[postgis-tickets] r16552 - GEOS ClipByBox2D stub notice now includes ticket link
Darafei
komzpa at gmail.com
Tue Apr 24 04:52:21 PDT 2018
Author: komzpa
Date: 2018-04-24 04:52:20 -0700 (Tue, 24 Apr 2018)
New Revision: 16552
Modified:
trunk/doc/reference_processing.xml
trunk/liblwgeom/lwgeom_geos.c
Log:
GEOS ClipByBox2D stub notice now includes ticket link
Closes #4039
Closes https://github.com/postgis/postgis/pull/239
Modified: trunk/doc/reference_processing.xml
===================================================================
--- trunk/doc/reference_processing.xml 2018-04-24 09:28:02 UTC (rev 16551)
+++ trunk/doc/reference_processing.xml 2018-04-24 11:52:20 UTC (rev 16552)
@@ -490,10 +490,9 @@
<title>Description</title>
<para>
-Clips a geometry by a 2D box in a fast but possibly dirty way. The output
-geometry is not guaranteed to be valid (self-intersections for a polygon
-may be introduced). Topologically invalid input geometries do not result
-in exceptions being thrown.
+Clips a geometry by a 2D box. The output geometry is not guaranteed to be valid
+(self-intersections for a polygon may be introduced).
+Topologically invalid input geometries do not result in exceptions being thrown.
</para>
<para>Performed by the GEOS module.</para>
@@ -500,6 +499,7 @@
<note><para>Requires GEOS 3.5.0+</para></note>
<para>Availability: 2.2.0 - requires GEOS >= 3.5.0.</para>
+ <para>Changed: 2.5.0 - wrapper around ST_Intersection to work around GEOS bugs. </para>
</refsection>
@@ -3151,9 +3151,9 @@
<title>Description</title>
<para> Returns a "smoothed" version of the given geometry using the Chaikin algorithm.
See <ulink url="http://www.idav.ucdavis.edu/education/CAGDNotes/Chaikins-Algorithm/Chaikins-Algorithm.html">Chaikins-Algorithm</ulink> for an explanation of the process.
- For each iteration the number of vertex points will double.
+ For each iteration the number of vertex points will double.
The function puts new vertex points at 1/4 of the line before and after each point and removes the original point.
- To reduce the number of points use one of the simplification functions on the result.
+ To reduce the number of points use one of the simplification functions on the result.
The new points gets interpolated values for all included dimensions, also z and m.</para>
<note><para>Note that returned geometry will get more points than the original.
@@ -3205,15 +3205,15 @@
<refsection>
<title>Description</title>
- <para>Filters away vertex points based on their m-value. Returns a geometry with only
+ <para>Filters away vertex points based on their m-value. Returns a geometry with only
vertex points that have a m-value larger or equal to the min value and smaller or equal to
- the max value. If max-value argument is left out only min value is considered. If fourth argument is left out the m-value
- will not be in the resulting geoemtry. If resulting geometry have too few vertex points left for its geometry type an empty
- geoemtry will be returned. In a geometry collection
+ the max value. If max-value argument is left out only min value is considered. If fourth argument is left out the m-value
+ will not be in the resulting geoemtry. If resulting geometry have too few vertex points left for its geometry type an empty
+ geoemtry will be returned. In a geometry collection
geometries without enough points will just be left out silently. If </para>
- <para>This function is mainly intended to be used in conjunction with ST_SetEffectiveArea. ST_EffectiveArea sets the effective area
+ <para>This function is mainly intended to be used in conjunction with ST_SetEffectiveArea. ST_EffectiveArea sets the effective area
of a vertex in it's m-value. With ST_FilterByM it then is possible to get a simplified version of the geoemtry without any calculations, just by filtering</para>
-
+
<note><para>There is a difference in what ST_SimplifyVW returns when not enough points meets the creterias compared to ST_FilterByM.
ST_SimplifyVW returns the geometry with enough points while ST_FilterByM returns an empty geometry</para></note>
<note><para>Note that the retuned geometry might be invalid</para></note>
Modified: trunk/liblwgeom/lwgeom_geos.c
===================================================================
--- trunk/liblwgeom/lwgeom_geos.c 2018-04-24 09:28:02 UTC (rev 16551)
+++ trunk/liblwgeom/lwgeom_geos.c 2018-04-24 11:52:20 UTC (rev 16552)
@@ -884,8 +884,14 @@
{
LWGEOM *result;
LWGEOM *tmp;
- LWGEOM *envelope = (LWGEOM*)lwpoly_construct_envelope(geom1->srid, x1, y1, x2, y2);
+ /* This lwgeom_intersection should be a call to GEOSClipByRect:
+ * g3 = GEOSClipByRect(g1, x1, y1, x2, y2);
+ * Unfortunately as of GEOS 3.7 it chokes on practical inputs.
+ * GEOS ticket: https://trac.osgeo.org/geos/ticket/865
+ */
+
+ LWGEOM *envelope = (LWGEOM *)lwpoly_construct_envelope(geom1->srid, x1, y1, x2, y2);
result = lwgeom_intersection(geom1, envelope);
lwgeom_free(envelope);
@@ -908,32 +914,6 @@
lwgeom_simplify_in_place(result, 0.0, LW_TRUE);
return result;
-
-#if 0 /* POSTGIS_GEOS_VERSION >= 35, enable only after bugs in geos are fixed */
- int32_t srid = get_result_srid(geom, NULL, __func__);
- uint8_t is3d = FLAGS_GET_Z(geom->flags);
- GEOSGeometry *g1, *g3;
-
- if (srid == SRID_INVALID) return NULL;
-
- /* A.Intersection(Empty) == Empty */
- if (lwgeom_is_empty(geom)) return lwgeom_clone_deep(geom);
-
- initGEOS(lwnotice, lwgeom_geos_error);
-
- if (!input_lwgeom_to_geos(&g1, geom, __func__)) return NULL;
-
- g3 = GEOSClipByRect(g1, x1, y1, x2, y2);
-
- if (!g3) return geos_clean_and_fail(g1, NULL, NULL, __func__);
-
- if (!output_geos_as_lwgeom(&g3, &result, srid, is3d, __func__))
- return geos_clean_and_fail(g1, NULL, g3, __func__);
-
- geos_clean(g1, NULL, g3);
-
- return result;
-#endif /* POSTGIS_GEOS_VERSION >= 35 */
}
/* ------------ BuildArea stuff ---------------------------------------------------------------------{ */
More information about the postgis-tickets
mailing list