[R-sig-Geo] matching does not work in SpatialPointsDataFrame

[Jean-Luc Dupouey]

I get an incorrect match between row names of coordinates and data in 
SpatialPointsDataFrame:

 > Mycoord <- data.frame(row.names=letters[1:5], x=rnorm(5), y=rnorm(5))
 >
 > Mycoord
             x          y
a -0.80138451  1.9291882
b -1.58229979 -0.7713053
c -1.28416923  1.3324122
d -0.22558639  0.7320874
e -0.08702876  1.7635046
 >
 > Myattr <- data.frame(row.names=sample(letters[1:5]),var1=rnorm(5))
 >
 > Myattr
         var1
d  1.2935138
c  0.1143173
a -0.2157230
b -1.9424786
e -0.7234636
 >
 > MyptsdataT <- SpatialPointsDataFrame(Mycoord, Myattr, match.ID=TRUE)
 >
 > MyptsdataT
               coordinates       var1
d  (-0.8013845, 1.929188)  1.2935138
c   (-1.5823, -0.7713053)  0.1143173
a   (-1.284169, 1.332412) -0.2157230
b (-0.2255864, 0.7320874) -1.9424786
e (-0.08702876, 1.763505) -0.7234636

Either I made a mistake somewhere (99.99%), either there is a bug (0.01%).

I read carefully and several times the help page for this function, and 
did not find any clue to my problem. It is written: "if (match.ID=)TRUE 
AND coords has rownames (i.e., coerced to a matrix, 
dimnames(coords)[[2]] is not NULL), AND data has row.names (i.e. is a 
data.frame), then the SpatialPointsDataFrame object is formed by 
matching the row names of both components".

It seems to me that my coordinates, data and parameters fulfill these 
requirements (even if these requirements look a bit strange: shouldn't 
it be dimnames(coords)[[1]] which must not be NULL?).

A quick look at the function SpatialPointsDataFrame suggests me that the 
problem is perhaps coming from the line:

if (!is(coords, "SpatialPoints"))  coords = coordinates(coords)

When coords is a dataframe, it returns a matrix with the rownames lost. 
Subsequent matching with data is not longer possible. Should'nt the test 
be reversed: if (is(coords, "SpatialPoints")) coords = coordinates(coords) ?

I am working with recent versions of R (3.1.2 (2014-10-31)) and sp (1.0-17).

Thanks for any help towards the resolution of this problem.

Jean-Luc Dupouey

-- 
INRA
Forest Ecology and Ecophysiology Unit
F-54280 Champenoux
France

mail: dupouey at nancy.inra.fr