Do the values returned by rgeos::gCentroid() and sf::st_centroid() differ?

Question

Question

Do the values returned by rgeos::gCentroid() and sf::st_centroid() differ? If so, how?

Context

After reading the relevant commands exported by rgeos section within the r-spatial/sf wiki, I was thrilled to see that I only needed the sf package - and no longer needed to import the rgeos package - to calculate the centroid of a given geometry.

However, the use of sf::st_centroid() gave me this warning, which is addressed here:

Warning message:
   In st_centroid.sfc(x = comarea606$geometry) :
   st_centroid does not give correct centroids for longitude/latitude data

That warning led me to test for equality between the two centroid retrieval methods, just to make sure that the coordinates are the same regardless of the method.

While my use of identical() and %in% resulted in non-identical matches, all.equal() and a map plotting the centroids from each method appear to be saying these two methods are nearly identical.

Is there any reason why one method would return a different set of values than the other?

Reproducible Example

# load neccessary packages
library( sf )
library( rgeos )

# load data
comarea606 <-
    read_sf( 
      dsn = "https://data.cityofchicago.org/api/geospatial/cauq-8yn6?method=export&format=GeoJSON" 
      , layer = "OGRGeoJSON" 
    )

# find the centroid of each polygon
comarea606$centroids <-
  st_centroid( x = comarea606$geometry ) %>%
  st_geometry()

# Warning message:
#   In st_centroid.sfc(x = comarea606$geometry) :
#   st_centroid does not give correct centroids for longitude/latitude data

# Ensure the st_centroid() method
# contains identical values to gCentroid()
sf.centroids <-
  st_coordinates( x = comarea606$centroids )

rgeos.centroids <-
  gCentroid(
    spgeom = methods::as( 
      object = comarea606
      , Class = "Spatial"
    )
    , byid = TRUE 
  )@coords


# ensure the colnames are the same
colnames( rgeos.centroids ) <-
  colnames( sf.centroids )

# Test for equality
identical(
  x = sf.centroids
  , y = rgeos.centroids
) # [1] FALSE

all.equal(
  target = sf.centroids
  , current = rgeos.centroids
) # [1] TRUE

# View the first six results
head( sf.centroids )
#           X        Y
# 1 -87.61868 41.83512
# 2 -87.60322 41.82375
# 3 -87.63242 41.80909
# 4 -87.61786 41.81295
# 5 -87.59618 41.80892
# 6 -87.68752 41.97517
head( rgeos.centroids )
#           X        Y
# 1 -87.61868 41.83512
# 2 -87.60322 41.82375
# 3 -87.63242 41.80909
# 4 -87.61786 41.81295
# 5 -87.59618 41.80892
# 6 -87.68752 41.97517

# Identify the numbers which aren't identical
sf.centroids[ , "X" ] %in% rgeos.centroids[ , "X" ]
# [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [11] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [21] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [31] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [41] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [51] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [61] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [71] FALSE FALSE FALSE FALSE FALSE FALSE FALSE
sf.centroids[ , "Y" ] %in% rgeos.centroids[ , "Y" ]
# [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [11] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [21] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [31] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [41] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [51] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [61] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
# [71] FALSE FALSE FALSE FALSE FALSE FALSE FALSE

# view results
par( 
  mar = c( 2, 0, 3, 0 )
  , bg = "black"
)
plot( 
  x = comarea606$geometry
  , main = "`sf` v. `rgeos`:\nComparing Centroid Coordinates"
  , col.main = "white"
  , col = "black"
  , border = "white"
)
plot( 
  x = comarea606$centroids
  , add = TRUE
  , pch = 24
  , col = "#B4DDF2"
  , bg  = "#B4DDF2"
  , cex = 1.2
)
points( 
  x = rgeos.centroids[ , "X" ]
  , y = rgeos.centroids[ , "Y" ]
  , pch = 24
  , col = "#FB0D1B"
  , bg  = "#FB0D1B"
  , cex = 0.6
)
legend(
  x = "left"
  , legend = c( 
    "Centroids from `sf::st_coordinate()`"
    , "Centroids from `rgeos::gCentroid()`"
  )
  , col      = c( "#B4DDF2", "#FB0D1B" )
  , pt.bg    = c( "#B4DDF2", "#FB0D1B" )
  , pch      = 24
  , bty      = "n"
  , cex      = 0.9
  , text.col = "white"
)
mtext(
  adj    = 0.99
  , line = 1
  , side = 1
  , cex  = 0.9
  , text = "Source: Chicago Data Portal"
  , col  = "white"
)

