move POINT geometries out of a list column, into a matrix #2059

edzer · 2022-12-13T20:29:30Z

This PR moves POINT geometries out of the list column, leaves a list column filled with NULL values in place, and moves the point coordinates in an n x 2 (or 3 for XYM/XYZ, or 4 for XYZM) matrix attribute called points. The class dim label (XY/XYZ/XYM/XYZM) is put in attribute points_dim. This happens when:

st_as_sf.data.frame is called with e.g. coords specifying the coordinate colums
WKB in converted into an sfc, i.e. when coming back from st_read() or any of the GEOS functions.
Right now, a[] restores the original list of sfg objects:

library(sf)
# Linking to GEOS 3.10.2, GDAL 3.4.3, PROJ 8.2.1; sf_use_s2() is TRUE
data.frame(z=rnorm(3), x=1:3, y=3:1) |>
  st_as_sf(coords = c("x", "y")) -> x
x$geometry |>
  unclass() |> 
  str()
# List of 3
#  $ : NULL
#  $ : NULL
#  $ : NULL
#  - attr(*, "points")= num [1:3, 1:2] 1 2 3 3 2 1
#  - attr(*, "points_dim")= chr "XY"
#  - attr(*, "n_empty")= int 0
#  - attr(*, "precision")= num 0
#  - attr(*, "crs")=List of 2
#   ..$ input: chr NA
#   ..$ wkt  : chr NA
#   ..- attr(*, "class")= chr "crs"
#  - attr(*, "bbox")= 'bbox' Named num [1:4] 1 1 3 3
#   ..- attr(*, "names")= chr [1:4] "xmin" "ymin" "xmax" "ymax"
x$geometry[] |>
  unclass() |> 
  str()
# List of 3
#  $ : 'XY' num [1:2] 1 3
#  $ : 'XY' num [1:2] 2 2
#  $ : 'XY' num [1:2] 3 1
#  - attr(*, "n_empty")= int 0
#  - attr(*, "precision")= num 0
#  - attr(*, "crs")=List of 2
#   ..$ input: chr NA
#   ..$ wkt  : chr NA
#   ..- attr(*, "class")= chr "crs"
#  - attr(*, "bbox")= 'bbox' Named num [1:4] 1 1 3 3
#   ..- attr(*, "names")= chr [1:4] "xmin" "ymin" "xmax" "ymax"

This may give some performance changes, both in memory footprint and runtime; still need to do benchmarks.

Comments welcome!

edzer · 2022-12-13T20:55:36Z

Memory size: decreases with a factor 18:

library(sf)
# Linking to GEOS 3.10.2, GDAL 3.4.3, PROJ 8.2.1; sf_use_s2() is TRUE
n = 1e6
data.frame(z = rnorm(n), x = runif(n), y = runif(n)) -> df
st_as_sf(df, coords = c("x", "y")) |>
  st_geometry() -> pts
(s1 = object.size(pts))
# 24002864 bytes
(s2 = object.size(pts[]))
# 432002312 bytes
as.numeric(s2)/as.numeric(s1)
# [1] 17.99795

edzer · 2022-12-13T22:20:21Z

This one is pretty obvious:

library(sf)
# Linking to GEOS 3.10.2, GDAL 3.4.3, PROJ 8.2.1; sf_use_s2() is TRUE
n = 1e6
data.frame(z = rnorm(n), x = runif(n), y = runif(n)) -> df
st_as_sf(df, coords = c("x", "y")) |>
  st_geometry() -> pts
pts   |> st_is_empty() |> system.time()
#    user  system elapsed 
#   1.692   0.072   1.764 
system.time(pts0 <- pts[])
#    user  system elapsed 
#   1.104   0.036   1.140 
pts0 |> st_is_empty()  |> system.time()
#    user  system elapsed 
#   1.853   0.057   1.909 
pts |> st_combine() |> system.time()
#    user  system elapsed 
#   0.002   0.000   0.002 
pts0 |> st_combine() |> system.time()
#    user  system elapsed 
#   3.119   0.155   3.275

paleolimbot · 2022-12-14T02:38:33Z

This is a very cool idea! The use of an attribute to maintain a semblance of backward compatibility with the structure is clever.

