DjangoBot: #django-dev

hey ericflo

hey brosner

came across a bit of a bug in #6095

Good to know, what's going on?

validation.py breaks with self-referenced m2m and through

brosner: Ohh, good edge case, I don't think there's a test for self-referential m2m and through

i tried formulating a patch, but got side tracked ;)

brosner: I'm on it :)

great :)

ericflo: welcome back, floguy :)

C8E: Didn't realize I was gone!

lol

your last mail in django-dev was on 17th may :P

C8E: Ahh I see. Yeah, I've been living over in user land doing django apps and not peering into the internals at all.

ericflo: ah, sorry. i can't follow dj-users

C8E: Haven't been posting there either, lol. Just off in my own little world I guess. :)

brosner: What's the use case for that, anyway? Thinking about a model being m2m related to itself through itself going through itself makes my brain hurt a lot.

ericflo: you've been around the world, afaik

ericflo: it doesn't go through itself. my case is a Product model with a related_products m2m with self going through a RelatedProduct model

C8E: Ahhh, nope, I'm leaving in a few days to begin on that crazy crazy journey.

ericflo: to mochi-city, ya mean? ;)

C8E: Yep, and then Japan, then back to Minnesota, then back to mochi-city, lol. It's gonna be quite a ride.

ehehe

ericflo: ja luky man :P

brosner: I see what you're getting at. That makes more sense.

ericflo: do you have this in a git repo i can look at. id like a patch of the patch :P

brosner: I do, but it's local and I don't have a port punched in my firewall.

brosner: I'm such a git-noob. I use it like SVN lol.

fork my repo on github and push there

brosner: I'm gonna update the patch locally first, then figure out how to share it better.

ok cool

or if you really wanted just do a git diff between the commits and send me that ;)

that would work too

brosner: gah I got distracted, back to patching

:)

brosner; This is a bigger bug than I thought. Self referenced M2M doesn't work right now at all through a model. I've got some tests, now for the fix...

so it was more than just the validation.py code?

brosner: Yeah, definitely.

brosner: Well I got it working, pretty simple patch, but it brings up another bug.

brosner: Symmetric-ness

yeah that was my next thing ;)

brosner: Working on that now

brosner: Haha, do you have anything else I should be looking for? I'm not good at these edge cases, I guess.

in my m2m i have set symmetrical to False as well, but didn't get that far

ok

I'm using a case of Person and intermediary of Friendship, so I think this is actually not such an edge case, but something that many people will want to use.

yeah pownce would have a use case for no symmetrical relationships defined on the m2m if that was how they did it

brosner: Actually, symmetric-ness doesn't really make sense with intermediary models since it's only invoked by add()

brosner: Which no longer exists.

ah right

brosner: If it comes back, all the machinery still works, from what I can tell.

the use of an inline in nfa would make the relationship not symmetrical by default

OK

Would there still be a way to make that widget apply changes symmetrically?

I gotta learn more about these inlines.

hmm

basically an inline is a wrapper around a queryset.

in an inline formset it implies the fk relationship

Makes sense

brosner: I've got a git diff now.

awesome

Is there any way to have git make a diff that SVN plays nice with?

Alex_Gaynor: Does SVN even handle diffs?

Alex_Gaynor: I always use patch

ericflo: Bah, I meant some of the SVN tools like Trac

it does work in trac

ericflo: yay #6095 is now even better ;)

Hrm, first I thought it was patches that were too big, then I thought it was git diffs, what does cause Trac not to display patches :D

not sure

it may even be fixed for their next release

I didn't think of that, hope so, the new version also has the nice JS foldout things for folders in the borwser :D

OK uploaded the new patch to the ticket, gotta go for now though. Hopefully this works better for 'ya, brosner.

awesome man. thanks for your quick help and fixes :)

NP

Hi, guys. I trying to revive the MS SQL backend (as in make it really work and update it to qs-rf) and one of the things needed is it is really an external backend (i.e. without need to patch Django itself) is, when working around some characteristics of the RDBMS, not use if settings.DATABASE_BACKEND == 'mssql' (or 'mysql) buy rely on some more generic backend property...

do you think a path like the one in th patch that would be worthwile pursuing?

In the case of the MS SQL it has one shortcomming in common with Oracle and another with MySQL, so that's why i've decided to switch these "if"s to test for the values of (newly-added) connection.features attributes

cramm: agreed. Those types of special cases shouldn't be done that way if you want it pluggable.

Although you could override method itself.

empty: that's precisely what I'm doing for special cases that just involve this new backend. But i've decided to generalice the cases where there are two backends to special case

cramm: I think that's good.

empty: i've seen django-sqlalchemy is being updated again :)

cramm: yeah I've been steady on it. I just do some things in local branches when they might be disruptive. I've got to the hard places now. :)

empty: cool

cramm: Are you going to be working with the current db backend projects for MS(I think there are 2 of them(

cramm: it seems to me that a mssql backend would be fairly easy.

but then again I haven't dealt with the backend code all that much.

cramm: are you targeting 2005 or 2000

I'm afraid to use overriding internal django methods that are big from the external backend because it would mean that the external backend would need to track Django development