# end of script #

Session Info

R version 3.4.4 (2018-03-15)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.2

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
[1] rgeos_0.3-26 sf_0.6-0    

loaded via a namespace (and not attached):
 [1] modeltools_0.2-21 kernlab_0.9-25    reshape2_1.4.3   
 [4] lattice_0.20-35   colorspace_1.3-2  htmltools_0.3.6  
 [7] stats4_3.4.4      viridisLite_0.3.0 yaml_2.1.18      
[10] utf8_1.1.3        rlang_0.2.0       R.oo_1.21.0      
[13] e1071_1.6-8       pillar_1.2.1      withr_2.1.2      
[16] DBI_0.8           prabclus_2.2-6    R.utils_2.6.0    
[19] sp_1.2-7          fpc_2.1-11        plyr_1.8.4       
[22] robustbase_0.92-8 stringr_1.3.0     munsell_0.4.3    
[25] gtable_0.2.0      raster_2.6-7      R.methodsS3_1.7.1
[28] devtools_1.13.5   mvtnorm_1.0-7     memoise_1.1.0    
[31] evaluate_0.10.1   knitr_1.20        flexmix_2.3-14   
[34] class_7.3-14      DEoptimR_1.0-8    trimcluster_0.1-2
[37] Rcpp_0.12.16      udunits2_0.13     scales_0.5.0     
[40] backports_1.1.2   diptest_0.75-7    classInt_0.1-24  
[43] squash_1.0.8      gridExtra_2.3     ggplot2_2.2.1    
[46] digest_0.6.15     stringi_1.1.6     grid_3.4.4       
[49] rprojroot_1.3-2   rgdal_1.2-18      cli_1.0.0        
[52] tools_3.4.4       magrittr_1.5      lazyeval_0.2.1   
[55] tibble_1.4.2      cluster_2.0.6     crayon_1.3.4     
[58] whisker_0.3-2     dendextend_1.7.0  MASS_7.3-49      
[61] assertthat_0.2.0  rmarkdown_1.9     rstudioapi_0.7   
[64] viridis_0.5.0     mclust_5.4        units_0.5-1      
[67] nnet_7.3-12       compiler_3.4.4   

Answer 1

This question simplifies to: how is all.equal() different from identical(), we go to the documentation for those functions:

all.equal(x, y) is a utility to compare R objects x and y testing ‘near equality’.

and for identical()

The safe and reliable way to test two objects for being exactly equal. It returns TRUE in this case, FALSE in every other case.

Lets look a little closer at all.equal.numeric(), which is what is called on these two objects, since both return "double" with typeof() . We see there is a tolerance argument in all.equal.numeric(), which is set to sqrt(.Machine$double.eps), by default. .Machine$double.eps is the smallest number that your machine can add to 1 and be able to distinguish it from 1. It's not exact, but it's on that order of magnitude. all.equal.numeric() essentially checks to see if all the values in a vector are near() all the values in another vector. You can look at the source code (which is mostly error checking) to see exactly how it does this.

To convince yourself that they are not, in fact, identical(), try looking at the output of sf.centroids - rgeos.centroids.

head(sf.centroids - rgeos.centroids)
#               X             Y
# 1 -5.056506e-10  2.623175e-09
# 2 -2.961613e-09 -4.269602e-09
# 3  4.235119e-10  4.358100e-09
# 4 -7.688925e-10 -1.051717e-09
# 5  1.226582e-09  1.668568e-10
# 6 -2.957009e-09  4.875247e-10

These two matrices are, most definitely, very nearly the same (but none of the values are exactly the same).