I do wonder if this change is going to result in enough performance gains to justify the maintenance cost of the additional representation. My adventures playing with this branch are below...basically all the ways I know about based on the benchmarks above. wk::xy() is more like data.frame(x, y, z) than what you have here...basically just because it doesn't require the matrix attribute shenanigan to work with existing data frames. The GEOS representation is a list() of external pointers to GEOSGeometry(). Geoarrow isn't quite ready for this yet but its kernels will probably be faster than both of those.

While I certainly don't mind if this representation exists in sf, my first impression is that it's rather complicated and not that fast. If performance is the motivation, I wonder if broadening the sf API to work with other S3 classes (e.g., wk::xy(), geos::geos_geometry(), whatever terra has on the go). Functionally that mostly means just making more things S3 generic (then people like me can provide fast implementations in wk and geos without you having to do anything!). Slightly harder would be loosening that assumption for the sf geometry column.

Another low-hanging performance fruit would be to drop the cached bbox and geometry classes (that's most of why the geos package is faster than sf in most cases).

library(sf)
#> Linking to GEOS 3.11.1, GDAL 3.6.0, PROJ 9.1.1; sf_use_s2() is TRUE

n = 1e5
data.frame(z = rnorm(n), x = runif(n), y = runif(n)) -> df
st_as_sf(df, coords = c("x", "y")) |>
    st_geometry() -> pts

pts0 <- pts[]
wk_xy <- wk::as_xy(df)
geos_pts <- geos::as_geos_geometry(wk_xy)

lobstr::obj_size(pts)
#> 2.40 MB
lobstr::obj_size(pts0)
#> 12.80 MB
lobstr::obj_size(wk_xy)
#> 2.40 MB
# not really accurate...only includes R allocs
lobstr::obj_size(geos_pts)
#> 7.20 MB

bench::mark(
    st_is_empty(pts),
    st_is_empty(pts0),
    wk::wk_meta(wk_xy)$size == 0,
    wk::wk_meta(geos_pts)$size == 0,
    geos::geos_is_empty(geos_pts)
)
#> # A tibble: 5 × 6
#>   expression                           min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>                      <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 st_is_empty(pts)                 33.83ms  34.48ms      28.9    1.16MB     12.8
#> 2 st_is_empty(pts0)                46.59ms   46.8ms      21.4    1.15MB     49.8
#> 3 wk::wk_meta(wk_xy)$size == 0      2.06ms   2.25ms     421.      3.1MB     53.9
#> 4 wk::wk_meta(geos_pts)$size == 0   4.21ms   4.43ms     221.     3.06MB     31.1
#> 5 geos::geos_is_empty(geos_pts)   756.29µs 840.21µs    1180.   406.18KB     22.7

bench::mark(
    st_combine(pts),
    st_combine(pts0),
    wk::wk_collection(wk_xy, wk::wk_geometry_type("multipoint")),
    geos::geos_make_collection(geos_pts, "multipoint"),
    check = FALSE
)
#> Warning: Some expressions had a GC in every iteration; so filtering is disabled.
#> # A tibble: 4 × 6
#>   expression                                                        min   median
#>   <bch:expr>                                                   <bch:tm> <bch:tm>
#> 1 st_combine(pts)                                               97.99µs 100.61µs
#> 2 st_combine(pts0)                                             139.98ms 181.44ms
#> 3 wk::wk_collection(wk_xy, wk::wk_geometry_type("multipoint"))   1.85ms   2.03ms
#> 4 geos::geos_make_collection(geos_pts, "multipoint")             6.09ms   6.51ms
#> # … with 3 more variables: `itr/sec` <dbl>, mem_alloc <bch:byt>, `gc/sec` <dbl>

rsbivand

This makes the POINT object (XY and XYZ) more like sp, but:

