architecture_notes.md

Chapter 12 — Architecture Notes: Slices at the C Level

---

1. The slice Object

In Python, a:b:c is syntactic sugar. When used inside square brackets [a:b:c], the CPython parser translates it into a call to the built-in slice() function.

# These are evaluated identically by CPython:
my_list[1:10:2]
my_list[slice(1, 10, 2)]

At the C level, a slice object is a simple struct (PySliceObject) containing three PyObject pointers: start, stop, and step.

---

2. The indices() Method

One of the trickier parts of slicing is dealing with missing bounds and negative indices (e.g., [-5:]). Calculating the actual start and stop integers for an array of a specific length is non-trivial.

To solve this, CPython exposes an indices(length) method on slice objects.

s = slice(-3, None, None)
print(s.indices(10))  # Output: (7, 10, 1)

The indices() method takes the length of the target sequence and calculates the exact, positive integer coordinates (start, stop, step) needed to execute the slice. It handles negative indices, out-of-bounds indices, and missing steps.

---

3. How __getitem__ Receives Slices

When you write a custom sequence in Python, your __getitem__ is bound to the C-level sq_item or mp_subscript slots. In Python 3, CPython routes most bracket lookups through mp_subscript (the mapping protocol).

1. User writes v[1:4].
2. CPython parser builds a slice(1, 4, None) object.
3. CPython looks up the __getitem__ method on your class.
4. It passes the slice object as the index parameter.
5. It is entirely up to the Python-level __getitem__ method to inspect the type of index using isinstance(index, slice) and react appropriately.