Paul Ramsey [Mon, 19 Jan 2009 21:33:14 +0000 (21:33 +0000)]
First revision of the GUI. Configure using --with-gui to enable full GUI build. New core/cli will build by default. Old utilities remain in place for now.
Mark Cave-Ayland [Thu, 15 Jan 2009 18:35:06 +0000 (18:35 +0000)]
Part 1 of GBT#94: Rename folders/files to be more consistent. Embed the MAJOR.MINOR version in the PostgreSQL shared library name (to allow different databases to contain different PostGIS versions) and also rename lwpostgis.sql to postgis.sql. Update documentation in various files to reflect the new names at the same time.
Mark Cave-Ayland [Thu, 15 Jan 2009 15:10:21 +0000 (15:10 +0000)]
Fix the output of "SELECT postgis_full_version()" on trunk by ensuring that USE_STATS is present, and correctly generating the POSTGIS_SCRIPTS_VERSION variable so that installations of trunk don't keep thinking that they need a scripts upgrade.
Mark Cave-Ayland [Tue, 13 Jan 2009 13:00:48 +0000 (13:00 +0000)]
Update lwgeom_gettypename() by renaming Curve to CircularString. Now lots of error messages scattered throughout liblwgeom should start to make sense...
Mark Cave-Ayland [Tue, 13 Jan 2009 12:59:03 +0000 (12:59 +0000)]
Rearrange liblwgeom.h as per Paul's comment on GBT#97: Rename LWCURVE to LWCIRCSTRING. Also rename lwfree_circstring() to lwcircstring_free() so that it matches the others (I guess this was missed in the first pass).
Mark Cave-Ayland [Tue, 13 Jan 2009 12:27:39 +0000 (12:27 +0000)]
Commit for GBT#97: Rename LWCURVE to LWCIRCSTRING (also know as restoring balance to the universe).
Having gone through and re-read sections of the CIRCULARSTRING code, it is amazing how much easier the code is to read without the confusion of a curve being either a CIRCULARSTRING or LINESTRING as per the SQL-MM spec. The resulting commit compiles & installs cleanly for me, passes "make check" and also Regina's torture script. Please let me know if I accidentally broke anything :)
Regina Obe [Mon, 12 Jan 2009 13:54:17 +0000 (13:54 +0000)]
Move over ST_Line functions and point out how pointless they are. Also correct incorrect statement that they throw errors when given non-line - they just return null.
Regina Obe [Fri, 9 Jan 2009 22:52:16 +0000 (22:52 +0000)]
Move over rest of miscellaneous -- NOTE st_find_srid does not exist - was never renamed from find_srid. Not sure its worth fixing since its rarely used anyway.
Paul Ramsey [Thu, 8 Jan 2009 17:04:00 +0000 (17:04 +0000)]
Alter lwfree_* to deep-free memory. Add variants of *_release to support shallow-free. Go through lwgeom/ and replace any lwfree_* instances with *_release where appropriate. (Surprisingly few.)
Allow CPPFLAGS/LDFLAGS parameters from the command line to be passed into the CUnit Makefile, since CUnit is the only dependency that cannot determine its own flags using a --with-X parameter. This allows CUnit to be installed in a non-standard location if required.
Rework the CUnit detection infrastructure as you can't locate include files based on an absolute path. This is because autoconf and the compiler have differing ideas of the current paths (this totally breaks the Win32 build since you have the extra mapping between UNIX-type MingW paths and Windows-type paths). Hopefully this should not affect existing users at all - if so, please post a report on -devel. I'll try running this on MingW at a later date.
I've also enforced "make check" to run the unit tests, rather than "make test". This is a fallout from the early days when strk committed the regression test suite to run using "make test" by accident which appears to have stuck :(
.
Fix comment in lwgeom_from_ewkt() and also add a quick error check to lwgeom_from_ewkt() and lwgeom_from_ewkb() - otherwise upon parse failure we could pass a half-baked geometry back to the caller which would not be good :(
Add -lm to the cu_tester link line - we must always add this to the link line for anything that links agains liblwgeom.a (at least on my Linux system here)
Commit quick fix for GBT#85: ST_Distance crashes on Circular String. Rather than devise and implement a set of distance functions for CIRCULARSTRINGs, I've simply added code to detect the condition and throw an "Unsupported geometry type" error, similar to as already exists within the codebase.
Paul Ramsey [Tue, 23 Dec 2008 20:20:48 +0000 (20:20 +0000)]
Rename new functions to ST_LineCrossingDirection and ST_LocateBetweenElevations for more explicitness and closer correspondance to SQL/MM, respectively. Add a new geometry counter for collections.
Regina Obe [Tue, 23 Dec 2008 13:24:16 +0000 (13:24 +0000)]
Break out multi as separate geometry types and start putting in. Remove unary/aggregate geom since now exactly the same as section that takes only 1 geometry.
Regina Obe [Mon, 22 Dec 2008 14:14:09 +0000 (14:14 +0000)]
Revision - change to not test geoms against other geoms if function only takes one geom, better commenting of output. Put back circularstring -- crashes on 1.3 still on ST_Distance in some cases. Haven't tested against trunk
Mark Cave-Ayland [Mon, 22 Dec 2008 12:15:12 +0000 (12:15 +0000)]
Remove the few C++-style comments that have crept into the codebase, and add a note to the STYLE document stating that C-style comments should be used.
Mark Cave-Ayland [Mon, 22 Dec 2008 11:51:53 +0000 (11:51 +0000)]
Remove the WKB_CONVERSION hack which was invented to try and make the JTS connector more reliable. Since all the other JTS code has gone, this can now go too. See http://postgis.refractions.net/pipermail/postgis-devel/2008-August/003495.html for a little discussion.
Regina Obe [Thu, 18 Dec 2008 15:58:23 +0000 (15:58 +0000)]
Revise to cross every geometry with every other to better test the functions that take 2 geometries. This increases the generated script to about 18 MB. Still need to trim off some of these tests (since it does the same for singular geom functions)
Regina Obe [Thu, 18 Dec 2008 15:25:54 +0000 (15:25 +0000)]
Numerous enhancements to make torture script generator smarter - e.g. don't use geometry select when function takes not geometries, don't put monkey in when you see pattern matrix use a real intersectionmatrix, concept of version
projects / postgis / log