[asdf-devel] [PATCH] add &allow-other-keys to LOAD-SYSTEM, COMPILE-SYSTEM, TEST-SYSTEM

[Tobias C. Rittweiler]

Robert Goldman <rpgoldman at sift.info>
writes:

> On 2/24/10 Feb 24 -9:00 AM, Tobias C. Rittweiler wrote:
>> Robert Goldman writes:
>> 
>>> On 2/24/10 Feb 24 -5:54 AM, Tobias C. Rittweiler wrote:
>>>>
>>>> Although the way to extend operations by additional initargs is somewhat
>>>> cumbersome, it's possible. Unfortunately, the sugar forms LOAD-SYSTEM,
>>>> COMPILE-SYSTEM, and TEST-SYSTEM do not take additional initargs.
>>>
>>> I see that the docstrings for these functions seem to have some form of
>>> markup in them that I do not recognize ( [operate][] ).  Can you comment
>>> on this?
>> 
>> No it's what was in there.
>
> Hm.  Possibly it's some creeping Markdownization.  At any rate, unless
> we're going to adopt some docstring markup scheme for ASDF, which I
> don't favor (ASDF should not have external dependencies), we should
> replace all the "( [operate][] )" strings with simply "OPERATE".
>
> I have mixed feelings about &allow-other-keys.  In general, I hate it,
> because it makes the code no longer self-documenting.  Anyone's guess
> how to call the function.

I plan to add necessary magic to slime so valid keywords will be shown
in the minibuffer as part of arglist display.


> So here's a question for you, should we:
>
> 1.  Add &allow-other-keys for these wrapper functions or
>
> 2.  Figure that if you are supplying arbitrary keywords, the safety
> belts are off and you should be forced to use OPERATE or OOS.
>
> I'm OK with either.

Actually, &allow-other-keys would not be necessary if these weren't
normal functions but generic functions. Then methods can add valid
keywords. And there's a programmatic protocol to get at all valid
keywords.

I'd favor that approach.

  -T.