#django

also with postgres you can stick arrays in your data models (dont know if thats a good thing)

Codd would cry

I use arrays when it makes sense to.

If there will never be more than a known number of choices, it isn't such a bad idea.

For instance: to store something that may refer to between 0-7 weekdays, I store the ISO weekday values as an array.

I used to make that assumption about gender, but apparently I was wrong :)

Remember: "It's the word of Codd, not the word of God."

#controversy

Anything that needs to refer to a different set of possibly changeable data can't be stored as an array: at least, not without breaking referential integrity.

Technically anything that is stored as an array, mutable or not, breaks 4NF

But as FunkyBob said...

Indeed: but practicality beats purity.

Agreed

hi all, i want to have a model with a field that is based on the primary key, which is just the primary key prepended by a bit of text. i.e. PREFIX99 for the 99th entry in the database. I've tried to override the save method, but the primary key doesn't exist yet, so I think it needs to happen post save but I'm not sure how to do it. Grateful for any assistance.

thmcmahon: make it a property

mcspud: then again, technically datetime fields break 1nf (or is it 2?)

thmcmahon - define it as a property on the model so it is computed when it is inspected

Koterpillar mcspud thanks guys, wasn't sure what it was called, I'll have a look at properties

thmcmahon - https://dpaste.de/YPP2

mcspud: legend, thanks so much

FunkyBob - They break 4NF technically

depends

on?

you can view them as UNIX times

mcspud: works perfectly, thanks.

That doesn

't mean anything, its just a format

4NF says any shared attribute needs to be split

So even "First Name" can technically break it if there are 2 people named "Peter"

how is that different from an integer field?

But again, practicality

ah, yes, ok, that one

Koterpillar - its not

In practice any reasonable finite set of choices that can be changed over time should be split into their own table, but `reasonable` is very subjective

hmm

the "set of choices" bit is one I'll pick on

Also, if you are going to be _ordering_ by one of these "choices" (i.e., a timestamp, or even a name), then putting that value into a seperate table is probably going to be counter-productive.

Yeah, it's not strictly a limited set of choices.

time is continuous, therefore the set is infinite

I mean, there are an unlimited number of ... yeah, that.

Koterpillar - any DB system will have the deepest level of granularity it can go do though

so at 1ms increments, in the purest sense of it you should do that via FK relationships to a "Time" table

I feel like building a DB with unbounded precision floats, that has AWS keys, automatically orders itself new hardware and blocks calls until it's fulfilled

in fact, scratch that

dont do that

let's fork pg to report virtual tables like "datetimes", "floats" and such, and making float fields seem like FKs to that table

SELECT * FROM floats; will still work, obviously

hahaha

We are going down a deep, dark hole here

Koterpillar - I feel like Joe Celko has written something about this before :P

note that this hole is bounded

happy to watch, not getting involved in this one, as fun as it sounds.

He probably has a valid usecase for it

haha

Off the top of anyones head, does anyone know if, when calculating an income stream for an income/expense report to what period you assume it is valid? I'm guessing you annualise it, so 100/month income from child support becomes $1200 etc

Hi hi hi. I am having trouble with the model interface. Imagine three model tables that together represent an Exam, some Exam Questions, and a student's Exam Question Answer to the exam questions. I see how I can start with exam questions answers and follow related links to Exam and Exam Questions. I want to use the database state to know what exam questions are not yet answered. That is I want to

construct a to-do list from db state. So, Exam has a incoming fk from Questions. A Question has incoming fk from Answer. (There are other things on answers that make it unique to a student.) I want to subtract Answers from Questions to know what's unanswered. How can I do this nicely with django models?

I can express it in SQL. I want to be django-y.

.filter(something__something__answer__pk__isnull=True)

Ooo.

I'm trying to think of what has an 'answer'. Answer points to a Question. Question could have a related_name link back to via "answers" (eg).

exam.questions.filter(answer__pk__isnull=True)

?

hello

I'm at https://docs.djangoproject.com/en/1.10/intro/tu...

On Mac os

I am getting a 404 instead of a welcome screen, yet I see no error in the console, what else could it be?

can you show the output of the console from the moment you runserver to the 404? https://dpaste.de/

jennifer: a 404 where?

Yes one second

how do i get a @property from a thumbs.values("title”, ‘???’) ?

jennifer: and you're trying to access http://127.0.0.1:8000/ in your web browser?

yes

On safari

tchalla, you'd need to use .only instead

whoa whoa whoa

It works on chrome but not safari, why is htis?

that's weird

$ python manage.py runserver 8005

try that, then access at port 8005

It is weird indeed, I dunno why

Why 8005? One sec

Because it's new.

What's new?

just a different port, just incase something else is interfering

Getting you to behave differently might lead to the solution.

Says can;'t connect to server

you might want to consider turning off ipv6 in network connections

or /etc/hosts

Wat? No.

No

I won't do that because I write my own web server using socket and it works

Using 8000 and 8080

Django should emit the address and port to connect to. Even on v6-only systems.

So I know it's likely not related to ipv6

I'm also on a Mac, I didn't make any major / serious change to the system, it's a new laptop that's supposed to "just work"

"know" "likely" :P

Well I've been surprised before

But then this would be one of those "this is just stupid" surprise

Like a game of counterstrike interfering with somebody's ability to open up word

I've seen that and it's just dumb, absolutely dumb

jennifer, probably best to just disable ipv6, just to be safe. I've seen some absolutely crazy behaviour from osx before

:(

cybo_: values with .only or just .only?

one of our guys' cookies wouldn't stick. all of us tried to crack it

Or I can just develop on chrome...

Fuck okayu

tchalla, just .only