覆盖变量名是否应该带来性能优势？

Question

When parsing objects out of large datasets, I often find myself aggregating information into set objects, then transforming them into lists to sort them.

For example, here might be a code snippet:

all_times = set([])

for row in dataset:
   time = parse_out_time(row)
   all_times.add(time)

sorted_times = sorted(list(all_times))

My question is about that last assignment. I could reassign the same variable name with the sorted list

all_times = sorted(list(all_times))

I know that python has automatic garbage collection to remove the data assigned to old variable names that are not reused. This approach seems like it would allow the Python interpreter to immediately de-allocate the memory belonging to the old set version of all_times. If we ran the above code in a loop over a few million datasets, this could be important.

Should writing over variable names that you will never use again give any performance benefits? Or is Python's garbage collector smart enough to immediately de-allocate the memory for variables that are never called again by the script?

Answer 1

Python doesn't do any static analysis of the code. It will maintain the reference count for an object bound to a variable until that variable goes out of scope (for instance, on return ), the variable is reassigned ( all_times = sorted(list(all_times) ) or is deleted ( del all_times ). In the case of a set , you could also do all_times.clear() to get rid of the contained data. All four are reasonable ways to get rid of a container that is no longer needed.

Notice in any case that the contained data is still in sorted_times . All you got rid of is the hash tables used by the set. Its likely not that big of a help either way.

Answer 2

Il don't think it would cost so much resources if you do it right (depending on your needs).

I mean, If you run this code into a loop, you will still only use 2 variables.

Things will be slightly different if you put this code into a function, and run the function into a loop. And it would consume a little more if you run this function as multithreaded tasks.