DjangoBot: #django

sounds like you forgot to set your pythonpath properly

defrex: http://www.b-list.org/weblog/2007/sep/22/standa...

thebitguru: i was reading the spec and it seems there's a deprecated many-queries functionality in execute(), i was wondering if that might be interfering

thebitguru: thanks

but it seems unlikely if your .queries outptu is correct

s/outptu/output/;

defrex: I have written my own version which assumes that your scripts are living inside the project. http://www.pastethat.com/Si3cO

Xore: yeah, the behavior is really weird

thebitguru: cool, thanks

Magus- got this, works, but i tried to put a - in front of pub_date to get it to order the other way and it didn't work. {% for news in object.news_set.all|slice:":10"|dictsort:"pub_date" %}

got the permalink decorator to work, but what confuses me is how is it supposed to be reusable if takes a absolute name of a view?

RaceKondition: because a view name doesn't change

the url for that view might

ameriblog1: you can solve the date sorting also by editing your default sort order for news

Magus- got it, never mind, i apolizie

Magus-: so.. the reusability of the app is determined by whether it is embedded in a project or a standalone app? because project app's view names might change..

Generally one doesn't put stuff intended to be reused within a project.

yeah, true

indeed, so the point is moot

can i compare the result of a filter with ifequal? it doesn't seem to be working.

e.g., {% ifequal types|index:forloop.counter0 "number" %}

can you get access to the logged in user in the template without passing it in from the view? Is there a special way to get to it?

"{{ types|index:forloop.counter0 }}" prints "number" exactly

So, if you're developing a reusable app, do you just move it outside of hte test project?

isleshocky77: no, though you can sort of automate the passing

isleshocky77: use RequestContext and context processors can supply data to templates

tobias: no, ifequal does not currently support filters - known issue, will be fixed :)

Magus-: cool, thanks.

if you don't mind applying patches you can fix it now

#5756

ah workaround looks nice

i know apache/mod_python is the preferred setup according to the docs. i wanted to know if others are using lighttpd/fastcgi in production for django

I'm sure people are

others use apache/mod_wsgi

irk

Viva mod_wsgi

actually the better method is nginx/mod_wsgi ;)

I'm really tempted to try WSGI. it looks hot.

belred: I'm using fastcgi with flup with good success. But I'm living in the shared hosting ghetto ;-)

FastCGI with Apache, mind you.

thunderbolt: how silly

wolfe: Which part?

geez... any two people here with the same setup???

shared hosting, you should pay me to set you up with a VPS ;)

wolfe: Ooh, does your business have a website with prices?

not up yet. corporeal has a few VPSs here with me that is the only customer which isn't in town :)

Ah

i can set up a production system anyway i want to. i was hoping there was a clear answer which was the best for speed and robustness

thunderbolt: if you do go for a VPS though, use a Xen provider

belred: they all have tradeoffs and advantages...

you'll find places which use UML or Virtuozzo will screw clients over

Thanks for the advice.

belred: i'm using apache/mod_python right now, but the fact that I have to restart apache when I update a django app just sucks. I'm looking for alternatives too.

people who use virtuozzo don't limit how many people are on a server, just liek shared hosting. With Xen, it is more static to offer a allocated amount of memory

tobias: you'll always have to restart the code no matter what you use

axiak: yes, but some options let you restart just the app, and, furthermore, don't break incoming requests for a few seconds :)

tobias: what? no

wolfe: I wish I remembered which VM they used at work, *smacks head*, I thought it was from Novell...

VMware.

WSGI being one of those

tobias: well it's not like you should be hosting anything else from that apache/mod_python service :)

Be back later, my meal has dinged out of the microwave.

you don't realize how apache works then nor any other daemon. daemons finish the request before ending the daemon

wolfe: what plans are you offering for VPS's?

what is the downside to using lighttpd/fastcgi for dynamic content?

axiak: I should run a separate copy of apache for every django app?

tobias: you don't have much of a clue on how to cycle http daemons

wolfe: how should i be doing it?

tobias: you set it to port 2344 on 127.1, if you need to restart gracefully, set in the config to port 2345

tobias: well it depends...at least a large segment of your code, yes

if you HUP nginx, it will gracefully restart, same with apache

it will finish the requests BEFORE shutting down the old config

it is up to the admin to update the application correctly

wolfe: will port 80 always be open?

tobias: I mean mod_python really starts to suck when you have many projects...but for one it's not like you'd be serving more than one code base anyway...

tobias: yes, it still accepts requests

axiak: but i am :)

wolfe: i'll try that, instead of a hard restart. thanks.

tobias: hence one of the tradeoffs of mod_python... :)

tobias: but the tradeoff is certainly _not_ that you have to restart your code :)

greenskeleton: I do yes. shared hosting too, but I only have my own web customers in the shared VPS :)

which is what has been keeping me from finishing dancingroo

tobias: it's that it doesn't allow you to break down projects into separate processes...

is there a speed difference for apache/mod_python and lighttpd/fastcgi for dynamic content? i head lighttpd is faster for static content

belred: there is a slight overhead for having multiple daemons using fastcgi/wsgi...but it's heavily amortized

belred: I think you'll find mod_python uses more memory but does better with higher concurrency and is single threaded

while you can try things like using multithreaded wsgi to get high concurrency with lower memory otherwise

memory won't be a problem for me... i'll just put 8+ MB ram on this machine

belred: what performance are you looking for?

i think 100+ req/sec is reasonable

i'll only put one project on it

belred: what concurrency?

you mean number of simultaneous users?

sure

belred: like, when you use ab -c XXX -n YYY

i don't remember seeing those numbers... i'm sure it's in the 1000's