[Grok-dev] Re: Global utilities registered too early?

Jan-Wijbrand Kolman janwijbrand at gmail.com
Tue Jun 19 08:53:47 EDT 2007


On 6/19/07, Martijn Faassen <faassen at startifact.com> wrote:
> Hey,
>
> Jan-Wijbrand Kolman wrote:
> > FYI: I put a print statement in registerUtility() in z.c.registry and
> > indeed it seems utilities that are registered thru grok are registered
> > *before any other* utility that gets registered in core zope packages
> > and "third party non-grok" pacakges. HTH.
>
> This is because Grok doesn't use actions yet, like ZCML does. Reading
> ZCML generates a whole bunch of actions, and only at the end of all the
> ZCML reading are all actions resolved (things may fail here with
> conflicts) and then executed.
>
> The grok directive triggers the grokkers. The grokkers directly register
> everything with the component architecture (including the global utilities).
>
> We need to change this. The path is reasonably clear, but I haven't
> found the time yet to work on this, and likely won't before EuroPython
> at the least:
>
> * change grok core so it uses Martian. This is not strictly necessary to
> do actions, but since we're going to touch all grokkers anyway, better
> get it over with first.
>
> * change all grokkers to generate actions. This will likely make us grow
> new infrastructure to do so. The actions should have discriminators
> compatible with those created by ZCML, so that conflicts work the ZCML
> way. We also should make sure overrides work the ZCML way (I think that
> already works if we do this, but we need to test it and analyze the
> consequences). Our actions can be home-grown; there is no need to reuse
> ZCML's actions, just to use discriminator tuples that will conflict
> appropriately with those from ZCML.

OK, thanks. For now what I will do is to register this utility thru
zcml after all.

-- 
Jan-Wijbrand Kolman


More information about the Grok-dev mailing list