#mapnik log: Monday 09, February 2009

2009 | 02

previous | next
02:20:24 *** migurski (n=migurski@dsl081-049-227.sfo1.dsl.speakeasy.net) has joined #mapnik
02:59:02 *** migurski has quit ()
04:04:52 *** rcoup_ has quit ()
04:27:50 *** D3f0 has quit (Remote closed the connection)
04:30:58 *** aub has quit ()
06:20:37 *** springmeyer (n=dane@c-24-19-50-92.hsd1.wa.comcast.net) has joined #mapnik
07:48:00 *** xcacou (n=aga@AToulouse-157-1-165-31.w83-193.abo.wanadoo.fr) has joined #mapnik
08:21:05 *** kunitoki (i=55121f9a@gateway/web/ajax/mibbit.com/x-c2361faf062ca282) has joined #mapnik
08:21:12 *** kunitoki_ (i=55121f9a@gateway/web/ajax/mibbit.com/x-28ecd17717c67dc3) has joined #mapnik
08:22:37 *** kunitoki1 (n=kraken@host210-49-dynamic.7-79-r.retail.telecomitalia.it) has joined #mapnik
08:23:09 *** kunitoki has quit (Client Quit)
08:24:09 *** kunitoki_ has quit (Client Quit)
08:36:55 *** kunitoki1 is now known as kunitoki
12:32:22 *** aub (n=aubrey@cpe-72-227-134-148.nyc.res.rr.com) has joined #mapnik
12:41:04 *** aub has quit ()
14:04:54 *** artem (n=artem@i-83-67-142-225.freedom2surf.net) has joined #mapnik
14:08:52 *** aub (n=aubrey@static-70-107-236-83.ny325.east.verizon.net) has joined #mapnik
14:09:32 *** aub_ (n=aubrey@static-70-107-236-83.ny325.east.verizon.net) has joined #mapnik
14:09:32 *** aub has quit (Read error: 104 (Connection reset by peer))
14:48:03 <kunitoki> artem: any chance to apply sqlite patches ? i know they are uber-alpha but may be simpler for tracking changes if they gets checked in
14:52:08 <artem> kunitoki: yes, I'm playing with your patches at the moment. I'll apply then asap . We need to decide how we're going to ref spatial index table. How about extending 'geometry_columns' to have extent and spatial index table name
14:52:23 <artem> ?
14:57:20 <kunitoki> well it is interesting
14:57:41 <kunitoki> anyway spatialite wkb already have extent
14:58:31 <kunitoki> if you look here:
14:58:48 <kunitoki> http://www.gaia-gis.it/spatialite-2.3/spatialite-2.3_manual.html#t3 in section 3.3
14:59:34 <kunitoki> but probably you mean table-wise-extent and not single-row-extent ?
15:00:27 <kunitoki> ah yes
15:00:31 <artem> yes, layer extent
15:01:19 <kunitoki> yes, if you look at my code, in the sqlite_datasource there is a #if 0 around that
15:01:41 <kunitoki> you can specify that table, and it looks for minx, miny, maxx, maxy
15:01:49 <kunitoki> i call it geocatalog
15:02:57 <kunitoki> anyway there is also the MBR Caching support in spatialite (which adds some triggers for managing cache of table extents)
15:03:13 <kunitoki> the only drawback is: "creation of both R*Tree and MbrCache on the same Geometry column is not supported"
15:03:20 <artem> kunitoki: yep, I saw it. I just winder if using 'standard' geometry_columns is better strategy
15:03:58 <kunitoki> yeah but 'standard' geometry_columns doesn't have extent. nor we require triggers to keep those extent in sync with the table
15:04:38 <kunitoki> we shuld use MbrCache for that purpose, but we lose the spatial index on that column
15:04:49 <kunitoki> which is unacceptable
15:05:36 <artem> I'm about to go out but I'll read about MbrCache .
15:06:58 <artem> I don't quite understand why MbrCache and r*tree index are self excluding
15:07:41 <artem> r*tree index is just a virtual table, mbr-cache is another, right ?
15:08:25 <kunitoki> yeah they are different kind of indexes. one is a virtual table, the other is in memory
15:08:56 <artem> i think it's spatialite restriction rather then sqlite
15:09:22 <kunitoki> yeah
15:09:26 <kunitoki>  i think too
15:09:41 <artem> btw, r*tree index works quite well
15:09:55 <kunitoki> have you implemented something on my code to use it ?
15:10:08 <artem> I'm reading ~40,000 linestring features (Norway OSM)
15:10:08 <kunitoki> i'm curious :=
15:10:24 <artem> yes, I added some hacks to your patch:)
15:10:37 <kunitoki> obviously. you welcome :D
15:11:10 <kunitoki> it's fast ?
15:11:22 <artem> we need a good query builder to merge client SQL with spatial index
15:11:37 <artem> yes, it is fast for this small data set
15:11:42 <kunitoki> mmmh that is one of my newer question for you
15:11:51 <kunitoki> but we will discuss about it
15:12:29 <kunitoki> cause most of the time i need to pass to the datasource (oracle) special queries with join and the use of more specialized where clauses
15:13:12 <artem> yes, sure . I really need to run now. I'll commit my hacks later on today and we can discuss how to take it  forward.
