pmxbot IRC Log Viewer

[14:35:28] <aclark> morning

[14:35:39] <MattBowen> aclark: !

[14:36:51] <aclark> MattBowen!

[14:37:54] <MattBowen> aclark: neat -- what were you doing with it?

[14:38:03] <MattBowen> aclark: i used it for a pubsub system backed by scala, but just a toy

[14:38:52] <aclark> using it on vanity.aclark.net to store entries

[15:01:50] <hazmat> i did some queuing work with redis a few weeks ago similiar to resque, its nice.. but i think i'm going to switch out for mongodb (findandmodify for atomic queue ops)

[15:02:07] <hazmat> i'm curious about the pubsub functionality of redis though

[16:08:53] <aclark> anyone want to critique this change for webscale! https://github.com/ACLARKNET/vanity_aclark_net/commit/b8aa113ad5d175d8482fd279939007d2eec87b84

[16:10:40] <aclark> kennethreitz: still in sf?

[16:14:30] <kennethreitz> aclark: got back yesterday :)

[16:17:16] <aclark> kennethreitz: ah! any interest in fixing vaincheese? people are opening tickets on vanity.aclark.net re: case sensitivity (not important at all of course :-))

[16:17:31] <kennethreitz> haha, nice :)

[16:17:41] <kennethreitz> Remind me later today :)

[16:17:48] <aclark> k

[16:18:15] <aclark> kennethreitz: i noticed you hadn't committed the change to version 1.1.2 in the code, FWIW

[16:18:29] <kennethreitz> local version

[16:18:36] <kennethreitz> er, didn't push it to github

[16:18:48] <aclark> kennethreitz: next i'm going to add ALL the packages on pypi somehow, so i can pick a random one and get it's dl count

[16:19:00] <aclark> only way i know is to look at /simple, unfortunately

[16:19:17] <j00bar> aclark: how much traffic did steve holden send your way?

[16:19:24] <kennethreitz> only have 5mb of cache on it i think

[16:19:36] <aclark> j00bar: hah, haven't looked yet… only made like 50 cents on adsense though ;-)

[16:19:54] <aclark> j00bar: it's hilarious he is opening tickets on it

[16:19:54] <j00bar> aclark: what does your filter() fn do?

[16:20:00] <j00bar> in your changeset?

[16:20:15] <aclark> j00bar: takes all the redis entries from lrange and returns the first uniq 20 (i hope)

[16:20:32] <j00bar> why are you using a list?

[16:20:41] <j00bar> (haven't read any of the other code)

[16:20:43] <aclark> j00bar: in filter or in redis?

[16:20:49] <j00bar> in redis

[16:21:17] <aclark> j00bar: because i when i read the redis docs i found lpush and that sounded like a good way to keep track of entries

[16:21:26] <j00bar> only if the order matters

[16:21:31] <aclark> it does

[16:21:44] <aclark> as i'm building a "recent" list

[16:21:47] <j00bar> ah.

[16:22:00] <aclark> at least i think it matters

[16:22:04] <j00bar> but you don't want duped entries.

[16:22:31] <aclark> right

[16:22:31] <j00bar> how large is the list?

[16:22:39] <aclark> well

[16:22:51] <aclark> i want to record the entries actually, because i'm going to do a "most requested"

[16:23:16] <j00bar> that sounds like a zset to me

[16:23:21] <aclark> so if someone requests Django 50 times and Pyramids 25 times…

[16:23:21] <aclark> oh?

[16:23:27] <j00bar> scored sets

[16:23:38] <aclark> as for size, it's infinite (or could be)

[16:23:39] <j00bar> values are unique but have a "score" associated with them

[16:23:46] <aclark> !google redis zset

[16:23:47] <pmxbot> http://twitter.com/antirez/status/5053225293 - Twitter / Salvatore Sanfilippo: The Redis ZSET type explai ...

[16:23:54] <j00bar> you can atomically increment the score associated with any member

[16:24:10] <j00bar> and you can grab by score ranges

[16:24:26] <aclark> j00bar: yeah your right, this needs to be redone non-aclark-ghetto-style, thanks

[16:24:35] <j00bar> http://redis.io/commands/#sorted_set

[16:25:06] <j00bar> and if you do stay with lists, lrange(key, 0, -1) for large sets is expensive if only for how much data it sends that you eventually throw away

[16:25:21] <aclark> i thought that too

[16:25:36] <j00bar> for small lists, back-and-forth for multiple redis lookups is slower

[16:25:51] <j00bar> for large lists, grabbing smaller chunks and concat'ing is faster

[16:26:06] <aclark> IC, interesting

[16:26:33] <j00bar> so if you were to identify the avg likelihood of there being dupes in any sequence of 20, (say it's 10), grab 30 entries

[16:26:37] <j00bar> then if you need more, grab more

[16:27:35] <aclark> efficient

[16:27:56] <j00bar> <3 teh redises

[16:32:01] <aclark> j00bar: it's really nice. i love stuff that is easy to get started with and redis could not have been simpler

[16:33:28] <j00bar> also hiredis++

[16:50:53] <aclark> time for "real work" ;-)

[16:52:30] <aclark> I'm already loving vanity.aclark.net for package pimping. never would have heard of ErikRose's http://pypi.python.org/pypi/blessings

[17:48:29] <j00bar> neat - tburke just showed me this: http://pypi.python.org/pypi/Dozer

[17:50:29] <aclark> j00bar: i've seen that, some dude came to a meetup and wanted to name his PythonPaste killer "Dozer" and people were like "Uhhh… it exists". Actually, I think it was me that said that because I looked the name up on PyPI

[17:50:45] <aclark> never tried it though

[17:54:39] <benji> I would think some sort of vanity badge would be popular. People could put it on their non-pypi package home page to show off.

[17:55:41] <aclark> benji: open a ticket! :-)

[17:56:22] <benji> aclark: you drastically underestimate my lazyness

[17:57:58] <j00bar> aclark: pmxbot needs to support opening tickets for you

[18:02:26] <aclark> j00bar: or DCPython_Hubot

[18:14:40] <j00bar> DCPython_Hubot sleeps in too late. doesn't even come into work until 1pm.

[18:25:37] <aclark> !slap DCPython_Hubot

[23:14:28] <aclark> jaraco: anyone ever thought about including pip w/distribute?

Log file Viewer

Help | Karma | Search:

#dcpython logs for Monday the 14th of November, 2011