> data.frame(z=rnorm(3), x=1:3, y=3:1) |> st_as_sf(coords = c("x", "y")) -> x
> as(x$geometry[], "Spatial")
SpatialPoints:
     coords.x1 coords.x2
[1,]         1         3
[2,]         2         2
[3,]         3         1
Coordinate Reference System (CRS) arguments: NA 
> as(x$geometry, "Spatial")
Error in (function (classes, fdef, mtable)  : 
  unable to find an inherited method for function 'coordinates' for signature '"NULL"'
> as(x, "Spatial")
Error in h(simpleError(msg, call)) : 
  error in evaluating the argument 'x' in selecting a method for function 'addAttrToGeom': unable to find an inherited method for function 'coordinates' for signature '"NULL"'
> sp_x <- as(x$geometry[], "Spatial")
> st_as_sfc(sp_x)
Geometry set for 3 features 
Geometry type: POINT
Dimension:     XY
Bounding box:  xmin: 1 ymin: 1 xmax: 3 ymax: 3
CRS:           NA
POINT (1 3)
POINT (2 2)
POINT (3 1)
> st_as_sfc(sp_x) |> unclass() |> str()
List of 3
 $ : 'XY' num [1:2] 1 3
 $ : 'XY' num [1:2] 2 2
 $ : 'XY' num [1:2] 3 1
 - attr(*, "precision")= num 0
 - attr(*, "bbox")= 'bbox' Named num [1:4] 1 1 3 3
  ..- attr(*, "names")= chr [1:4] "xmin" "ymin" "xmax" "ymax"
 - attr(*, "crs")=List of 2
  ..$ input: chr NA
  ..$ wkt  : chr NA
  ..- attr(*, "class")= chr "crs"
 - attr(*, "n_empty")= int 0

edzer · 2022-12-14T11:23:35Z

@rsbivand thanks, that should now work; I finished implementation so far only to pass sf's own tests, this was not in there.

bart1 · 2022-12-22T08:42:15Z

I just encountered an other case where this pull request could provide a big speedup. When using print.sfc it can be quite slow (20 seconds for 10^7 points) as the dimensions are calculated in an apply function: https://github.com/r-spatial/sf/blob/main/R/sfc.R#L195 . If this information can directly be taken from the attribute points_dim this should speed up considerably

edzer · 2022-12-22T08:53:21Z

... which is rather stupid (the 20 seconds), given that #2046 suggests sf reading breaks (unintended) in case of mixed dimensions...

kadyb · 2023-02-28T22:05:03Z

Today ALTREP support for lists appeared in R SVN. Do you think it can somehow affect {sf} if it uses lists as data structure?

edzer · 2023-02-28T22:12:28Z

Thanks for reporting! I'm not an altrep expert, but would expect no miracles; I think it's mostly useful for (i) data structures out of memory (e.g. in a memory mapped area) or (ii) unrealised arrays (e.g. 1:1e12, where you're going to compute the value when you need it rather than realize it in memory). The first would help with massive point sets (but merely make it possible, not fast), the second doesn't help us here.

The speedup of this PR (if any) is due to the flat memory model of matrices vs. the arbitrary memory layout of lists; accessing lists elements is more expensive because of that; the smaller memory footprint may be even more important.

paleolimbot · 2023-03-30T19:47:33Z

FWIW I actually think ALTREP for lists will be very useful here. It will let you use the data structure you've prototyped here without breaking backwards compatibility with users of the previous data structure (i.e., everybody).

edzer · 2023-03-30T20:22:35Z

Great idea, do you know of an example of such a thing?

edzer · 2023-06-29T18:52:52Z

Great example; that should run better, now.

bart1 · 2023-06-30T14:16:33Z

Thanks @edzer , I tested a bit more and noticed two other examples changing for unexpected behavior. Sorry for having these issues iteratively, the popup up in some of the unit test. Let me also know if your not looking for these detailed tests yet

