Names, identity, mutability, copying and hashability

1. Shallow and deep copying

   Stas' Python Cookbook - selected path

   Lists: Copying (shallow vs deep), including the listB = listA warning

   Done
2. Tuples, sets and dictionaries

   Stas' Python Cookbook - selected path

   Tuples, Sets and Dictionaries: construction, mutation, frozenset, key rules and insertion order

   Done
3. Objects, values and types

   Python language reference

   Section 3.1: identity, type and value; mutable versus immutable; the a = 1; b = 1 reuse rule

   Done
4. Default equality and hashing

   Python language reference

   object.__eq__ and object.__hash__: default identity equality; defining __eq__ without __hash__ removes hashing

   Done

Focus on

• Assignment binds a name to an object. is tests identity and == tests value; CPython reuses small immutable objects, but the reference says reuse may or may not happen, so never branch on it.
• Two names can own one mutable object. A shallow copy shares every nested object; only a deep copy duplicates them. Copying an immutable such as a tuple or str may return the same object.
• A key must stay hashable and equal for its lifetime in the container. A tuple is hashable only when all its items are; NaN never equals itself, so a fresh NaN misses a stored NaN key.
• The cookbook says tuples are hashable because they are immutable. That shortcut is wrong for tuples that hold mutables; hash((1, [2])) raises TypeError.

Reproduce closed-book

Traps

• Comparing values with is. Constant folding makes 257 is 257 true inside one compilation unit, and CPython warns (SyntaxWarning) on is with a literal, so is proves nothing about equality.
• dict.fromkeys(keys, []) shares one list across all keys; appending through one key shows up under every key.
• Reading d[float('nan')] after storing under another NaN. Lookup needs equality with the stored key, and NaN never equals NaN.

Practice

The section mini-project covers this: diagnosing and fixing the seeded aliasing defect with a regression test.

Self-test

1. After a = [1, [2]]; b = a; c = copy.copy(a); b.append(3); b[1].append(4), what are a and c?

   Show answer Got it Missed

2. Why does hash((1, 2)) work while hash((1, [2])) raises TypeError?

   Show answer Got it Missed

3. d = {}; k = float('nan'); d[k] = 1. Why does d[float('nan')] raise KeyError?

   Show answer Got it Missed

4. Give a case where x is y holds but x == y is False.

   Show answer Got it Missed

Leads to

• 3.2 Argument binding, scope and closures Follow mutation across call boundaries and mutable defaults.
• 3.9 Reference counts, cycle collection and weak references Connect aliasing to object lifetimes and weak references.

Comes after

• 2.5 Make the experiment reproducible and explain its limitsAfter completing the first supervised ML project, deepen the Python object model and runtime behavior behind reliable code.

Resources for this unit

• Stas' Python Cookbook - selected path
  Stas Bekmanprimarycourse
  Run Part I on dictionaries, iteration, functions, closures, decorators, classes and context managers; Part II on modules, files, paths, environment and arguments; and Part IV on versions, dependencies, packaging and code quality. Finish by building one src-layout package from a blank directory.
• Python language reference
  Python Software Foundationrequiredreference
  Check Data model, Execution model, Import system and Exceptions whenever a claim depends on Python semantics. These pages are the authority when the course and a runtime observation disagree.