How to replace coordinates of Polygons

Dear R geographers:

I'm able to replace the coordinates of any given polygon, but not construct 
apply() family commands or for() loops to accomplish the same task. How can I 
construct a simple loop or apply() command that will sequentially replace all 
of the coordinates in a SpatialPolygons object?

As far as I understand, the coordinates of a polygon are kept in a slot 
referred as follows:
map at polygons[[x]]@Polygons[[y]]@coords

Using which, I can create mapply(foo, indexp, indexP) commands to extract the 
original coordinates and transform them.

# Define a pair of vectors indexp, indexP that together index every Polyon in 
the SpatialPolygons object map.
# Then extract coordinates as follows:
coords <- function(x,y) {map at polygons[[indexp]]@Polygons[[indexP]]@coords}
# Or operate a function foo on the old coordinates
coords.alt <- mapply(foo, 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,1], 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,2])

So I can create lists and matrices of coordinates, and transform the 
coordinates.

It is possible to write a command that replaces coordinates using integers; 
but these commands fail when I attempt to use mapply() to replace existing 
coordinates. For example,

p at polygons[[1]]@Polygons[[1]]@coords <- foo(1,1)

will work fine, but the following will not.

newcoords <- function(indp, indP) {
   p at polygons[[indp]]@Polygons[[indP]]@coords <- foo(indp, indP)
   }
mapply(newcoords, indexp, indexP)


Have I missed something in the manual on coordinates that would enable me to 
transform the coordinates of all the Polygons in a SpatialPolygonsDataFrame 
according to some regular function of the old coordinates (x,y)? Is the only 
alternative to build a new SpatialPolygons data frame from scratch, beginning 
with the new coordinates?

If it would be useful, I can provide sample code that works on wrld_simpl.

Please advise. Thank you!

On Mon, 19 Jul 2010, Ben Mazzotta wrote:

Dear R geographers:

I'm able to replace the coordinates of any given polygon, but not 
construct apply() family commands or for() loops to accomplish the 
same task. How can I construct a simple loop or apply() command that 
will sequentially replace all of the coordinates in a SpatialPolygons 
object?

Just build lists of new Polygon then Polygons objects, and build a new 
SpatialPolygons at the end. I don't think that you'll see big memory 
problems, and the approach you propose will probably have poor 
timings, because the containing objects will probably be rebuilt for 
each update.

Are you looking for the elide() method for SpatialPolygons in 
maptools? Do you need an extra operation added to it?

Roger

As far as I understand, the coordinates of a polygon are kept in a 
slot referred as follows:
map at polygons[[x]]@Polygons[[y]]@coords

Using which, I can create mapply(foo, indexp, indexP) commands to 
extract the original coordinates and transform them.

# Define a pair of vectors indexp, indexP that together index every 
Polyon in the SpatialPolygons object map.
# Then extract coordinates as follows:
coords <- function(x,y) 
{map at polygons[[indexp]]@Polygons[[indexP]]@coords}
# Or operate a function foo on the old coordinates
coords.alt <- mapply(foo, 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,1], 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,2])

So I can create lists and matrices of coordinates, and transform the 
coordinates.

It is possible to write a command that replaces coordinates using 
integers; but these commands fail when I attempt to use mapply() to 
replace existing coordinates. For example,

p at polygons[[1]]@Polygons[[1]]@coords <- foo(1,1)

will work fine, but the following will not.

newcoords <- function(indp, indP) {
   p at polygons[[indp]]@Polygons[[indP]]@coords <- foo(indp, indP)
   }
mapply(newcoords, indexp, indexP)


Have I missed something in the manual on coordinates that would 
enable me to transform the coordinates of all the Polygons in a 
SpatialPolygonsDataFrame according to some regular function of the 
old coordinates (x,y)? Is the only alternative to build a new 
SpatialPolygons data frame from scratch, beginning with the new 
coordinates?

If it would be useful, I can provide sample code that works on 
wrld_simpl.

Please advise. Thank you!

Thank you very much, Roger.

I was afraid that might be the answer. Once I have the new Polygon and 
Polygons objects, how can I merge all the slots from the old Polygon and 
Polygons to the new Polygon and Polygons objects? Ultimately I'd like to 
bring the other fields from the old object into the new one.

The context is this: I am transforming the coordinates using the cartogram() 
function from {Rcartogram}. The argument to cartogram() is a rectangular 
matrix, which contains sample densities of a statistic at regular intervals 
over the map. Cartogram() returns a lookup table in the form of two matrices, 
which specify x_new as a function of (x_old, y_old) and y_new, again as a 
function of (x_old, y_old). After cleaning up missing observations, I can 
interpolate new coordinates for every pair of old coordinates.

Therefore, I'd like to create a new version of the world map by transforming 
every pair of coordinates in the original polygon.

On 07/19/2010 12:09 PM, Roger Bivand wrote:

On Mon, 19 Jul 2010, Ben Mazzotta wrote:

Dear R geographers:

I'm able to replace the coordinates of any given polygon, but not 
construct apply() family commands or for() loops to accomplish the same 
task. How can I construct a simple loop or apply() command that will 
sequentially replace all of the coordinates in a SpatialPolygons object?

Just build lists of new Polygon then Polygons objects, and build a new 
SpatialPolygons at the end. I don't think that you'll see big memory 
problems, and the approach you propose will probably have poor timings, 
because the containing objects will probably be rebuilt for each update.