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
require(dplyr)
#> Loading required package: dplyr
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
n <- 4
data.frame(z = rnorm(n), x = runif(n), y = runif(n)) -> df
dfsf<-st_as_sf(df, coords = c("x", "y"))
# The filtered data turns empty:
dfsf |> filter(rep(T,4))
#> Simple feature collection with 4 features and 1 field (with 4 geometries empty)
#> Geometry type: GEOMETRY
#> Dimension:     XY
#> Bounding box:  xmin: NA ymin: NA xmax: NA ymax: NA
#> CRS:           NA
#>             z                 geometry
#> 1 -0.32671172 GEOMETRYCOLLECTION EMPTY
#> 2 -0.05574357 GEOMETRYCOLLECTION EMPTY
#> 3 -1.34424667 GEOMETRYCOLLECTION EMPTY
#> 4 -0.24700773 GEOMETRYCOLLECTION EMPTY
# there is no empty point inserted
dfsf$geometry[2]<-st_point()
dfsf
#> Simple feature collection with 4 features and 1 field (with 4 geometries empty)
#> Geometry type: POINT
#> Dimension:     XY
#> Bounding box:  xmin: 0.02291049 ymin: 0.2227787 xmax: 0.6861906 ymax: 0.4822213
#> CRS:           NA
#>             z                     geometry
#> 1 -0.29728105  POINT (0.3049286 0.4746749)
#> 2 -0.33397322  POINT (0.5350945 0.2988531)
#> 3 -0.93209304  POINT (0.6861906 0.4822213)
#> 4  0.08929689 POINT (0.02291049 0.2227787)
sessioninfo::session_info()
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.1.2 (2021-11-01)
#>  os       Ubuntu 22.04.2 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Europe/Amsterdam
#>  date     2023-06-30
#>  pandoc   2.19.2 @ /usr/lib/rstudio/resources/app/bin/quarto/bin/tools/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package     * version date (UTC) lib source
#>  ...
#>  dplyr       * 1.1.2   2023-04-20 [1] CRAN (R 4.1.2)
#>  ...
#>  sf          * 1.0-14  2023-06-29 [1] Github (r-spatial/sf@3ec76b8)
#>  ...
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.1
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

edzer · 2023-06-30T18:50:54Z

This should now also run.

bart1 · 2023-06-30T20:28:57Z

That works for the examples above but filter fails if there is an empty location:

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
n <- 4
data.frame(z = rnorm(n), x = runif(n), y = runif(n)) -> df
dfsf<-st_as_sf(df, coords = c("x", "y"))
dfsf$geometry[2]<-st_point()
dfsf |> dplyr::filter(rep(T,4))
#> Error in st_as_sf.data.frame(NextMethod(), coords = attr(.data, "sf_column"), : missing values in coordinates not allowed

I also see a few other changes in behavior/test error but I have not been able to make small reproducible examples out of those yet

bart1 · 2023-07-02T21:26:24Z

It seems dplyr filter drops the CRS:

require(dplyr)
require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
data(meuse, package = "sp")
meuse_sf = st_as_sf(meuse, coords = c("x", "y"), crs = 28992, agr = "constant")
st_crs(meuse_sf)
#> Coordinate Reference System:
#>   User input: EPSG:28992 
#>   wkt:
....
#>         BBOX[50.75,3.2,53.7,7.22]],
#>     ID["EPSG",28992]]
st_crs(dplyr::filter(meuse_sf, copper>80))
#> Coordinate Reference System: NA
sessioninfo::session_info()
#> ─ Session info ───────────────────────────────────────────────────────────────
.....
#>  sf          * 1.0-14  2023-07-01 [1] Github (r-spatial/sf@203187d)
.....
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.1
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

^{Created on 2023-07-02 with reprex v2.0.2}

address #2059 (comment)

bart1 · 2024-01-22T14:08:36Z

Hi, I have been doing some more testing. In general I use the test suite from move2 (most test/functions use sf in the background in some way) to find these changes from before (for the move2 package this pull request would be quite beneficial). Then I try to isolate these errors to sf workflows. Let me know there is no desire to merge this request then I will stop testing.

I have found a few more cases that show some strange/undesired? behavior.

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
d<-st_as_sf(data.frame(x=c(1:4), y=c(1:3,NA)), coords=c("x","y"), na.fail = F)
st_is_empty(st_geometry(d))
#> [1] FALSE FALSE FALSE FALSE
st_is_empty(d) 
#> [1] FALSE FALSE FALSE FALSE
# should be c(F,F,F, T)?, now is c(F,F,F,F)

dput(st_geometry(d))
#> structure(list(NULL, NULL, NULL, NULL), points = structure(c(1, 
#> 2, 3, 4, 1, 2, 3, NA), dim = c(4L, 2L)), points_dim = "XY", n_empty = 0L, precision = 0, crs = structure(list(
#>     input = NA_character_, wkt = NA_character_), class = "crs"), bbox = structure(c(xmin = 1, 
#> ymin = 1, xmax = 4, ymax = 3), class = "bbox"), class = c("sfc_POINT", 
#> "sfc"))
dput(st_geometry(d[1:4,])) 
#> structure(list(structure(c(1, 1), class = c("XY", "POINT", "sfg"
#> )), structure(c(2, 2), class = c("XY", "POINT", "sfg")), structure(c(3, 
#> 3), class = c("XY", "POINT", "sfg")), structure(c(4, NA), class = c("XY", 
#> "POINT", "sfg"))), n_empty = 0L, precision = 0, crs = structure(list(
#>     input = NA_character_, wkt = NA_character_), class = "crs"), bbox = structure(c(xmin = 1, 
#> ymin = 1, xmax = 4, ymax = 3), class = "bbox"), class = c("sfc_POINT", 
#> "sfc"))
# sub setting directly reduces to the old structure, keeping the matrix might be nicer and makes sure the benefits are throughout scripts

