#POSTGIS IRC Log - 2008-05-28

For logs after Feb 3, 2007, all times are GMT-8. Prior logs are GMT-9.
Back to Logs
03:35:17 low: hi there. is there any way to get the position of the closest object to another object ?
03:39:29 mcayland: i think there may be some nearest neighbour search ideas on the wiki?
04:18:09 low: mcayland: will take a look, thx.
04:53:00 simplexio: low: slowest wat to do it is st_distance( geom,geom ) to all geams in table, faster way to do it for point is like... where geom && st_extend( geompoint ) nad then order by distance( geom,geom )
04:53:05 simplexio: -typos
06:23:17 nside: hello
06:24:00 nside: is it common to partition gis data in postgis?
06:24:38 nside: let's say i have billions of user accounts, i'd partition the records according to some hash functions
06:26:16 nside: i guess the same thing is doable with geometries.. using some kind of rtree/quadtree structure
06:30:57 low: hi again
06:34:32 mloskot: nside: what is the purpose of clustering?
06:34:52 mloskot: it's unclear to me "what for"
06:51:54 rupert: hi folks.. I have around 72 tables on a debian1 ( pg8.2.7, pg1.3.1 ). I want to pg_dump all the tables BUT without the postgis functions.. then transfer it to a debian2 ( pg8.3.1, pg1.3.3 ). Do I have to specify each and every table after the "-t" flag?
06:54:29 mcayland: rupert: if you are dumping in custom format, look at the -L / -l options to pg_restore
06:58:15 rupert: -L "use specified table of contents for ordering output from this file"
06:59:36 rupert: mcayland: so does it mean.. i can order the way it insert my tables? say i have "district", "district_20080507" originally on my pg_dump file, i want to only restore the "district" table.. il put it on this "-L" flag?
07:00:51 nside: mloskot: to handle billions of records
07:01:12 mloskot: what's problem with handling then in a single table
07:01:23 nside: the table gets too big, won't fit on a single hd
07:01:28 mcayland: rupert: nearly. custom format pg_dump files contain a catalogue of contents which you output using pg_restore -l
07:01:33 mloskot: have you measured?
07:01:36 nside: eyp
07:01:38 nside: yep
07:01:55 mcayland: you can then edit this, leave only the bits you want to restore, and then feed the list back into pg_restore using -L
07:01:56 mloskot: then you have answer - divide
07:02:18 mloskot: using tablespace, schemas, etc
07:02:44 nside: that's my question: how can one partition geometries
07:03:14 mloskot: imo, not feasible
07:04:01 rupert: mcayland: ok.. thanks for the clarification. i will try it out now..
07:04:15 mloskot: to divide singl geom
07:04:22 mloskot: physically
07:09:02 mcayland: nside: if one drive is too big, you could consider a software RAID/LVM to combine multiple physical drives into one virtual drive
07:09:26 mcayland: nside: i mean one *drive too small* or your data is too big
07:09:28 simplexio: small you mean=
07:09:30 simplexio: ?
07:09:41 mcayland: simplexio: ; )
07:10:23 mcayland: or depending upon your application, you may be able to partition at app level. but it will need some manual work.
07:10:28 simplexio: i dont think thats is problem nowdays, i mean how much data you need to fill quite "cheap" 5x500G raid5 set
07:10:30 nside: let's say i want to store 100 tera
07:10:53 simplexio: : )
07:11:10 mcayland: nside: for a single table???!
07:11:14 simplexio: where you get that much data
07:11:37 nside: not a single table, that's why i want to partition
07:12:00 simplexio: rule number one, if you want it to be fast. shile index from table must be smaller than amount ram in server
07:12:41 mcayland: nside: have you looked at postgresql tablespaces?
07:12:43 nside: i'd like to have many postgres instances distributed over 10-20 servers
07:13:03 nside: mcayland: why?
07:13:35 mcayland: nside: so you can assign sets of tables to sets of physical disks
07:13:37 simplexio: i have about 99% from finnish roads and all data associat to them and dataset is 10G
07:14:18 mloskot: nside: IMO reasonable suggestion is that you should check the postgresql manual and analyse all tools provided by PostgreSQL
07:14:33 nside: i looked at hypertable and hbase, but unfortunately there's no indexing
07:14:48 mloskot: nside: there are no extra tools specific to PostGIS regarding physical clustering
07:14:56 mloskot: Just use what PostgreSQL has in the box.
07:15:13 nside: mloskot: i saw some nice tools built by skype
07:15:17 simplexio: mloskot: thanks, that tablespace was something new for me
07:15:29 mloskot: nside: http://www.postgresql.org/docs/8.3/interactive/high-availability.html
07:15:31 sigq: Title: PostgreSQL: Documentation: Manuals: PostgreSQL 8.3: High Availability, Load Balancing, and Replication ( at www.postgresql.org )
07:15:43 nside: it's at https://developer.skype.com/SkypeGarage/DbProjects
07:15:45 sigq: Title: SkypeGarage/DbProjects - Skype Developer Zone ( at developer.skype.com )
11:12:26 sheats: so i am getting "server process ( PID 17494 ) was terminated by signal 11: Segmentation fault" in my logs from doing a COLLECT function call
11:12:48 sheats: full details here... http://dpaste.com/53400/
11:12:50 sigq: Title: dpaste: #53400 ( at dpaste.com )
11:14:11 sheats: it only happens on one pcn instance, others work fine...
11:15:11 sheats: i was wondering if anyone could provide some insight on what I can look at...
13:33:29 CIA-6: pramsey * r2787 /trunk/doc/postgis.xml: Fix for issue 30, ST_Buffer doco mistake
13:33:33 CIA-6: pramsey * r2788 /branches/1.3/doc/postgis.xml: Fix for issue 30, ST_Buffer doco mistake
13:42:42 CIA-6: pramsey * r2789 /trunk/doc/postgis.xml: Bug reporting documentation ( issue 27 )
13:42:48 CIA-6: pramsey * r2790 /branches/1.3/doc/postgis.xml: Bug reporting documentation ( issue 27 )
13:44:46 CIA-6: pramsey * r2791 /branches/1.3/doc/postgis.xml: Geomunion name change note, issue 26.
13:44:51 CIA-6: pramsey * r2792 /trunk/doc/postgis.xml: Geomunion name change note, issue 26.
13:56:09 CIA-6: pramsey * r2793 /trunk/spatial_ref_sys.sql: Fix towgs84 for 31300 and 31370 ( issue 23 )
13:56:12 CIA-6: pramsey * r2794 /branches/1.3/spatial_ref_sys.sql: Fix towgs84 for 31300 and 31370 ( issue 23 )
14:56:15 CIA-6: pramsey * r2795 /branches/1.3/lwgeom/ ( 7 files ): Enhanced speed _ST_DWithin( g,g,d ) that returns as soon as g and g are within d of each other, rather than using distance naively. Change ST_DWithin to use enhanced op. ( Issue 20 )
15:00:51 CIA-6: pramsey * r2796 /trunk/lwgeom/ ( 7 files ): Enhanced speed _ST_DWithin( g,g,d ) that returns as soon as g and g are within d of each other, rather than using distance naively. Change ST_DWithin to use enhanced op. ( Issue 20 )