Python’s always been a cool language for many reasons: it uses a lot of simple, English-language keywords, it cuts down on many of the braces and semicolons so prevelant in other programming languages, and it’s cross-platform. Python code is beautiful. But besides these obvious perks, Python syntax and semantics go far deeper. In this post, I want to highlight 3 aspects of the language: functions, object definitions, and runtime introspection.

Function signatures are highly expressive

You’re probabbly familiar with ordered arguments:

def foo(a, b, c):
  print(f'a={a}, b={b}, c={c}')

foo(1, 2, 3) # a=1, b=2, c=3

But have you ever passed in arguments by keyword?

def foo(a, b, c):
  print(f'a={a}, b={b}, c={c}')

foo(1, 2, c=3) # a=1, b=2, c=3
foo(1, c=3, b=2) # a=1, b=2, c=3

You can also define a function with an arbitrary number of positional arguments like so:

def foo(*args):
  print(f'a={args[0]}, b={args[1]}, c={args[2]}')

foo(1, 2, 3) # a=1, b=2, c=3

And you can do the same for keyword arguments:

def foo(**kwargs):
  print(f'a={kwargs["a"]}, b={kwargs["b"]}, c={kwargs["c"]}')

foo(a=1, b=2, c=3) # a=1, b=2, c=3
foo(a=1, c=3, b=2, d=4) # d is ignored; a=1, b=2, c=3

You can even declare position-only arguments by terminating those args with a slash /:

def foo(a, b, /, c):
  print(f'a={a}, b={b}, c={c}')

foo(1, 2, 3) # a=1, b=2, c=3
foo(1, 2, c=3) # a=1, b=2, c=3
foo(1, b=2, c=3) # error; b is a position-only argument

And you can do the same for keyword-only arguments by prefixing them in the argument list with an asterisk *:

def foo(a, b, *, c):
  print(f'a={a}, b={b}, c={c}')

foo(1, 2, c=3) # a=1, b=2, c=3
foo(1, b=2, c=3) # a=1, b=2, c=3
foo(1, 2, 3) # error; c is a keyword-only argument

These features help you as a developer make the intent of your API more obvious by constraining how a function can be called. Python also offers the ability to annotate arguments with types, replace missing arguments with default values, and include a special triple-quote string at the start of your function for clear documentation.

Everything is an object

Unlike C++ or Java, Python treats all semantic-level constructs the same; functions, classes, types, instances – they’re all object’s from the interpretter’s point of view. All objects have a name, a type, and a value. So when we write something like:

a=1

class B:
  pass

def c(): 
  return a

we’ve just initialized three objects: a which is an int with a value of 1; B which is a type object with a value of the B class’s contents; and c which is a function with a value containing the function code, its signature, and some other data. We can verify this by running:

for var in [a, B, c]:
  print(type(var))

which outputs

<class 'int'>
<class 'type'>
<class 'function'>

Since classes and functions are just plain-old object’s, that means you aren’t limited by their block-style definition syntax. While certainly convenient, you can dynamically build your own types and functions.

Let’s build a class D using the three objects above:

D = type('D', (B,), dict(v1=a, v2=B, v3=c))
d = D()

This is the same as:

class D(B):
  v1=a
  v2=B
  v3=c

but it gives us the flexibility to programmatically define what D should look like.

Python is sometimes like JavaScript, in that you can often add properties on the fly. (I.E.: myobj.new_attribute = value). In cases where the interpretter allows you to do this, it’s usually because that object has a __dict__ attribute defined. This attribute contains all the object’s regular attributes and methods as well as hidden introspective variables (next section). You can override the getattr and setattr to methods to take managing the object’s __dict__ in your own hands. Finally, the property(f_get, f_set, f_del) method lets you define properties on objects with a custom getter (f_get), setter (f_set), and delete function (f_del). This function is available as a higher order function so you can write @property on the line above an instance method to make convert it into a propery accessor.

Runtime introspection is super simple

Runtime introspection is the ability to look at the state of the programming runtime within the code that is being run. In simple cases, you might want to find out how deep you are in the stacktrace to avoid overflows. In complex cases, you could be allocating memory by hand, declaring new types, or changing the code being executed! First, some basics:

Python code is run in a two-stage processes: First, the source code file is opened, and Python compiles it into a code object. Then the interpretter begins executing the code at its top-level scope. Whenever you’re just interpretting, compilation happens on the fly instead of ahead of time. This may sound complex, but conveniently, the code is a Python object type. That means we can introspect the code inside itself!

code is much more common that you may realize. Every Python function contains a __code__ property. Anytime you run a parameter-less function, you practically get identical behavior as if you had manually evaluated the function’s code using Python’s eval builtin:

def foo():
  print(f'a={a}, b={b}, c={c}')

a=1; b=2; c=3
foo()
# same as
eval(foo.__code__, globals(), dict(a=1, b=2, c=3))

Pretty cool!

Conclusion

I’m sure you can now see why so many computer science students use Python. It has awsome theoretical and practical power! Is there a feature of the Python programming language that you like? Please share it in the discussion.