I’ve been learning a lot about Python recently. Here are some interesting things I’ve learned.

  • What’s the difference between @classmethod and @staticmethod? @classmethod gets it’s class (cls by convention) as an implicit first argument. This means that it can modify class state if it wants. Static methods can’t do this.

  • Python supports generic function in the style of R S3 methods through the @singledispatch decorator.

  • If you need to iterate through something in reverse, use reversed().

  • If you need to iterate through two things at once, use zip(). It will stop the iteration when the smaller of the two iterators stops.
    • You can pass a zip object to the dictionary constructor like this: dict(zip(a, b)) to make a dictionary with keys from a and corresponding values from b.
  • The general term for a piece of data that signals the end of data (like a \0 terminator in C) is a sentinel value. Notice that this is an example of old-style in-band signalling, rather than new-style out-of-band signalling, like an extra attribute (size) or exception (StopIteraton).
    • In python, you can automatically convert an sentinel value to a StopIteration by passing it as the second argument to iter().

  • Partial function application is easy in Python. Just use functools.partial().

  • Every argument in a Python function can be either positional or keyword unless you explicitly indicate otherwise. To indicate the end of positional-only arguments, use the / argument. To indicate the beginning of keyword-only arguments, use the * argument.

  • Use namedtuples to improve readibility as an alternative to enum.Enum.

  • If you find yourself adding/removing items from the beginning of a list with del, .pop(0) or .insert(0, item), then you should be using a deque.

  • Use with contextlib.suppress(e): to temporarily suppress an exception e.

  • To terminate a generator, never do raise StopIteration. This won’t work anymore. Instead, simply use return.

  • “Frozen” (as in frozenset()) doesn’t really have a special meaning in Python aside from immutable.

  • Heaps offer a similar kind of partial ordering that you get in a tournament bracket.

  • In Python, names have no type, but do have scope. By contrast, values have a type but no scope (i.e. they live in the heap).
    • Note that Python is strongly typed, but typing applies only to values, not names.

  • deques are implemented as doubly linked lists, so they support popping from the ends in O(1), but indexing into the middle is O(n).

  • Use @dataclasses.dataclass if you want to define a struct-like class.
    • For simple structs that don’t need field names, just use tuples.
  • Python technically supports both constructors and destructors with .__init__() and .__del__(), but it’s more common to use context managers if you need to abstract away cleanup code.