[Grok-dev] Re: MultiAdapter doc bug?

[Brandon Craig Rhodes]

Martijn Faassen <faassen at startifact.com> writes:

> Tim Terlegård wrote:
>> On Jun 24, 2008, at 11:01 AM, Philipp von Weitershausen wrote:
>>> Tim Terlegård wrote:
>>>>
>>>>    >>> home = IHome((cave, fireplace))
>
> Brandon has been proposing it ... The proposal was changed to
> something different-looking in that discussion, though.

The change was because people pointed out that there do exist adapters,
in some people's applications at least, which adapt <type tuple> to
something else.  When you say,

  home = IHome((cave, fireplace))

you are, strictly speaking, asking that a tuple be adapted to the IHome
interface, and everyone agreed that this is sensible and should be
allowed; therefore, the tuple cannot be used to say "multi-adaptation!".

Instead, we negotiated that a default value would *have* to be supplied
with a "default=" keyword argument, *not* with a second argument, like:

  home = IHome(cave, default=None)

Once the default is moved to a keyword argument instead of an optional
second argument, multi-adaptation can be done like:

  home = IHome(cave, fireplace)

or:

  home = IHome(cave, fireplace, default=None)

Obviously this would break nearly the entire Zope code base, so the
proposal was to write an .adapt() method on Interface that takes the
new-style arguments for people who wanted to use them; and maybe then
rewrite everything to use adapt, then redefine __call__ to work the same
way, then remove .adapt() or make it the same as __call__.

And, as Martijn says, the first step is for me to sit down and write a
branch of grok.interface or grok.component (or both?) that defines the
new .adapt() method.  The problem last year was that, even after looking
a bit at the code, I didn't really understand it well enough to add the
extension.  I'll look again and see if things make more sense.

I remember that there are single adapters; multi adapters; for both,
there are named adapters in addition to the usual anonymous ones; each
kind of adaptation should be able to accept a default= argument; and
... there was *something* else that I didn't understand at all, and it's
what got in my way, and it's what kind of stopped me cold.  Was it
utilities and whether they can be name or not?  I can't remember.

-- 
Brandon Craig Rhodes   brandon at rhodesmill.org   http://rhodesmill.org/brandon