IRC Logs for #cmvt Tuesday, 2013-04-09

*** pvrdwb has joined #cmvt06:00
*** pvrdwb has quit IRC06:26
*** pvrdwb has joined #cmvt14:58
circuits25e6c03aedc4 by prologic: Got WFS to support the output of GeoJSON using the OGR GeoJSON driver.16:54
circuits86493896f00f by prologic: No longer need a template16:54
circuitsba8a11c005c1 by prologic: Testing performance of 1k (layerx) and 250m (layery) datasets with MapServer tile rendering16:54
circuits4c9c5eb8afc2 by prologic: Playing with performance of vector intersections with spatiallite vs. python + shapely16:54
*** pvrdwb has joined #cmvt19:10
DanielMacbookCramacbook crashed, first time ever19:22
prologicI've crashed my iMac twice in the time I've been here19:24
prologicI saw your IM(s)19:25
prologicGreat :) Looks like I've influenced you guys into using GeoTiff :)19:25
prologicwhich is way way way way faster! :)19:25
prologicMaybe you don't need to do this stuff on a cluster anymore? :)19:25
DanielMacbookCraback.  yes you can take credit for anything we do from now on19:54
DanielBairdi suspect the cluster modelling stuff is all about R and maxent data structures, but I'd like to revisit my other web systems and switch them over to geotiff.  probably wouldn't take too long, it's just a line or two in the mapfile19:56
prologicyeah well20:09
prologicdealing with "machine" data is a whole lot faster than dealing with "human data"20:10
prologicThat's why we also invented BSON20:10
prologicnot human readable, but can be easily converted to JSON20:10
prologicBSON == Binary JSON20:10
prologicMongoDB uses it internally20:10
DanielBairdremember when xml was cool?20:14
DanielBairdi was all into XSLT and whatnot20:15
DanielBairdactually that might perhaps have predated your entry into the wonderful world of software development...20:16
prologicyou'd be surprised :)20:20
prologicI've been programming since I was 11 :)20:21
prologicwe're going to want to use this I think20:27
DanielBairdsuspect leaflet is the right way to go.. I used geojson with open layers, it looks fine with leaflet too20:44
prologicI don't care either way :)20:45
prologicleaflet is a lot more lightweight though20:45
prologicand seems faster :)20:45
prologicbut I was more talking about the geojson + leaflet stuff20:45
DanielBairdopen layers lets you define multiple vector layers and have normal layer control over them.  i think leaflet treats all the vector objects as without a layer (or as all in one layer), but i think using feature collections will make up for that20:45
prologicI can spit geojson at you now20:45
prologicand still working out the finer details of spitting geojson at you given some request parameters and operations20:46
prologice.g: intersections, give me the nth feature, etc20:46
prologictossing up between custom and using MapServer's WFS (which sucks)20:46
prologicI haven't defined any RESTful services you can plug into yet20:47
prologicas I'm not sure what we're going to need20:47
prologicI've just been toying around with smaller bits and pieces (getting to know how stuff works and should work at a lower level)20:47
prologichow's the AngularJS + Leaflet shell going?20:49
prologicthat good huh? :)20:49
prologicwell I haven't seen any commits so ;020:49
DanielBairdit's probably fine, i just haven't had many actual at-the-keyboard-typing minutes to spend on it20:50
prologicwhat's the other low level things we need20:50
prologicintersection the datasets (raster) values themselves I think20:50
prologicI don't have that figured out yet20:50
prologice.g: the 5km/1km/250m datasets20:50
prologicI presume we need to take a bounding box20:50
prologicor a polygon of a region20:50
prologicand get the values of the corresponding raster dataset20:51
DanielBairdi wonder if we will have to categorise datasets to say whether they're suitable to chart against each other20:51
prologicor a point (marker)20:51
prologicgod only knows :)20:51
prologicbut that sounds reasonable yeah?20:51
DanielBairdmm the point thing will be a bit odd.  probably people really mean "the 250m square this point falls into"20:52
prologicintersect out raster datasets given either a point, polygon or bounding boxx?20:52
prologicgiving us a subset array to work with20:52
prologicyes - *nods*20:52
prologicpoint == the square it intersects with20:52
DanielBairdyes there will need to be some way to do numerical exclusion on a region20:52
prologicthe 5km/1km/250m square20:52
prologicok I'll work on this next20:53
prologicbe a pretty useless app without this I think20:53
DanielBairdif we only care about named regions, then you can pre-calculate the summary info for every nameable region, which needs complex polygons but can be pre-calculated.20:54
DanielBairdif we are letting the user define their own regions, we'll need to do live summary info, but might be able to restrict the user to a rectangle20:54
DanielBairdwe'll have to resolve the off-by-0.5 thing somehow.20:55
DanielBaird(worry i went back to points)20:55
DanielBaird* sorry, not worry20:56
DanielBairdanyway we'll fiddle that when we get to it20:58
prologicgood lord21:11
prologicclipping a raster is not an easy process21:11
DanielBairdeasier to do a rectangular clip then test each pixel individually, maybe?21:12
prologicno I think the process is going to be the same overall21:12
prologicI just didn't expect so much code21:12
prologicand so many steps in the process21:12
prologicbasically you treat your data sources (rasters)21:12
prologicas images21:12
prologicand then convert your polygon to an image as well21:13
prologicthen use the polygon/point/whatever as a mask21:13
prologiccomplicated :)21:13
prologicI'll have a play21:13
prologicalso with the summary stuff21:13
prologicI'm not particularly running into any performance issue per say21:13
DanielBairdthere's no help in a geo library>21:13
prologicso we might be able to just compute said values on the fly21:13
prologicand just cache them21:14
prologicbut obviously if things get slow then we can precompute them21:14
DanielBairdno need to pre-optimise21:14
prologichelp in what?21:14
DanielBairdin doing clipping to a map boundary21:14
DanielBairdseems like something a geo lib would have covered21:14
prologicyou would think so21:15
prologicI think there is actually21:15
prologicin the gdal library21:15
prologicbut I haven't read up on it yet21:15
prologiccya tomorrow22:37

Generated by 2.11.0 by Marius Gedminas - find it at!