What is the term to describe this property of how lists work?

Question

I am looking for the proper term to describe this well-known property of collection objects, and more importantly, the way the stack diagram changes when variables are used to reference their elements:

>>> x = 5
>>> l = [x]
>>> x += 1
>>> l
[5]
>>> x
6

What is the name of what the list is doing to the variable x to prevent it from being bound to any changes to the original value of x? Shielding? Shared structure? List binding? Nothing is coming back from a Google search using those terms.

Here's an example with more detail (but doesn't have a definition, unfortunately).

_{Credit to ocw.mit.edu}

Answer 1

What is the name of what the list is doing to the variable x to prevent it from being bound to any changes to the original value of x? Shielding? Shared structure? List binding? Nothing is coming back from a Google search using those terms.

Because the list isn't doing anything, and it isn't a property of collections.

In Python, variables are names.

>>> x = 5

This means: x shall be a name for the value 5.

>>> l = [x]

This means: l shall be a name for the value that results from taking the value that x names (5), and making a one-element list with that value ([5]).

>>> x += 1

x += 1 gets rewritten into x = x + 1 here, because integers are immutable. You cannot cause the value 5 to increase by 1, because then it would be 5 any more.

Thus, this means: x shall stop being a name for what it currently names, and start being a name for the value that results from the mathematical expression x + 1. I.e., 6.

That's how it happens with reference semantics. There is no reason to expect the contents of the list to change.

---

Now, let's look at what happens with value semantics, in a hypothetical language that looks just like Python but treats variables the same way they are treated in C.

>>> x = 5

This now means: x is a label for a chunk of memory that holds a representation of the number 5.

>>> l = [x]

This now means: l is a label for a chunk of memory that holds some list structure (possibly including some pointers and such), which will be initialized in some way so that it represents a list with 1 element, that has the value 5 (copied from the x variable). It cannot be made to contain x logically, since that is a separate variable and we have value semantics; so we store a copy.

>>> x += 1

This now means: increment the number in the x variable; now it is 6. The list is, again, unaffected.

Regardless of your semantics, you cannot affect the list contents this way. Expecting the list contents to change means being inconsistent in your interpretations. (This becomes more obvious if you rewrite the code to l = [5]; x = l[0]; x += 1.)

Answer 2

I would call it "immutability" of the contained object.

I think you compare your situation with the following one:

x = []
l = [x]
x += [1]
print l # --> [[1]]

The difference is:

In this situation (the mutable situation), you mutate your original object x which is contained in the list l.

In your situation however, you have x point to an immutable object (5), which is then added to the list. Afterwards, this reference is replaced by the one to 6, but only for x, not for the list.

So x += <something> either modifies x or replaces it with another object depending on the nature of the object type.

---

EDIT: It as well has nothing to do with the nature of lists. You can achieve the same with 2 variables:

x = 5
y = x
print x, y, x is y
x += 1
print x, y, x is y

vs.

x = []
y = x
print x, y, x is y
x += [1]
print x, y, x is y

The first one will change x due to the immutability of ints, resulting to x is y being false, while in the second one, x is y remains true because the object (the list) is mutated and the identity of the object referenced to by x and y stays the same.

Answer 3

The only thing list is doing to x in your example is reading it; it interacts with the variable in no other way. In fact the list generated by the expression [x] does not interact with the variable x at all.

If you had to introduce a piece of jargon for it, it would be value-capture, or perhaps simply independence.

I think the reason why there isn't a special term for this is that it is (a) not something that requires much discussion (b) is an aspect of strict by-value semantics where variables always hold references. Those kinds of semantics are pretty much the norm now (except where it is by reference with variables actually naming bits of memory containing objects). I think OP was expecting by-name or lazy semantics.

Answer 4

The behavior you're describing has to do with references. If you're familiar with c, you're probably also familiar with "pointers." Pointers in c can get really complicated, and Python uses a data model that greatly simplifies things. But having a bit of background in c helps understand Python's behavior here, which is closely related to the behavior of c pointers. So "pointer," "reference," and "dereference" are all terms that are related to what you're talking about, although none of them is quite a "name" for it. Perhaps the best name for it is "indirection" -- though that's a bit too abstract and inclusive; this is a very specific kind of indirection. Perhaps "reference semantics"? Here's a slide from a talk by GvR himself that uses that term, and a google search turns up a few useful hits.

But if you don't have any background in c, here's my best explanation. In short, you can think of a Python name as a pointer to an object. So when you assign a name to a value, you're "pointing" that name at the value. Then, when you assign a new value to the name, you point it at a new value; but the old value isn't changed at all as a result. This seems natural when thinking about names as pointers here; the "value of the name" is changed, but the "value of the value" is not.

The thing that's a bit confusing is that += behaves inconsistently. When you use += on a number, the result makes perfect sense using the above metaphor:

x = 5
y = x
x += 1
print x, y 
# 6 5

The behavior is exactly the same as it would be if you did x = x + 1.

But sometimes += is overloaded in a way that does in-place mutation. It's a pragmatic, but slightly inconsistent, idiom:

x = [5]
y = x
x += [6]
print x, y
# [5, 6] [5, 6]

So here, the "value of the value" is changed. This is actually not my favorite thing about Python, but there are good reasons for it.

Answer 5

What you're effectively doing is constructing a list with one element that references the same object as x. Then you bind a new value to x while the list still references the old element. This is because += returns a reference to a new object (6) and leaves the old 5 object intact. Integers are immutable in Python.