#POSTGIS IRC Log - 2009-05-04

For logs after Feb 3, 2007, all times are GMT-8. Prior logs are GMT-9.
Back to Logs
03:31:10 sigq: geosfeed: Ticket #250 ( defect closed ): geos_contains( multipolygon, polygon ) returns false when it should return true <http://trac.osgeo.org/geos/ticket/250#comment:3>
09:52:16 CIA-34: pramsey * r4063 /branches/1.3/ ( ChangeLog NEWS doc/postgis.xml ): Changes in advance of 1.3.6
09:53:53 dassouki: http://www.neatorama.com/2009/04/27/the-geography-of-the-seven-deadly-sins/
09:53:54 sigq: Title: The Geography of the Seven Deadly Sins - Neatorama ( at www.neatorama.com )
09:56:08 CIA-34: robe * r4064 /trunk/ ( 3 files in 2 dirs ): more doxygen comment cleanup
09:56:59 CIA-34: pramsey * r4065 /tags/1.3.6/: tag 1.3.6
09:58:10 CIA-34: pramsey * r4066 /tags/1.3.6/Version.config: Alter version in tag, per our silly procedure.
11:04:00 CIA-34: kneufeld * r4067 /branches/1.3/Version.config: bumped branch svn to 1.3.7SVN
11:38:50 zanberdo: I seek guidance. I have two tables, one with a point geometry and another with a multi polygon geometry. I want to update a field for each of the records in the table with the points with a field from the table with the multi polygon. I what to qualify the update where the point in table 1 is within the multi polygon of table two. ( more )
11:40:36 zanberdo: I know I can query records from table 1 that are within the polygon of table 2 using ST_WITHIN as: select t1.field from table1 as t1, table2 as t2 where ST_WITHIN( t1.the_geom, t2.the_geom )
11:41:18 zanberdo: but I'm not clear how to structure the update statement. I want to update table1.city with table2.name where table1.the_geom within table2.the_geom...
11:41:20 zanberdo: and suggestions?
11:44:52 strk: geos svn hung
11:45:13 strk: confirmations / refusals ?
11:45:51 : * strk expects a refusal from CIA
11:48:05 CIA-34: strk * r2459 /trunk/source/ ( 2 files in 2 dirs ): Extend to accept a vector of const Coordinate pointers for snap pointers, fix typo making it clone snap points rather than source points on snapping...
11:51:04 CIA-34: strk * r2460 /trunk/source/ ( 4 files in 4 dirs ): Port overlay.snap.GeometrySnapper
12:58:21 CIA-34: strk * r2461 /trunk/source/ ( 6 files in 4 dirs ): Port overlay.snap.SnapOverlayOp, update GeometrySnapper to be more useful
13:24:29 CIA-34: strk * r2462 /trunk/ ( 25 files in 7 dirs ): Move overlay.validate package files where they belong
13:59:15 CIA-34: strk * r2463 /trunk/source/ ( 5 files in 4 dirs ): Port SnapIfNeededOverlayOp
14:17:23 strk: uhm
14:17:29 strk: and now that the whole system is setup...
14:17:50 strk: I don't feel like sacrifying the fantastic BinaryOp : )
14:50:00 strk: pramsey: BinaryOp was our answer to the robustness issues. At that time it was JTS following GEOS IIRC
14:50:22 strk: more or less. or at least it was a joint effort.
14:50:39 pramsey: OK... and now?
14:52:01 strk: now I ported the JTS-proper way of doing that thing
14:52:15 strk: only, neither the CAPI nor the XMLTester code use that
14:52:32 strk: the JTS-proper way is in Geometry::{overlayOperationHere}( xxx )
14:52:44 strk: while both XMLTester and CAPI use the BinaryOp( ) construct instead
14:53:03 strk: could switch both back to the original
14:53:07 strk: in XMLTester is still commented out
14:53:16 strk: ie:
14:53:16 strk: //GeomAutoPtr gRealRes( gA->Union( gB ) );
14:53:16 strk: GeomAutoPtr gRealRes = BinaryOp( gA, gB, overlayOp( OverlayOp::opUNION ) );
14:53:30 pramsey: right, and what's the *difference* ?
14:53:32 strk: gA->Union( gB ) <--- would automatically use the snapping heuristic now
14:53:35 pramsey: ( this is the part I dno't get )
14:53:40 strk: that's a good question : )
14:54:05 strk: probably not much difference
14:54:09 strk: but evil is in details
14:54:14 pramsey: no kiddlinh
14:54:28 strk: BinaryOp is full of compile-time options
14:54:34 strk: what to use and in which order...
14:54:41 strk: some steps are enabled, some disabled
14:54:52 strk: was a real-world-case driven configuration
14:54:59 strk: and fine-tuned for performance IIRC
14:55:15 strk: so my feeling is that out-of-the-box the JTS_proper way would be slightly slower
14:55:40 pramsey: well, my problem is I'm clueless
14:55:57 strk: ic
14:56:02 pramsey: so I can't help with useful advice other than to see if the list has any more wisdom
14:56:06 strk: well, I was just mentioning for the other thing you asked me : )
14:56:20 pramsey: and to write up a .txt file talking about what we're doing and why w're doing it
14:56:26 strk: as if you want to skip JTS it'll create this same situation later : )
14:56:26 pramsey: and committing it to the repo
14:56:34 pramsey: ah.
14:57:08 pramsey: well, since the splitting problem can be done w/o alterations for the line/line case that opportunity is currently on hold
14:57:36 strk: yeah, can probably take lines out of polys, split and then rebuild the polis
15:00:16 CIA-34: strk * r2464 /trunk/source/ ( 3 files in 2 dirs ):
15:00:16 CIA-34: Bring Geometry port up a revision by having it use SnapIfNeededOp for overlay
15:00:16 CIA-34: operations. Note that currently this doesn't affect the C-API nor the XMLTester
15:00:16 CIA-34: codes, both using the BinaryOp original class ( not JTS-ported ) doing about the
15:00:16 CIA-34: same thing as the SnapIfNeededOp class.
15:19:12 CIA-34: strk * r2465 /trunk/source/ ( 3 files in 2 dirs ): Geometry up to 1.112 with a minor bugfix and performance improvement in Geometry::covers
15:39:28 CIA-34: strk * r2466 /trunk/source/operation/IsSimpleOp.cpp: Fix memory bug
15:44:41 strk: ok, I've a response now to the question: "what's the difference between BinaryOp and SnapIfNeededOp"
15:44:51 strk: the latter fails like hell our testsuite
15:45:02 strk: better sleep over it
16:00:36 CIA-34: strk * r2467 /trunk/source/geom/util/GeometryTransformer.cpp: Fix memory leak on exception
16:09:07 dassouki: has anyone used pgrouting here?
20:25:24 CIA-34: robe * r4068 /trunk/postgis/ ( 7 files ): doxygen friendly comments
20:40:34 CIA-34: robe * r4069 /trunk/postgis/lwgeom_functions_basic.c: doxygen friendly comments