[R-sig-Geo] Shapes rearranged upon write

[Sam Rabin]

Lyndon — 

I couldn't figure out how to get x2 out of the function and into my workspace so I could save it as a shapefile. I instead converted your code into a script to run it, and it works like a charm. Thanks very much!

	Sam



On Nov 2, 2011, at 12:27 PM, Lyndon Estes wrote:

> Hi Sam,
> 
> I wrote the following function a while ago to account for this problem
> of dataframe row reordering relative to the polygon IDs:
> 
> joinAttributeTable <- function(x, y, xcol, ycol) {
> # Merges data frame to SpatialPolygonsDataFrame, keeping the correct
> order. Code from suggestions at:
> # https://stat.ethz.ch/pipermail/r-sig-geo/2008-January/003064.html
> # Args:
> #   x: SpatialPolygonsDataFrame
> #   y: Name of data.frame to merge
> #   xcol: Merge column name
> #   ycol: Merge column name
> # Returns: Shapefile with merged attribute table
> 
>  x$sort_id <- 1:nrow(as(x, "data.frame"))  # Column containing
> original row order for later sorting
> 	
>  x.dat <- as(x, "data.frame")  # Create new data.frame object
>  x.dat2 <- merge(x.dat, y, by.x = xcol, by.y = ycol)  # Merge
>  x.dat2.ord <- x.dat2[order(x.dat2$sort_id), ]  # Reorder back to original
> 	x2 <- x[x$sort_id %in% x.dat2$sort_id, ]  # Make new set of polygons,
> dropping those which aren't in merge
>  x2.dat <- as(x2, "data.frame")  # Make updated x2 into a data.frame
> 	row.names(x.dat2.ord) <- row.names(x2.dat)  # Reassign row.names from
> original data.frame
>  x2 at data <- x.dat2.ord  # Assign to shapefile the new data.frame
>  return(x2)
> }
> 
> I haven't used this in a while, so I hope it still works.  I think
> others could suggest a more efficient, existing method for dealing
> with this as well.
> 
> Cheers, Lyndon
> 
> 
> On Wed, Nov 2, 2011 at 12:12 PM, Sam Rabin <srabin at princeton.edu> wrote:
>> Hello!
>> 
>> I have a shapefile (.shp) containing the Landsat grid coverage of South America. I also have 97 other shapefiles, each of which covers one of the Landsat grid cells (the 97 being a subset of the cells in the first shapefile). I managed to get R (via maptools) to perform a calculation on each of the 97 and put the results in a data frame. Among other things, this data frame contains a column for grid cell ID that matches up with a column in the Landsat grid shapefile. I used merge() to combine these and then saved a new shapefile, following the instructions at http://help.nceas.ucsb.edu/R:_Spatial#Append_a_second_set_of_attributes_to_a_spatial_data_file.27s_attribute_table. Here is the code I used:
>> 
>>        require(maptools)
>>        landsat1 = readShapePoly("landsat_grid")
>>          # Make a backup
>>          landsat_orig = landsat1
>>        landsat_2_spdf = merge(landsat1 at data,newcalc,by.x="PR",by.y="pr2",all.x=TRUE,sort=FALSE)
>>        landsat1 at data = landsat_2_spdf
>>        writeSpatialShape(landsat1, "methodA")
>> 
>> Looking at the new data with
>> 
>>        View(landsat1 at data),
>> 
>> it seemed like everything went great. The ID's matched up perfectly, and I checked some of the calculated figures and they matched, too. The polygons from landsat_grid.shp that were not included in the 97 were assigned NA for the column with the calculated figures, appropriately.
>> 
>> Unfortunately, when I imported the new shapefile to my GIS program, all the tiles were rearranged. screenshot_expectedcoverage (http://tinyurl.com/3nbm3q7 ) shows fine-scale data for what I was expecting (Amazonia; imagine a polygon drawn around all the data there), and screenshot_methodA (http://tinyurl.com/3htvlga) shows what I got. (Only the 97 polygons are shown.)
>> 
>> I tried changing sort to TRUE in the fourth line (and the output shapefile name to "methodB"), but the tiles are still messed up, just differently — see screenshot_methodB (http://tinyurl.com/3h95vr3).
>> 
>> In both shape files, the attribute tables look fine, but the tiles are just totally rearranged. They're not all scattered over the world, either. In screenshot_rearrangement (which shows all tiles, not just the 97) (http://tinyurl.com/6xfabvm), you can see that they cover South America perfectly — they all line up with the polygons from landsat_grid.shp (not shown).
>> 
>> What do I need to do to fix this?
>> 
>> Thanks very much in advance.
>> 
>> 
>> 
>> Sam Rabin
>>     Graduate student
>>         Princeton University
>>         Ecology & Evolutionary Biology
>> 
>> 
>> 
>> 
>> 
>> 
>>        [[alternative HTML version deleted]]
>> 
>> 
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>> 
>>