15:13:23 *** artem has quit ()
15:45:23 *** darragh (n=darragh@83.70.173.25) has joined #mapnik
15:49:20 *** darragh has parted #mapnik ()
16:32:34 *** D3f0 (n=defo@190.176.194.209) has joined #mapnik
16:40:38 *** xcacou has quit (Remote closed the connection)
16:45:01 *** kunitoki has quit ("leaving")
17:56:59 *** kunitoki (n=kraken@host210-49-dynamic.7-79-r.retail.telecomitalia.it) has joined #mapnik
18:50:59 *** cmarqu has quit (Remote closed the connection)
18:54:56 *** rcoup (n=rcoup@ip-118-90-27-63.xdsl.xnet.co.nz) has joined #mapnik
19:10:25 <Mrfo> springmeyer, im running into some issues using that pointsymbolizer and text style you created before
19:10:46 <springmeyer> okay
19:11:02 <Mrfo> i think because the point icon is alot smaller than the text, i get times where i get only points
19:11:50 <Mrfo> lemmie take a quick snapshot
19:14:20 *** D3f0 has quit ("Konversation terminated!")
19:18:56 <Mrfo> http://img15.imageshack.us/img15/1006/mapwithoutny7.png -- this is the map without the points
19:19:32 <Mrfo> http://img509.imageshack.us/img509/1199/mapwithpointsbj8.png -- this is with them on
19:20:13 <Mrfo> the area to note is at the bottom centerish area. theres more dots than cities
19:22:08 <springmeyer> yes, I see that
19:22:39 <springmeyer> makes sense because, like you say, the collision avoidance affects the 'larger' text more than the points
19:23:03 <springmeyer> Mrfo: do you ideally/cartographically want more text or less points?
19:23:21 <springmeyer> nice maps btw!
19:23:46 <Mrfo> ummmm, less points would be good. im fine with fewer cities being on there as long as theres not extra dots
19:23:52 <Mrfo> thanks :)
19:26:17 <Mrfo> i tried the shieldsymbolizer, but the dx,dy for the shieldsymbolizer shifts the entire shield
19:26:48 <springmeyer> hmm, ya, this is an obvious need but I don't see an easy solution
19:28:56 <springmeyer> i think a 'min_distance' parameter on the PointSymbolizer would get you closer (if that were added)
19:29:17 <springmeyer> but would still lead to some points being placed where text was not (due to size differences)
19:30:50 <Mrfo> yeah. i can't think of much either
19:30:54 <Mrfo> its no biggie though
19:31:43 <springmeyer> This seems like a common usecase: some smart way of truly associating the point with the text is needed
19:32:15 <springmeyer> and ya, the control is not their in the ShieldSymbolizer either, even though that gets closer to this idea ^^^
19:37:38 <Mrfo> yeah. not sure what the solution is
19:37:46 <Mrfo> but ill see if i can think of something
19:41:00 *** artem (n=artem@i-83-67-142-225.freedom2surf.net) has joined #mapnik
19:43:42 <springmeyer> ya, perhaps some kind of PointLabelSymbolizer that borrows from all the nice TextSymbolizer rules but enforces that a point will be places only if the text label is placed
19:43:58 <nikq> Mapnik Trac: Changeset [881]: + sqlite-input-plugin.patch (kunitoki) + wkb-sqlite.patch (kunitoki) +  ... | http://trac.mapnik.org/changeset/881
19:44:06 <Mrfo> yeah that would be pretty cool
19:44:24 <springmeyer> and then it could have additional options to allow you to control what direction the label is placed in relation to the point
19:44:38 <Mrfo> yeah i was just about to mention that
19:44:54 <Mrfo> ideally you could get a better fit if you didn't care about directions either
19:45:03 <springmeyer> up/down , n/e/s/w or 'best fit'
19:45:10 <springmeyer> ya, right :)
19:46:07 <springmeyer> but in cluttered maps it is often nice for the eye/undstanding to be able to find a label then look for the point off of a predicable side
19:46:23 <springmeyer> or at least I try to achieve that effect when manually making labels
19:46:33 <Mrfo> thats a good point
19:46:51 <Mrfo> ideally they both have their uses
19:47:02 <springmeyer> Mrfo: maybe a ticket discribing this for the 1.0.0 milestone?
19:47:19 <Mrfo> sure i can do that
19:55:17 <springmeyer> Mrfo: how does your server search go?
19:56:16 <Mrfo> heh, not too good. costs a bit more than i figured
19:56:57 <springmeyer> oh, I mean your choice of tool to serve mapnik tiles....
