessing path i took",time.time()-start)
        start=time.time()
        self.processed_file_names[i]
        print("accessing with processed_files_names took",time.time()-start)
@property
def processed_file_names(self):
    return [
        osp.join(self.stage, self.pre_transform_hash, f"{w}.h5")
        for w in self.cloud_ids
    ]

`
Accessing the value in assigned_process paths took around 1e-5 second however if I don't assign a variable:

each iteration took over 2 seconds, why? Is processed_paths called every time in the for loop and never cached?
I know my question is quite dumb and not exactly pytorch geometric related