Re: Refactor get/set_property to take the item as first argument (issue 573670043 by address@hidden)

[Han-Wen Nienhuys]

On Sat, Apr 11, 2020 at 2:45 PM Han-Wen Nienhuys <address@hidden> wrote:
> > What you call a "giant rewrite" is basically this patch.  It's a purely
> > syntactical patch only extending rather than reducing possible
> > approaches and perfectly compatible with different implementations.  The
> > work for that patch is already done, the impact is similar to what
> > happened when changing around the syntax of unsmob a few times ian the
> > past.  It will allow for prolonged work on possibly different
> > implementations without having to constantly deal with merge conflicts.
> >
> > It does not depend on picking a particular implementation.  You
> > expressed interest in what I was planning to do with the opportunity
> > this patch opens, I answered.  I do not, however, see the point in
> > discussing and dissing (partially) unwritten code based on an incomplete
> > understanding in a handwaving discussion.
>
> I think it is not out of the ordinary to discuss plans invasive plans
> before implementing them.  I think this change is invasive because it
> changes the idiom across the source code, and I disagree that this
> change is net neutral. C++ method calls are a more natural idiom plain
> function calls. If we can get a significant performance boost, that
> may be worth it, but I think we should first understand if that is the
> case.  As said (see above), I am skeptical.

Also, property access accounts for about 10% of runtime, last time I
looked. This can probably be reduced by some factor, but even if it
were infinitely fast, it is not going to make a huge difference to the
overall runtime.

-- 
Han-Wen Nienhuys - address@hidden - http://www.xs4all.nl/~hanwen