My ELI5 for this has always been that static vs dynamic concerns variables; strong vs weak concerns objects.

- Static + Weak = C

- Static + Strong = Haskell

- Dynamic + Strong = Python

- Dynamic + Weak = JS

> strong vs weak concerns objects

and, in particular, explicit vs. implicit type casts of those objects.

But C and Haskell don't really have objects...

It is strongly typed in the sense that 3 + "5" is an error. It also has duck typing which is weak. So, which is it? I would say it is more weak than strong.

There are two axes: strong/weak and dynamic/static.

Examples:

Python: strong, dynamic

PHP: weak, dynamic

C: weak, static

Haskell: strong, static

Duck-typing doesn't imply weak typing, but rather dynamic typing. Can you explain why you consider dock-typing as being weak-typing?

I once had someone try to tell me that python was "untyped" - I guess people assume a lot when they come from a first-class statically-typed language.

I wonder if my professor of the Programming Languages course at university would laugh or cry at such a definition.

I expect that everybody doing this job has a CS or equivalent degree and knows the difference between strong / weak and static / dynamic typing. Reading other comments here demonstrates this is not the case. This is not a bad thing (luckily we only need a screen and a keyboard) but hopefully they'll have learned something new today.

Btw, there is a lot I don't know despite my degree. No hard feelings.

“Untyped” is the phrase used in literature too

https://www.sciencedirect.com/science/article/pii/S147466701...

Practically, untyped languages would include asm and forth.

In Python, objects have types. Variables are best thought of as void* pointers to objects.

I've found over the years that people have an easier time understanding how Python manages variables when you tell them that, rather than "in-memory boxes with data" as is the case in other languages, Python variables are more like "name tags that you attach to an object".

Pointers that also carry the underlying type information.

I don’t remember what the CPython implementation looks like, but even if the information is there, it’s not really exposed to end users. “type(foo)” doesn’t return the type of the name “foo”, but of the object that “foo” points to.

I meant that it's a pointer that is aware of the type of object it's pointing to by virtue of all objects knowing their own type. Same difference I guess.

Unlike void pointers or Object references in Java you can't "forget" an object's type.

Ah, I think I see what you mean. I still think it's more accurate to say that names are just pointers that don't have any typing info themselves (or if they do, it's an implementation detail that doesn't actually get surfaced and isn't a part of the language definition). In Python it doesn't quite make sense to talk about the type of a "name" or "variable" or "pointer" because, because it points to an object, and that's where the canonical type info actually lives.

iirc untyped is not incorrect, untyped lambda calculus for instance is where your value type is unknown until your runtime like dynamic type, untyped doesn't mean you don't have type in your language

There's no such thing as "dynamic types". A type is by definition associated with an expression in a language, but those expressions are part of the program's static source code. Python has some kind of checking of the runtime values involved in certain operations, but that's not typing.

Types are not associated with expressions in Python: evaluating an expression produces a typed value, whose type, and not only its value, depends on the type and/or values of the inputs. Example:

   def comparativeInsult(a,b):
      if b>a:
         return "try again"
      if a>b:
         return a/b
      else:
         return None

The type is, for all practical purposes, a part of the value of a Python object.

Such a thing is not a type. Not by the usual definition, which predates not only python but mechanical computation.