Hacker News Reader

Comment by joatmon-snoo | Hacker News Reader

Strictly speaking, lists are not arrays.

I'm not aware of what the specification for a list would be, this is the closest I'm aware of: https://docs.python.org/3/reference/datamodel.html#objects-v..., 3.2 > Sequences > Mutable Sequences > Lists, "The items of a list are arbitrary Python objects. Lists are formed by placing a comma-separated list of expressions in square brackets"

CPython implementation is here: https://github.com/python/cpython/blob/e42b705188271da108de4...

If I'm talking about arrays with someone else, I would expect us to be thinking of arrays in the sense of C. Both C++ and Java use the same definition. Other languages that I'm aware of (e.g. Javascript, Go) have their own notions of arrays that you can't conflate with the C notion of an array.

From the C11 spec, §6.2.5.20: "An array type describes a contiguously allocated nonempty set of objects with a particular member object type, called the element type."

View on HN

Lists are very often used in the way arrays are, though.

The heterogeneous use-case (that in C, etc. is usually a struct) is often filled by either tuples or (ordered) dictionaries in Python.

If you really want something a bit closer to a C array, you can fairly easily create a UserList that enforces the non-empty requirement and only allows items of a particular type.

If you want some closer to C arrays, you can use Python array.array, which is part of the standard library.

https://docs.python.org/3/library/array.html

https://stackoverflow.com/questions/176011/python-list-vs-ar...

Thanks! I had forgotten about that module.

I think you have a point about C or C++ notion of arrays having the same type.

I'll probably rewrite that paragraph, since the main point is to tell people unfamiliar with Python that lists aren't actually linked lists.

So... Because lists are dynamically typed and heterogeneous, does that mean the underlying C is basically a contiguous segment of memory of python object references? Then in cpython the actual list contents are Python objects sitting on the heap in a non-contiguous manner?

And when the list needs to be resized and doubles itself, it's just doubling the space for object references?

> So... Because lists are dynamically typed and heterogeneous, does that mean the underlying C is basically a contiguous segment of memory of python object references?

Yes. Every Python object is held by the interpreter as a pointer to a PyObject struct on the (C) heap.

Not something I would put into production, but had a fun application of this on a side project. I had the `id` of a function (but not the function object itself) and needed to recover the function. In CPython, `id` of a function corresponds to it's memory address (not sure if that can be overridden). By casting the id to a PyObject, I was able to recover the original.

    func = ctypes.cast(int(address), ctypes.py_object).value

Was kinda cool actually seeing that work.

> Because lists are dynamically typed and heterogeneous, does that mean the underlying C is basically a contiguous segment of memory of python object references?

Yep, it is an array of pointers: https://github.com/python/cpython/blob/3.7/Include/listobjec...

> And when the list needs to be resized and doubles itself, it's just doubling the space for object references?

Basically yes. The growth pattern is not as aggressive as doubling though: https://github.com/python/cpython/blob/3.7/Objects/listobjec...

I keep forgetting just how readable cpython source is. I'm still mentally stuck in my beginner years, where it was all Greek to me.

Thanks for sharing these links.

Interesting:

"The growth pattern is: 0, 4, 8, 16, 25, 35, 46, 58, 72, 88, ..."

What is the big-o amortized cost of using this pattern? Doubling has a easy proof that it's O(1) amortized.

The growth pattern is still exponential, instead of going x -> 2 * x, you go x -> 9/8 * x + 6. I think it should still be O(1) amortized, you just need to adjust your proof for doubling a bit.

Mathematically the best is not doubling but fibonacci. for the start doubling is used because it's much faster to compute.

Reply on news.ycombinator.com