e<-st_as_sf(data.frame(x=c(1:3, NA), y=c(1:3,NA)), coords=c("x","y"), na.fail = F)

st_is_empty(st_geometry(e)) 
#> [1] FALSE FALSE FALSE  TRUE
# works correctly
st_is_empty(st_geometry(dplyr::filter(e, rep(T,4)))) 
#> Error in scan(text = lst[[length(lst)]], quiet = TRUE): scan() expected 'a real', got 'ParseException:'
#> Error in (function (msg) : ParseException: Unexpected EOF parsing WKB
# fails,  one difference I see is that a Z dimension is added to points_dim.
sessioninfo::session_info(pkgs = "attached")
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.3.2 (2023-10-31)
#>  os       Ubuntu 22.04.3 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Europe/Amsterdam
#>  date     2024-01-22
#>  pandoc   3.1.1 @ /usr/lib/rstudio/resources/app/bin/quarto/bin/tools/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package * version date (UTC) lib source
#>  sf      * 1.0-14  2024-01-22 [1] Github (r-spatial/sf@d38b3dc)
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.3
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

^{Created on 2024-01-22 with reprex v2.0.2}

bart1 · 2024-01-23T09:38:27Z

Another small change/corner case casting and empty sfc fails:

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
st_cast(st_sfc(),"POINT")
#> Error in cls[2, ]: incorrect number of dimensions
sessioninfo::session_info("attached")
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.3.2 (2023-10-31)
#>  os       Ubuntu 22.04.3 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Europe/Amsterdam
#>  date     2024-01-23
#>  pandoc   3.1.1 @ /usr/lib/rstudio/resources/app/bin/quarto/bin/tools/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package * version date (UTC) lib source
#>  sf      * 1.0-14  2024-01-22 [1] Github (r-spatial/sf@d38b3dc)
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.3
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

This used to work fine:

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
st_cast(st_sfc(),"POINT")
#> Geometry set for 0 features 
#> Bounding box:  xmin: NA ymin: NA xmax: NA ymax: NA
#> CRS:           NA
sessioninfo::session_info("attached")
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.3.2 (2023-10-31)
#>  os       Ubuntu 22.04.3 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Europe/Amsterdam
#>  date     2024-01-23
#>  pandoc   3.1.1 @ /usr/lib/rstudio/resources/app/bin/quarto/bin/tools/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package * version date (UTC) lib source
#>  sf      * 1.0-15  2023-12-18 [1] CRAN (R 4.3.2)
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.3
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

^{Created on 2024-01-23 with reprex v2.0.2}

bart1 · 2024-01-23T13:02:04Z

Great to see progress! @edzer am I right to notice that this resolves the cast error I posted today but does not address the remarks from yesterday?

bart1 · 2024-01-23T15:22:53Z

Additionally I found this issue with replacing values:

require(sf)
d<-st_as_sf(data.frame(x=1:3,y=1:3), coords = c("x","y"))
e<-st_as_sf(data.frame(x=4:5,y=4:5), coords = c("x","y"))
# Works correctly if both are of the list type:
replace(st_geometry(d)[T,], c(F,T,T), st_geometry(e)[T,])
#> Geometry set for 3 features 
#> Geometry type: POINT
#> Dimension:     XY
#> Bounding box:  xmin: 1 ymin: 1 xmax: 5 ymax: 5
#> CRS:           NA
#> POINT (1 1)
#> POINT (4 4)
#> POINT (5 5)
# However not of the replacement is to the matrix type:
replace(st_geometry(d)[T,], c(F,T,T), st_geometry(e))
#> Geometry set for 3 features  (with 2 geometries empty)
#> Geometry type: POINT
#> Dimension:     XY
#> Bounding box:  xmin: 1 ymin: 1 xmax: 1 ymax: 1
#> CRS:           NA
#> POINT (1 1)
#> POINT EMPTY
#> POINT EMPTY
# Replace is equivalent to:
dd<-st_geometry(d)[T,]
ee<-st_geometry(e)
dd[c(F,T,T)]<-ee
dd
#> Geometry set for 3 features  (with 2 geometries empty)
#> Geometry type: POINT
#> Dimension:     XY
#> Bounding box:  xmin: 1 ymin: 1 xmax: 1 ymax: 1
#> CRS:           NA
#> POINT (1 1)
#> POINT EMPTY
#> POINT EMPTY
sessioninfo::package_info("attached")
#>  package * version date (UTC) lib source
#>  sf      * 1.0-16  2024-01-23 [1] Github (r-spatial/sf@4ce501e)

