[Python-3000] Draft pre-PEP: function annotations

[Phillip J. Eby]

At 07:04 AM 8/15/2006 -0700, Paul Prescod wrote:

>On 8/14/06, Guido van Rossum <<mailto:guido at python.org>guido at python.org> 
>wrote:
>>
>>Haven't I said that the whole time? I *thought* that Collin's PEP
>>steered clear from the topic too. At the same time, does this preclude
>>having some kind of "default" type notation in the standard library?
>
>The PEP steered TOO far of this topic. If it is total free-for-all then 
>when and if we do come up with a standard syntax (whether in 2006 or 2010) 
>it will conflict with deployed code that used the same syntax to mean 
>something different then the standard. And even if there is never, ever, 
>going to be a standard, it must be possible for tools reading the 
>annotations to know whether the user intended their markup to conform to 
>metadata-syntax 1, where "int" means "32 bit int" or metadata syntax 2 
>where it means "arbitrary sized int". Similarly, they must know whether 
>the annotater intended to use metadata syntax 1 where "tuple" means "fixed 
>size, heterogenous" or syntax 2 where it means "immutable list".

On the contrary - it is precisely this looseness that the PEP meant to 
specify, and that I support.  The alternative is too restrictive.

Meanwhile, the absence of predefined semantics does *not* preclude a 
default type notation existing in the standard library, any more than the 
absence of a predefined semantics for docstrings or function attributes 
prevents the stdlib from containing docstring processors or tools that 
operate on function attributes.