19:57:08 <CIA-23> mapnik-utils: dane.springmeyer * r548 /trunk/nik2img/ (CHANGELOG.txt nik2img.py): Add alpha support for rendering a map by pulling the map definition from a python script 'load_pymap()'
19:57:42 <springmeyer> but I'd be interested about what server hardware too
19:58:45 <Mrfo> ohh
19:58:46 <Mrfo> lol
19:59:04 <Mrfo> ummm, still using tilecache atm. i tried some stuff with modtiles
19:59:31 * crschmidt would not recommend tilecache if you have complex/slow tiles
19:59:36 <crschmidt> and lots of users
20:00:40 <Mrfo> i think mod tile is going to be the best solution, its just going to take a bit to set it up
20:01:39 <Mrfo> although given how often the tiles need to update, i dont know if theres any perfect solution really
20:03:20 <Mrfo> crschmidt, is tilecache not a good solution with complex tiles?
20:03:58 <Mrfo> it seems to render about as fast as mapnik viewer or nik2img
20:04:44 <crschmidt> TileCache renders inside the serving process
20:04:57 <crschmidt> so if you have many users all browsing areas they haven't seen before, they'll use up all your apache children
20:05:08 <crschmidt> mod_tile uses a single backend renderer
20:05:23 <crschmidt> So you don't have lots of different processes trying to render at the same time
20:05:36 <crschmidt> with TileCache, you can get 100 processes all fighting for the same couple CPUs
20:05:40 <crschmidt> with mod_tile, you don't have that
20:05:44 <Mrfo> ah
20:05:57 <crschmidt> If you have fast-to-render tiles, that's okay
20:06:09 <crschmidt> because the tiles render quick enough that you don't get a backlog
20:06:15 <crschmidt> (this is why it's not a problem for most other services)
20:06:24 <crschmidt> but if your tiles are slow, then you get that backlog
20:06:36 <Mrfo> makess ense
20:06:38 <Mrfo> *sense
20:06:47 <crschmidt> It might be possible to make a rendered tile source for TileCache
20:07:03 <crschmidt> but it's probably not worth it
20:16:49 <CIA-23> mapnik: artem * r881 /trunk/ (11 files in 5 dirs):
20:16:49 <CIA-23> mapnik: + sqlite-input-plugin.patch (kunitoki)
20:16:49 <CIA-23> mapnik: + wkb-sqlite.patch (kunitoki)
20:16:49 <CIA-23> mapnik: + very preliminary spatial index support (idx_<tablename>_<geometry_field>)
20:17:24 *** ser has quit (kubrick.freenode.net irc.freenode.net)
20:17:45 *** ser (n=ser@sergiusz.pawlowicz.name) has joined #mapnik
20:24:35 <jburgess> Could we revert the use of the "-fast" gcc flag in r881? I believe it is only supported on the Apple version of gcc.
20:24:38 <nikq> http://trac.mapnik.org/changeset/881, at , by artem: + sqlite-input-plugin.patch (kunitoki)+ wkb-sqlite.patch (kunitoki)+ very preliminary spatial index support (idx_&lt;tablename&gt;_&lt;geometry_field&gt;)
20:27:51 *** D3f0 (n=defo@190.176.212.26) has joined #mapnik
20:41:47 <springmeyer> jburgess: +1 ya I bet that was just for testing
20:42:11 <artem> jburgess: my bad sorry, I'll revert that change
20:42:39 <jburgess> cheers :)
20:42:43 <springmeyer> I added the ability to set the OPTIMIZATION level, not sure if that is useful enought to keep or modify to accept an arbitrary flag like --fast ?
20:42:56 <springmeyer> up to you guys...
20:44:09 <nikq> Mapnik Trac: Changeset [882]: + reverting back accidental change | http://trac.mapnik.org/changeset/882
20:45:40 <artem> springmeyer : yes, OPTIMIZATION flag is good
20:46:46 <springmeyer> okay. I questioned it once I realized that -O3 wouldn't work
20:59:27 <CIA-23> mapnik: artem * r882 /trunk/SConstruct: + reverting back accidental change
21:05:47 <artem> springmeyer: -O3 works fine for me
21:06:51 <springmeyer> ah, okay, cool
21:06:58 <springmeyer> does cairo work?
21:07:15 <springmeyer> #193 is what happened to me with -O3
21:07:16 <nikq> Ticket #193: Bus Error during Image render at cairo extract_surface(), http://trac.mapnik.org/ticket/193
21:07:52 <springmeyer> but perhaps I assumed wrongly that it was the -O3 flag and it was actually a cairo install problem...
21:08:10 <artem> ah, ok. interesting
21:09:01 <artem> it works find with -fast which is just a shortcut to -O3 -fomit-frame-pointer ...
21:09:09 <artem> fine
21:09:17 <springmeyer> okay, great
21:41:13 <dukeku> urgh, mapnik box died
21:48:49 <artem> kunitoki: ping did you try using spatial index ?
22:22:21 *** artem has quit ()
22:24:19 *** artem (n=artem@i-83-67-142-225.freedom2surf.net) has joined #mapnik
23:18:30 *** weizhuo (n=chatzill@nat/yahoo/x-04358482f4ad2bbe) has joined #mapnik
23:46:37 *** artem has parted #mapnik ()