Fix value replacement for sfc_POINT

ensure vctrs operations work

bart1 · 2025-04-10T11:17:17Z

I just check the most recent version of this branch st_is_empty seems to work incorrectly:

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 8.2.1; sf_use_s2() is TRUE
st_sfc(st_point(),st_point(c(1,2))) |> st_is_empty()
#> [1]  TRUE FALSE
data.frame(x=c(NA,1),y=c(NA,2)) |> st_as_sf(coords=c('x','y'), na.fail=F) |> st_is_empty()
#> [1] TRUE TRUE
sessioninfo::session_info()
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.4.2 (2024-10-31)
#>  os       Ubuntu 22.04.5 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Europe/Amsterdam
#>  date     2025-04-10
#>  pandoc   3.2 @ /usr/lib/rstudio/resources/app/bin/quarto/bin/tools/x86_64/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package     * version date (UTC) lib source
#>  class         7.3-23  2025-01-01 [4] CRAN (R 4.4.2)
#>  classInt      0.4-11  2025-01-08 [1] CRAN (R 4.4.0)
#>  cli           3.6.4   2025-02-13 [1] CRAN (R 4.4.2)
#>  DBI           1.2.3   2024-06-02 [1] CRAN (R 4.4.0)
#>  digest        0.6.37  2024-08-19 [1] CRAN (R 4.4.0)
#>  e1071         1.7-16  2024-09-16 [1] CRAN (R 4.4.0)
#>  evaluate      1.0.1   2024-10-10 [1] CRAN (R 4.4.0)
#>  fastmap       1.2.0   2024-05-15 [1] CRAN (R 4.4.0)
#>  fs            1.6.5   2024-10-30 [1] CRAN (R 4.4.0)
#>  glue          1.8.0   2024-09-30 [1] CRAN (R 4.4.0)
#>  htmltools     0.5.8.1 2024-04-04 [1] CRAN (R 4.4.0)
#>  KernSmooth    2.23-26 2025-01-01 [4] CRAN (R 4.4.2)
#>  knitr         1.49    2024-11-08 [1] CRAN (R 4.4.0)
#>  lifecycle     1.0.4   2023-11-07 [1] CRAN (R 4.4.0)
#>  magrittr      2.0.3   2022-03-30 [1] CRAN (R 4.4.0)
#>  proxy         0.4-27  2022-06-09 [1] CRAN (R 4.4.0)
#>  Rcpp          1.0.14  2025-01-12 [1] CRAN (R 4.4.2)
#>  reprex        2.1.1   2024-07-06 [1] CRAN (R 4.4.0)
#>  rlang         1.1.5   2025-01-17 [1] CRAN (R 4.4.2)
#>  rmarkdown     2.29    2024-11-04 [1] CRAN (R 4.4.0)
#>  rstudioapi    0.17.1  2024-10-22 [1] CRAN (R 4.4.0)
#>  sessioninfo   1.2.2   2021-12-06 [1] CRAN (R 4.4.0)
#>  sf          * 1.0-21  2025-04-10 [1] Github (r-spatial/sf@08e0e1b)
#>  units         0.8-7.1 2025-04-10 [1] Github (r-quantities/units@3fa5ae6)
#>  withr         3.0.2   2024-10-28 [1] CRAN (R 4.4.0)
#>  xfun          0.51.2  2025-02-25 [1] Github (yihui/xfun@e2197a9)
#>  yaml          2.3.10  2024-07-26 [1] CRAN (R 4.4.0)
#> 
#>  [1] /home/bart/R/x86_64-pc-linux-gnu-library/4.4
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

