#django

houman, then format the date correctly..

considering i18n, i would have to pick date.year, date.month etc myself and form a string like that myself right?

jrief_, I would do that as 2 queries.

jrwren, I woudl do that as 2 queries..

yea.. right person that time. :)

how would I do it as 2? its currently a nasty select n+1.

Hi, I have a url /report/html which renders nice report, but when I try to save this, it defaults to 'html', can I override this?

afaik, i cannot do it as 2 because that inner select in teh where clause references the outer t1.id

Z_A_M: what does this line do: p = get_object_or_404(PN, pk=policy_id)

activus1: It looks for a 'PN

drp

it looks for a 'PN' object with a primary key of = policy_id

if it doesn't exist it throws a 404 page

Z_A_M: thanks

activus1: yup! A lot of that stuff is in the django docs btw.

jrwren, ...

max_timestamp_id = t1.objects.order_by('-timestamp')[0].id

results = t1.objects.select_related('t2').filter(id=max_timestamp_id)

PKKid: that is not the same.

how i that not the same?

you eman if more than 1 object has the max timestamp?

mean*

You could do this if you expect more than 1 object to share the max_timestamp

max_timestamp = t1.objects.order_by('-timestamp')[0].timestamp

results = t1.objects.select_related('t2').filter(timestamp=max_timestamp)

is it possible to add/remove modelAdmin's inlines ?

I want to exclde some inlines if they are unnessecary (already filled in)

there isn't a single max_timestamp_id, there is a max for each t1

jrwren, I think I understand.. 2 sec

t1 is one to many t2.

jrwren, How many objects do you expect t1, t2 to have?

~5k

and then anywhere from 1 to dozens on teh many size to t2

is there anything I can do to automatically fill in a field of a model from models.py? I know I can do default='Whatever', but that doesn't actually fill the field in

jrwren, Also, I believe the ORM cannot do that query, at least not without storing a list of max_ids and using the __in filter.. which is a bit icky.

i'm thinking about .extra(where=['???timestamp=(select max(t3.timestampe) from ??? as t3 where t3.???= ??? )']

but I don't know how I can fill in the table and field names from orm in the extra where

i guess I can just hardcode, but I'd need to know the table alias name used if any... I guess django.db.connection.queries shows me there is no such alias

Z_A_M: how do I return just plain text to the screen so I can see what was passed in ?

jrwren, I wouldn't dig too deep into the ORM internals to write this. A raw query seems much more readable/maintainable imo.

return HttpResponse('POST')

got it

PKKid: thank you for your help

jrwren, Also note: this seems like a slow query either way. It might be worth flagging the max_timestamp rows when new rows are added. That would make this query blazing fast, easier to read (at the cost of slower writes to the DB)

the select n+1 is insanely slow.

the subquery is very fast.

if the rows were flagged, then you could do something simple like "SELECT * FROM t1 WHERE t1.max_flag"

it also seems like this could be re-written using GROUP BY. Unless im missing something..

i thought group by at first too, but couldn't figure how to write it, so I ended up with this WHERE subquery

good point on the max_flag, we'd have to maintain it, but should simplify things.

jrwren, I think this is a similar query, using GROUP BY and HAVING: https://gist.github.com/3540307

filer=t1, auser=t2 in your example

jrwren, And here is a blogpost showing that the ORM somewhat supports HAVING: http://eflorenzano.com/blog/2008/11/06/secrets-...

close, but no, that query isn't the same.

if it were groupby,I'd only want to groupby t2.id (kinda like turning that one to many into a one to one based on max(timestamp) )

but if I groupby t2.id only, sql errors that t1.id must appear in group by since it was in the aggregate

maybe a group by and a join

yes.

i must have first translated that gist incorrectly. I think I see now.

I should be able to follow that post and get it to work.

hrm, or not.

the group by on t1 works, but once I join t2 it does not.

hmm

im curious now.. :)

Starting with django 1.4, I am getting an issue where I am getting apache processes hanging forever. They seem to be stuck trying to get a lock in the _remove_receiver method (from the mod_wsgi stack traces). Has anyone come across anything similar to this? We can reproduce the issue in 1.4 but not using 1.3. http://pastebin.com/iPe0HjRv

jrwren, Still here..

?

PKKid: I'm going with extra(where=...

and the subquery.

i couldn't get a groupby having working.

i could probably do it with 2 queries, the first being groupby and then the second being the old way with an additional IN but that IN value would have the 5000 values, not what I want.

jrwren, is it working?

looks like it.

i still need to profile the big picture, but this piece is great.

jrwren, Yea, that IN value thing would suck. :)