Lesson 5. Truthiness, None and Equality

Mission link: Two idioms that read as "if this has a value" behave differently on 0, "" and []. Choosing between them by feel produces bugs that only appear on valid input.
Primary source: The Python Standard Library, Truth Value Testing
Prerequisites: Lesson 1, Lesson 4

Warm-up

1. ▢ You need to know whether an optional setting was provided at all, as opposed to reading it with a fallback. Which dictionary idiom says that?

2. ▢ What is the difference between a is b and a == b?

Know this

Every object in Python can be used in a boolean context, and each type decides for itself what that means (truth value testing).

Falsy: False, None, zero of any numeric type (0, 0.0, 0j), and every empty container ("", (), [], {}, set(), range(0)).
Truthy: everything else, including "0", "False", [0], {} with anything in it, and any object of a class that does not say otherwise.

The two strings in that list of truthy values are worth a second look. "0" is a non-empty string, so it is true.

if x: and if x is not None: are different questions

def apply(retries=None):
    if retries:                 # WRONG when retries == 0
        ...
    if retries is not None:     # asks whether an argument was given
        ...

Pass retries=0 and the first branch is skipped, even though 0 was a deliberate, valid choice. The same trap applies to "" as a deliberate empty prefix, and to [] as a deliberate empty list.

Use if x: when you mean "is there anything here", which is the right question for a container you are about to iterate. Use if x is not None: when you mean "was a value supplied", which is the right question for an optional parameter.

None is a singleton, so is is the correct operator for it and == None is discouraged: a class can define __eq__ to claim equality with anything, and is None cannot be lied to.

The or default trap

def connect(timeout=None):
    timeout = timeout or 30     # rejects 0

or returns its first truthy operand, so 0 or 30 is 30. The caller asked for no timeout and got thirty seconds. This idiom is fine when zero and empty are genuinely not valid inputs, and a bug the moment they are. The safe spelling is explicit:

timeout = 30 if timeout is None else timeout

and and or return one of their operands rather than a boolean, and both short-circuit: the right side is not evaluated if the left settles the answer. That is what makes if data and data[0] safe on an empty list.

Equality is defined by the type

1 == 1.0            # True, numeric types compare across types
True == 1           # True, bool is a subclass of int
"1" == 1            # False, no implicit conversion between str and int
[1, 2] == [1, 2]    # True, sequences compare element by element

Python does not coerce strings and numbers for comparison, so "1" == 1 is simply false rather than an error or a surprise.

Comparisons chain, and they mean what mathematical notation means:

0 <= x < 10         # one expression, x evaluated once

Ordering comparisons between unrelated types raise TypeError, which is deliberate: 1 < "a" is a bug, not a question.

Practice

1. ▢ Which of these are falsy?

   0, "0", [], [[]], "", " ", {}, {0: 0}, None, False, 0.0, "False"
   

2. ▢ Predict what this prints for each call: apply(), apply(0), apply(5).

   def apply(retries=None):
       if retries:
           print("using", retries)
       else:
           print("using default")
   

3. ▢ Fix this line so that a caller can ask for no timeout at all, and say why the original fails.

   timeout = timeout or 30
   

4. ▢ Which comparison is correct for detecting that an optional argument was omitted?

   • a) if value == None:
   • b) if value is None:
   • c) if not value:
   • d) if value != True:

5. ▢ Both lines below are common in real code. For each, say what input makes it wrong.

   if len(items) > 0:
       ...
   if items != []:
       ...
   

Real-world reps

• [ ] Write the apply function from practice 2 and call it with no argument, with 0, and with 5. Then change the condition to is not None and call all three again.
• [ ] Search code you know for or used to supply a default. For each hit, decide whether zero or an empty string is a legal input. Every yes is a live bug.
• [ ] Tomorrow: pick a function with an optional parameter and write down, in words, whether it distinguishes "not supplied" from "supplied as empty". If the function cannot, decide whether it should.

Going further

• Truth Value Testing: the full rule, including how a class opts out
• Comparisons: chaining, identity, and what raises
• Boolean operations: why and and or return an operand rather than a boolean
• Resources

Not landing? Reread the primary source at the top, since this lesson compresses it and compression is where understanding leaks. Check the glossary for any term that felt slippery.

If the lesson itself is unclear rather than the material, that is a defect: open an issue.