edzer added 2 commits December 13, 2022 11:50

first shot at implementing sfc_POINT with matrix

ec68f40

pass check with points in matrix attribute

d9c5de0

edzer requested review from paleolimbot and rsbivand December 13, 2022 20:29

speed up st_combine

7210201

rsbivand reviewed Dec 14, 2022

View reviewed changes

fix as(x, "Spatial") for points in matrix attr

b16648e

update st_as_sfc.SpatialPoints; see also #2060

1bfb7ea

edzer mentioned this pull request Dec 14, 2022

st_as_sfc.SpatialPoints relatively slow #2060

Closed

Merge branch 'main' into pointx

ea0e62a

edzer added 4 commits December 23, 2022 22:38

Merge branch 'main' into pointx

d0dab0f

merge main

fc83931

Merge branch 'main' into pointx

5275389

Merge branch 'main' into pointx

54a5835

rsbivand mentioned this pull request Feb 7, 2023

Writing a large gpkg file taking forever #1409

Closed

Merge branch 'main' into pointx

3401d0d

edzer added 2 commits March 13, 2023 10:45

Merge branch 'main' into pointx

f30aa65

remove min/max warnings on empty point sets

d81f8ca

bart1 mentioned this pull request Mar 27, 2023

filter.sf order of magnitude slower compared to filter #1889

Open

Merge branch 'main' into pointx

f96e0fa

address comment from @bart1

b8dc2b3

filter empty point; #2059

203187d

edzer added 3 commits August 21, 2023 15:39

Merge branch 'main' into pointx

1bcce69

keep crs in dplyr::filter

4ecabf0

address #2059 (comment)

Merge branch 'main' into pointx

d38b3dc

edzer added 2 commits January 23, 2024 10:48

Merge branch 'main' into pointx

230cde6

update for this branch

4ce501e

olivroy mentioned this pull request Feb 15, 2024

Rename test files to work better with usethis. (align them with R files) #2342

Merged

Fix value replacement for sfc_POINT

5126cd6

bart1 mentioned this pull request Feb 16, 2024

Fix value replacement for sfc_POINT #2345

Merged

edzer and others added 10 commits February 22, 2024 11:24

Merge pull request #2345 from bart1/pointx

c818629

Fix value replacement for sfc_POINT

Merge branch 'main' into pointx

4722081

ensure vctrs operations work

f687956

Merge pull request #2350 from bart1/pointx

b90139f

ensure vctrs operations work

Merge branch 'main' into pointx

d2800ea

Merge branch 'pointx' of github.com:r-spatial/sf into pointx

4a40825

Merge branch 'main' into pointx

7fd0ab1

align with main

82fe210

Merge branch 'main' into pointx

6643388

replace c() with st_combine() in example

08e0e1b

move POINT geometries out of a list column, into a matrix #2059

Are you sure you want to change the base?

move POINT geometries out of a list column, into a matrix #2059

Uh oh!

Conversation

edzer commented Dec 13, 2022

Uh oh!

edzer commented Dec 13, 2022

Uh oh!

edzer commented Dec 13, 2022

Uh oh!

paleolimbot commented Dec 14, 2022

Uh oh!

rsbivand left a comment

Choose a reason for hiding this comment

Uh oh!

edzer commented Dec 14, 2022

Uh oh!

bart1 commented Dec 22, 2022

Uh oh!

edzer commented Dec 22, 2022

Uh oh!

kadyb commented Feb 28, 2023

Uh oh!

edzer commented Feb 28, 2023

Uh oh!

paleolimbot commented Mar 30, 2023

Uh oh!

edzer commented Mar 30, 2023

Uh oh!

edzer commented Jun 29, 2023

Uh oh!

bart1 commented Jun 30, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

edzer commented Jun 30, 2023

Uh oh!

bart1 commented Jun 30, 2023

Uh oh!

bart1 commented Jul 2, 2023

Uh oh!

bart1 commented Jan 22, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bart1 commented Jan 23, 2024

Uh oh!

bart1 commented Jan 23, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bart1 commented Jan 23, 2024

Uh oh!

bart1 commented Apr 10, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

bart1 commented Jun 30, 2023 •

edited

Loading

bart1 commented Jan 22, 2024 •

edited

Loading

bart1 commented Jan 23, 2024 •

edited

Loading