readOGR and PostGIS: "Error getting field 0"

Dear Roger,

Thanks a lot for your very fast answer! Here are the requested outputs:

Le 30/11/2012 10:39, Roger Bivand a ?crit :

On Fri, 30 Nov 2012, Mathieu Basille wrote:

Dear R users,

I have a working PostGIS database that I routinely access through R,
using readOGR or readGDAL. I have this point layer, however, that I can't
read in R:

ctrdcoast <- readOGR("PG:dbname=florida host=localhost user=pguser", 
layer

= "gis.world_ctrdcoast")
OGR data source with driver: PostgreSQL
Source: "PG:dbname=florida host=localhost user=pguser", layer:
"gis.world_ctrdcoast"
with 35398 features and 2 fields
Error in readOGR("PG:dbname=florida host=localhost user=pguser", layer =
"gis.world_ctrdcoast") :
 Error getting field 0

The layer is basically a table with just 2 fields: geom (a point feature)
and gid (a serial primary key). I can export this layer without any
problem using ogr2ogr, and the layer looks fine to me (it displays fine
in QGIS):

ogr2ogr -f "ESRI Shapefile" ctrd.shp PG:"dbname=florida host=localhost
user=pguser" "gis.world_ctrdcoast"

Any idea where this is going wrong? Thanks in advance for any suggestion.
Mathieu.

The error message is from ogrReadColumn(), line 221 in src/ogrsource.cpp,
for example here:

https://r-forge.r-project.org/scm/viewvc.php/pkg/src/ogrsource.cpp?view=markup&root=rgdal


What does ogrInfo() report?

$ ogrinfo "PG:dbname=florida host=localhost user=pguser" gis.world_ctrdcoast 
-summary
INFO: Open of `PG:dbname=florida host=localhost user=pguser'
     using driver `PostgreSQL' successful.

Layer name: gis.world_ctrdcoast
Geometry: Unknown (any)
Feature Count: 35398
Extent: (-179.916500, -59.083000) - (179.968500, 83.535000)
Layer SRS WKT:
(unknown)
FID Column = gid
Geometry Column = geom

I noticed there was a call to ogrInfo in the readOGR function. After this 
line, here is what's reported:

Browse[2]> ogr_info
Source: "PG:dbname=florida host=localhost user=pguser", layer: 
"gis.world_ctrdcoast"
Driver: PostgreSQL number of rows 35398
Feature type: wkbPoint with 2 dimensions
NA
Number of fields: 0
[1] name     type     length   typeName
<0 rows> (or 0-length row.names)

Does it make a difference if you add an extra
field?

As a matter of fact, it does! If I add a serial field named 'bla', it works 
perfectly:

ctrdcoast <- readOGR("PG:dbname=florida host=localhost user=pguser", layer 

= "gis.world_ctrdcoast")
OGR data source with driver: PostgreSQL
Source: "PG:dbname=florida host=localhost user=pguser", layer: 
"gis.world_ctrdcoast"
with 35398 features and 1 fields
Feature type: wkbPoint with 2 dimensions

summary(ctrdcoast)

Object of class SpatialPointsDataFrame
Coordinates:
               min      max
coords.x1 -179.9165 179.9685
coords.x2  -59.0830  83.5350
Is projected: NA
proj4string : [NA]
Number of points: 35398
Data attributes:
  Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
     1    8850   17700   17700   26550   35400

head(ctrdcoast at data)

 bla
1   1
2   2
3   3
4   4
5   5
6   6

I'm wondering whether in this case the GID is not being seen as a
field as such, but the FID of the point. Possibly ogr2ogr steps round this.

It looks like a correct guess, given the output of ogrinfo. In addition, the 
exported shapefile from ogr2ogr present one field (named FID), while if I 
load the table into QGIS, I can see the field with its correct name (gid). 
Admittedly, while I'm pretty sure that QGIS uses GDAL to load tables, I have 
no idea how it deals with it.

I don't have a PostGIS system running, so may need help debugging. If you
debug(readOGR), what happens here:

         if (drop_unsupported_fields) {
              iflds <- as.integer((1:ogr_info$nitems)-1)
              iflds <- iflds[keep]
              fldnms <- ogr_info$iteminfo$name[keep]
              if (any(!keep)) warning(paste("Fields dropped:",
                  paste(ogr_info$iteminfo$name[!keep], collapse=" ")))
         } else {
              if (any(!keep)) stop(paste("Unsupported field type:",
                  paste(ogr_info$iteminfo$typeName[!keep], collapse=" ")))
              iflds <- as.integer((1:ogr_info$nitems)-1)
              fldnms <- ogr_info$iteminfo$name
         }

Browse[2]> ogr_info$nitems
[1] 0
Browse[2]> iflds
[1]  0 -1
Browse[2]> fldnms
character(0)

I think ogr_info$nitems is 1, so we get 0, but if ogr_info$nitems is 0, (0,
-1).

Does that mean that we necessarily need a field in the table (in addition to 
geom + gid)? It's not hard to 'fix' by adding a nonsensical field, but I find 
it quite counterintuitive.

Thanks again for the help!
Mathieu.

Roger

sessionInfo()

R version 2.15.1 (2012-06-22)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
[1] LC_CTYPE=fr_FR.UTF-8       LC_NUMERIC=C LC_TIME=fr_FR.UTF-8
[4] LC_COLLATE=fr_FR.UTF-8     LC_MONETARY=fr_FR.UTF-8
LC_MESSAGES=fr_FR.UTF-8
[7] LC_PAPER=C                 LC_NAME=C                  LC_ADDRESS=C
[10] LC_TELEPHONE=C             LC_MEASUREMENT=fr_FR.UTF-8
LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] raster_2.0-31 rgdal_0.7-22  sp_1.0-2

loaded via a namespace (and not attached):
[1] compiler_2.15.1 grid_2.15.1     lattice_0.20-6  tools_2.15.1


# SELECT PostGIS_Full_Version();

   postgis_full_version
-------------------------------------------------------------------------
POSTGIS="2.1.0SVN r10597" GEOS="3.3.3-CAPI-1.7.4" PROJ="Rel. 4.7.1, 23
September 2009" GDAL="GDAL 2.0dev, released 2011/12/29" LIBXML="2.8.0"
LIBJSON="UNKNOWN" TOPOLOGY RASTER