Are you looking for the elide() method for SpatialPolygons in maptools? Do 
you need an extra operation added to it?

Roger

As far as I understand, the coordinates of a polygon are kept in a slot 
referred as follows:
map at polygons[[x]]@Polygons[[y]]@coords

Using which, I can create mapply(foo, indexp, indexP) commands to extract 
the original coordinates and transform them.

# Define a pair of vectors indexp, indexP that together index every Polyon 
in the SpatialPolygons object map.
# Then extract coordinates as follows:
coords <- function(x,y) {map at polygons[[indexp]]@Polygons[[indexP]]@coords}
# Or operate a function foo on the old coordinates
coords.alt <- mapply(foo, 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,1], 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,2])

So I can create lists and matrices of coordinates, and transform the 
coordinates.

It is possible to write a command that replaces coordinates using 
integers; but these commands fail when I attempt to use mapply() to 
replace existing coordinates. For example,

p at polygons[[1]]@Polygons[[1]]@coords <- foo(1,1)

will work fine, but the following will not.

newcoords <- function(indp, indP) {
   p at polygons[[indp]]@Polygons[[indP]]@coords <- foo(indp, indP)
   }
mapply(newcoords, indexp, indexP)


Have I missed something in the manual on coordinates that would enable me 
to transform the coordinates of all the Polygons in a 
SpatialPolygonsDataFrame according to some regular function of the old 
coordinates (x,y)? Is the only alternative to build a new SpatialPolygons 
data frame from scratch, beginning with the new coordinates?

If it would be useful, I can provide sample code that works on wrld_simpl.

Please advise. Thank you!

On Mon, 19 Jul 2010, Ben Mazzotta wrote:

Thank you very much, Roger.

I was afraid that might be the answer. Once I have the new Polygon 
and Polygons objects, how can I merge all the slots from the old 
Polygon and Polygons to the new Polygon and Polygons objects? 
Ultimately I'd like to bring the other fields from the old object 
into the new one.

If they have the same IDs, no problem, just make a 
SpatialPolygonsDataFrame with data=as(old, "data.frame").

The context is this: I am transforming the coordinates using the 
cartogram() function from {Rcartogram}. The argument to cartogram() 
is a rectangular matrix, which contains sample densities of a 
statistic at regular intervals over the map. Cartogram() returns a 
lookup table in the form of two matrices, which specify x_new as a 
function of (x_old, y_old) and y_new, again as a function of (x_old, 
y_old). After cleaning up missing observations, I can interpolate new 
coordinates for every pair of old coordinates.

Therefore, I'd like to create a new version of the world map by 
transforming every pair of coordinates in the original polygon.

Firstly, the input is a planar grid of values of a variable, so the 
queries on the two output grids need to accommodate the loss of 
resolution implied there. Try on something simple first. If you can 
get a wrapper around cartogram() to take a SpatialPolygonsDataFrame, 
make a grid from it for the chosen variable (including NA handling), 
then do cartogram, followed by interpolation for Polygons object by 
Polygons object (remember objects may be multiple Polygon objects, and 
may have holes), you may be able to get there. Constructing the output 
objects from computed coordinates looks like the least of your 
challenges.

Good luck!

Roger

On 07/19/2010 12:09 PM, Roger Bivand wrote:

On Mon, 19 Jul 2010, Ben Mazzotta wrote:

Dear R geographers:

I'm able to replace the coordinates of any given polygon, but not 
construct apply() family commands or for() loops to accomplish the 
same task. How can I construct a simple loop or apply() command 
that will sequentially replace all of the coordinates in a 
SpatialPolygons object?

Just build lists of new Polygon then Polygons objects, and build a 
new SpatialPolygons at the end. I don't think that you'll see big 
memory problems, and the approach you propose will probably have 
poor timings, because the containing objects will probably be 
rebuilt for each update.

Are you looking for the elide() method for SpatialPolygons in 
maptools? Do you need an extra operation added to it?

Roger

As far as I understand, the coordinates of a polygon are kept in a 
slot referred as follows:
map at polygons[[x]]@Polygons[[y]]@coords

Using which, I can create mapply(foo, indexp, indexP) commands to 
extract the original coordinates and transform them.

# Define a pair of vectors indexp, indexP that together index every 
Polyon in the SpatialPolygons object map.
# Then extract coordinates as follows:
coords <- function(x,y) 
{map at polygons[[indexp]]@Polygons[[indexP]]@coords}
# Or operate a function foo on the old coordinates
coords.alt <- mapply(foo, 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,1], 
map at polygons[[indexp]]@Polygons[[indexP]]@coords[,2])

So I can create lists and matrices of coordinates, and transform 
the coordinates.

It is possible to write a command that replaces coordinates using 
integers; but these commands fail when I attempt to use mapply() to 
replace existing coordinates. For example,

p at polygons[[1]]@Polygons[[1]]@coords <- foo(1,1)

will work fine, but the following will not.

newcoords <- function(indp, indP) {
   p at polygons[[indp]]@Polygons[[indP]]@coords <- foo(indp, indP)
   }
mapply(newcoords, indexp, indexP)


Have I missed something in the manual on coordinates that would 
enable me to transform the coordinates of all the Polygons in a 
SpatialPolygonsDataFrame according to some regular function of the 
old coordinates (x,y)? Is the only alternative to build a new 
SpatialPolygons data frame from scratch, beginning with the new 
coordinates?

If it would be useful, I can provide sample code that works on 
wrld_simpl.

Please advise. Thank you!