PEP 526 - var annotations and the spirit of python

[Christian Gollwitzer]

Am 05.07.18 um 12:04 schrieb Steven D'Aprano:
> On Thu, 05 Jul 2018 09:17:20 +0200, Christian Gollwitzer wrote:
> 
>> Am 04.07.18 um 17:31 schrieb Steven D'Aprano:
>>> On Wed, 04 Jul 2018 13:48:26 +0100, Bart wrote:
>>>
>>>> Presumably one type hint applies for the whole scope of the variable,
>>>> not just the one assignment.
>>>
>>> You know how in C you can write
>>>
>>>       int x = 1;  # the type applies for just this one assignment x =
>>>       2.5;    # perfectly legal, right?
>>>
>>>
>> Not sure what point you are trying to make, but your example compiles in
>> C, if you replace the '#' comment sign with '//'.
> 
> 
> Oops.
> 
> 
> But... it compiles? Seriously?
> 
>> Only it doesn't do
>> what you might think: the 2.5 is down-converted to an integer, therefore
>> x will be 2 in the end. There will be a compiler warning but no error.
> 
> Sometimes I wonder how C programmers manage to write a bug-free "Hello
> World" program. No wonder it is described as a type-unsafe language or a
> weakly-typed language.

Even this compiles:

#include <stdio.h>
int main() {
	int x=1;
	x="This is serious";
	printf("%d\n", x);
	return 0;
}
Apfelkiste:Tests chris$ gcc intx.c && ./a.out
intx.c:4:3: warning: incompatible pointer to integer conversion assigning to
       'int' from 'char [16]' [-Wint-conversion]
         x="This is serious";
          ^~~~~~~~~~~~~~~~~~
1 warning generated.
15294370


Assignment in C to an integer will only fail when the source is a 
struct. Everything else can be "converted" to an integer by the compiler.

> I understand upcasting ints to floats, that's cool (even if a few
> languages take a hard line on that too, I don't). I understand Python's
> dynamic typing approach. I don't understand C requiring type
> declarations, then down-casting floats to integers.

Without prototypes, it would be worse; the compiler would put the 
bit-pattern of a float onto the stack and reinterpret that as an 
integer. With prototypes there is a "sensible" result, namely the 
integer part of the float. At least it doesn't crash ("2.5" in C is a 
double constant, which is usually 64 bit, whereas int is usually 32 bit 
only).
> 
> At least it shows a warning. But hell, who pays attention to C compiler
> warnings? There's *so many of them*.

In programs that are maintained, the warnings are typically taken 
seriously. Some warnings can be annoying, e.g. warnings from generated 
code about unused variables and such, but if you see a flood of warnings 
in regular code, that is a sign that the code has bad quality or wasn't 
ever tested on the platform you try it to compile.

> 
> (And I bet that by default the warning is disabled, amirite?)

Some people develop with -Werror (treat warnings as error and abort).

	Christian