#django

OH, i guess that wasnt to me :D

my mistake

right, was just wondering what the convention is, I guess something like modelA/pk/modelB/pk

HowardwLo: Then you might as well use the default auth. But if you need registration pages and such, something like django-registration will help a lot.

but django-registration is no longer maintained/developer afaik

apollo13: Nah, ubernostrum stepped down in 2013.

developed*

joelones: Yeah, ideally you'd not use the PKs but some other unique and human grokable value (also, security concerns about incrementing ID leakage).

karanlyons: well, I authenticate using Email instead of username. trouble is, the form error says “username” and not “email”. I guess I need to work out my own login() and authenticate() ?

So something like /my-parent/my-child/.

HowardwLo: What version of Django are you using?

karanlyons: 1.5

So you made USERNAME_FIELD the email address, right?

karanlyons: actually, I have a custom user with Email field

django 1.6.5 hangs while attempting to render my model admin

no idea what to do :(

nszceta: do what your heart says

burn it all

I'm kidding, just paste the code and the traceback

no traceback!!!

o_O...

HowardwLo: Yeah, and that inherits from AbstractBaseUser?

nszceta: Start wolf fencing out apps until you figure out which one is causing problems, then wolf fence out models in that app.

I know what model is causing it

I can't figure out why

nszceta: If you can, post all the code up to dpaste and let us take a look.

HowardwLo: https://docs.djangoproject.com/en/1.5/topics/au...

Synthead: How's it going?

karanlyons: librabbitmq doesn't support python 3?

Ah, shit.

Nah, it doesn't, I don't think.

Will do sir. Be back in a few

I'm still on 2.7 because of other dependencies, so I wouldn't have noticed.

karanlyons: does that mean that my celery tasks and django would have to be python 2?

karanlyons: or can I run celery as python 2 with django on python 3?

That's a good question.

You *could* definitely do it, but there are some possible complications involved.

karanlyons: one sec, i’m digging through the code to see what the hell is going on

karanlyons: I imagine that celery has to be on python 2 to see librabbitmq on python 2, and since celery integrates tightly with django, django and everything else would need to be python 2 also

Synthead: I'm investigating to see if there isn't something we know to work that we can swap in for librabbitmq in Python 3.

karanlyons: I'm using python-amqp at the moment, but it seems like everyone's in favor of librabbitmq

Synthead: librabbitmq is faster for being C, but py-amqp is developed by ask (who develops celery) as well.

Try uninstalling librabbitmq, and any other Python rabbitmq packages, and just running py-amqp.

librabbitmq is, I think, the most performant Python client, but sadly has no Py3 support yet.

in multitable inheritance, an index_together declared in the parent model get's propagated to the childs (when the child doesn't itself declares a Meta) ...and a validation error arises stating the fields are not available in the child model

is this something someone else can replicate? am I doing something wrong?

im a little confused at my own frankenstein code. I seem to have email in forms but not in models….. but its saving the email properly

Kronuz: Can you show us the code? Cause it'd be helpful to see the inheritance flow.

HowardwLo: You shouldn't need forms at all if you're extending djang.contrib.auth's User model. Unless you're making custom registration pages or something.

karanlyons: what would you use instead of pks?

karanlyons: i might’ve. I cant remember what I did, but I had a problem with django-facebook and hacked the crap out of it to get it to work

and im sure it wasnt the right way :/

joelones: Ideally something else that's unique. If you don't have anything like that, then you *can* use PKs, they're just not as nice for human beings to use.

hallo. i’m creating a custom manager for one of my models. i have an __init__ method for the manager and am getting the following error: AttributeError: ‘MyManager’ object has no attribute '_inherited'

karanlyons: I'll set up a test here with the vanilla django

(I have some patches applied on django)

shit for all i know, i probably added the field using South -.-

joelones: Then there's an issue with auto incrementing IDs exposed to the public, which is that anyone can trivially loop through all your IDs, which is a bad attack surface to have.

joelones: So if you're going to expose PKs, you want them to be non continuous to some extent. Something like simpleflake is great because it's non continuous but also includes a seconds since epoch, so your new rows will be continuous in the database.

HowardwLo: Haha, oh man, I know that feeling of returning to a forgotten hack. Good luck!

what does suppress the backwards relations ? for related_name in ManyToManyField

w00tner: I'm afraid I don't understand the question. Could you rephrease it?

karanlyons: I think it's something else in my code (the source of the problem)

Kronuz: Got it, we're here if you need us.

karanlyons: is it difficult to move emails and hashed/salted passwords from one column to another? I think I stored the users under the wrong column :(

QUOTE: ManyToManyField.related_name Same as ForeignKey.related_name. f you have more than one ManyToManyField pointing to the same model and want to suppress the backwards relations, set each related_name to a unique value ending with '+'

HowardwLo: Not difficult, but not trivial.

when i would like to supress backwards relations. what does that mean and when i will be using it?

You'd want to create the new column with blank=True, loop through the rows and copy the data from the old column, remove the old column, and then remove the blank=True decleration.

I *think*, hard to say exactly without seeing your schema/code.

karanlyons: thank you! :)

karanlyons: found it!

must be a bug in django

w00tner: related_names have to be unique, because otherwise they'd clash.

karanlyons: https://dpaste.de/6Kqj

must be a bug when using abstract base classes and multitable inheritance

it only happens if Place has an abstract base class which declares index_together

(might also happen with other stuff in Meta)

w00tner: If you end the related_name with "+", Django will see that plus and won't create the backwards relation, which means that the destination of your relationship won't have a way to trace back to the source.

... I've tested so far with v1.6.6dev

karanlyons: my problem is, I was using django-facebook for the facebook login, but I seemed to have the regular email registration save into the user column for facebook....

thanks karanlyons

karanlyons: so now I have to flush out the email registration, and tie it in with django-facebook

Kronuz: Oh, I bet Django isn't properly hooking up the _meta to the class.

w00tner: No problem.

Hello everyone.

HowardwLo: so User.username has what should be User.email_address?

I'm trying to convince my company using Django can be a very good deal for their front-office.

probably since it sees it's an abstract, it let's it pass to the children (but it shouldn't since the children inherits from a concrete)

But, as a company, they need warranties and for myself I need to know some "real" limitations of Django in a business project.

karanlyons: for facebook, I’m not exactly sure what they use Username for, but under personal information: First name, Last name, Email, thats where I shoved the email

Do you have any clues?

so I’m authenticated via a profile? or something

Kronuz: Actually, I think the issue is your Place model isn't abstract. But I'm spitballing, I'm not testing anything.

karanlyons: gah, wish i knew more

Debnet: What are you trying to build? Is the question "is Django stable" (yes) or "is Django right for our project?" (maybe)?

I understand, precisions.

karanlyons: yes, it's not abstract, thus, the meta inherited by the children using multi-table inheritance shouldn't get those values in the parent (Place) model's Meta

...but django somehow gets confused (as the Meta is declared in the abstract parent)

HowardwLo: I'm just having trouble groking what your setup is like. So you're using FacebookModel with a OnetoOne to the User?

if I move the index_together to the concrete model, then it works