fosslc/ingres

Ingres is a feature rich and robust database (RDBMS)

ingres's Issues

Be aware of Endian differences in spatial data.

The internal spatial data will have to be Endian aware once differing platforms are supported.

Geospatial causes diffs in SEP

The following files have diffs:

• star/phase1/ddl/sta02.log
• star/phase1/ddl/sta03.log
• star/phase1/ddl/sta01.log
• be/datatypes/dt07.log
• be/datatypes/dtu63.log
• be/datatypes/dtu29.log
• be/miscfunc/msfn029.log
• be/lar/lar66.log
• be/lar/lar19.log
• be/access/am85.log
• be/util/utl06.log
• be/util/utl02.log
• be/util/utl07.log
• be/util/utl08.log
• be/util/utl04.log
• be/util/utl05.log
• be/util/utl03.log
• fe/local/copy/cpy29.log
• fe/local/vision/vis35.log
• fe/local/vision/vis34.log

Disable SRID handling for copydb when geospatial is off

From Joe's email:

Hi,

I ran copydb from a 10.1.0 build 109 from int.lnx to usilsuds (2.6/0803)
and the copy.in file has "SRID 0" for each column, which I presume is
some Geospatial incantation:

create table abbjo03_bf(
s_bug_no integer SRID 0 not null default 0,
starprob varchar(10) SRID 0 not null default ' ',
bug_no integer SRID 0 not null default 0,
severity varchar(4) SRID 0 not null default ' ',
part varchar(20) SRID 0 not null default ' ',
category varchar(20) SRID 0 not null default ' ',
bug_sir varchar(1) SRID 0 not null default ' ',
descrip varchar(480) SRID 0 not null default ' '
)

I turned off GEO in my build so the DBMS doesn't like the extra specs,
but it seems to me they shouldn't be there in the first place.

Joe

Default to -nosrid if conf_WITH_GEO is not defined

This is a minor enhancement to the build, but worthwhile adding to our list of to-do's if we have time. The enhancement would be to flip -nosrid to default to on if the builder didn't build with geospatial enabled.

Since Geospatial support is part of core, this is an corner case for a minority of people so it's not urgent.

Correctly calculate length for all geometries where possible

Currently polygons cannot return a length. Fix this and check other types to return lengths where possible.

Make further us of GEOM family types

There are quite a few places where all spatial types are being checked. If possible, change them to check for the family type.

Passing along GEOS error messages

Currently when there is an error in GEOS the message is just written to the DBMS log, and there is no way to pass this to the user. The reason for this being that you have to register a geos_Error function when initializing GEOS. This error function can't take an ADF_CB as a parameter, hence we can't pass this error message along to the user. Ideally we would like to be able to do this.

rtree exact text match for applicable functions is limiting

Only the overlaps, inside and intersects functions will work with rtree. And only exact matches, e.g. st_overlaps will not use rtree.

Make range optional in rtree creation statement.

Since it is not currently being used.

iilink does not use -rpath properly

The problem is that fastload executes the dmfjsp executable which is a setuid program owned by the ingres userid. The tests are run using the testenv userid. In newer Linux releases the LD_LIBRARY_PATH is dropped when executing setuid programs as a different userid. Therefore libgeos-3.3.0.so can't be found.

The solution is to specify the library using a relative path specification and using the -rpath link option.

The iimerge executable is created using the iilink script. The iilink script is designed to create either an executable where all libraries are static, or all libraries are shared libraries. In the last release the iimerge was changed to use all static libraries to create the iimerge. The geospatial change was added to an iimerge which was not linked with the necessary -rpath flags because it was expecting all static libraries.

Geocoding support for Ingres

This ticket tracks work to add geocoding support for Ingres.

add spatial_ref_sys to main_catalog_list

From Karl Schendel:
spatial_ref_sys needs to be added to the "main_catalog_list"
array in upgradedb.qsc.

This governs handling for the table during an upgrade - for instance if the table definition is different or it doesn't exist yet.

Raster support

This ticket is to track adding raster support to Ingres. More details will follow as we develop the feature.

Compiling GDAL/OGR with Ingres can't find Proj.4

Track down why Proj.4 can't be found when compiling GDAL/OGR --with-ingres in certain circumstances.

Sep tests geo11, geo13 and geo14 have diffs

Most of these diffs are due to floating point rounding errors on different platforms due to the large precision used in astextround. Investigate PointOnSurface that produces a complete different result in geo13.log

Mapserver support for Ingres

This ticket is to track work to add a native mapserver driver for Ingres.

Investigate issues with rpm install on systems without geos/proj.4

The expected behaviour is that we'll add geos and proj.4 if they aren't present. For now, geos with our changes is not available as a binary for most distributions so we're packaging the library. Thus we shouldn't see issues.

Be aware of Endian differences in spatial data.

The internal spatial data will have to be Endian aware once differing platforms are supported.

Geos and proj in non /usr/local location results in link failure.

From Kristoff Picard via Techinfo mailing list.

I did not install the Geos and Proj.4 libraries in /usr/local, but in some private directories and set the needed varaiables to point to the libraries and headers.(as described in the wiki)
This worked fine to get everything compiled, but when iilink is executed it doesn't find these libraries.
The LDLIB... lines in iisysdep contain "-lgeos -lgeos_c -lproj" , but we would need an extra -L definitions as these libraries are not installed in one of the standard locations.
I ran the link command manually to get it running.

Routing support in Ingres

This ticket tracks work to add routing support to Ingres.

Equivalence class errors in some spatial functions.

There seems to be a problem in the Intersects and Overlaps functions. select count(*) from areawater_merge a, edges_merge e where Intersects(e.shape,a.shape)=1 and e.ogr_fid=3871598\g Executing . . . E_OP0487 consistency check

no joining equivalence class found when expected

However, Intersects return value is 0: select count(*) from areawater_merge a, edges_merge e where Intersects(e.shape,a.shape)=0 and e.ogr_fid=3871598\g The same goes for the Overlaps function. On the other hand, function Contains works with both 1 and 0.

select lu.property_i from land_use_2006 as lu, s_fld_haz_ar fz where lu.general_la in (100,113,150,160,200) and (fld_zone='A' or fld_zone='AE' or fld_zone='AO' or fld_zone='V') and overlaps(lu.shape, fz.shape)=1 \g Executing . . . E_OP0889 Eqc is not available at a CO node.

However, when changing Overlaps(x,y) = 0: select lu.property_i from land_use_2006 as lu, s_fld_haz_ar fz where lu.general_la in (100,113,150,160,200) and (fld_zone='A' or fld_zone='AE' or fld_zone='AO' or fld_zone='V') and overlaps(lu.shape, fz.shape)=0 \g