A couple tickets that should get some discussion

[James Bennett]

I just noticed a couple tickets posted tonight by Brian Harring which
look like they'll need some discussion before a clear decision can be
reached on them; since he's raised some good points, I thought I'd
point them out so we can get that discussion going:

http://code.djangoproject.com/ticket/3439 -- Improving the
dispatcher's performance (why *are* we still using such an old version
of PyDispatcher? Did it just get forgotten deep down in the code?)
http://code.djangoproject.com/ticket/3441 -- Improving template performance

--
"Bureaucrat Conrad, you are technically correct -- the best kind of correct."

[Jacob Kaplan-Moss]

On 2/5/07 10:39 PM, James Bennett wrote:
> http://code.djangoproject.com/ticket/3439 -- Improving the
> dispatcher's performance (why *are* we still using such an old version
> of PyDispatcher? Did it just get forgotten deep down in the code?)

I'm not entirely sure, actually...

I'd prefer to track pydispatcher instead of going out on our own, but I'm not
sure about the ramifications of that. Can someone investigate swapping in the
latest pydispatcher and compare it against Brian's patch?

> http://code.djangoproject.com/ticket/3441 -- Improving template performance

In [4461] -- thanks.

Jacob

[Brian Harring]

On Tue, Feb 06, 2007 at 08:05:16AM -0600, Jacob Kaplan-Moss wrote:
>
> On 2/5/07 10:39 PM, James Bennett wrote:
> > http://code.djangoproject.com/ticket/3439 -- Improving the
> > dispatcher's performance (why *are* we still using such an old version
> > of PyDispatcher? Did it just get forgotten deep down in the code?)
>
> I'm not entirely sure, actually...
>

> Can someone investigate swapping in the
> latest pydispatcher and compare it against Brian's patch?

Poked at pydispatcher 2.0; upshot, they grew some tests since 1.0.

Downside, the 2.0 release doesn't even build due to tests being
left out of the release- said issue has been in v2 since its release
(july), a bug was filed in august about it, and still no resolution.

Not too confident about upstream being responsive/handling issues, in
other words. Test coverage is pretty basic also- no assertion in terms of
callback ordering for example (so the f1/f2 question from the ticket
remains), no send coverage, etc.

The actual v1 to v2 changes, ignoring the broken release, are
inclusion of a basic test suite and whitespace changes- no real code
changes, thus the performance issues are still there.

> I'd prefer to track pydispatcher instead of going out on our own, but I'm not
> sure about the ramifications of that.

Personally, suspect y'all are already on your own from a maintenance
standpoint- as said, upstream looks to have shutdown. Either way,
updated the patch to correct a bug upstreams test suite spotted, and
folded upstreams limited test suite in.

Regardless of if the optimizations go in you probably want to lift the
test suite integration.

~harring