summarise.sf running into a null polygon geometry causes R to crash

[alankjackson]

Describe the bug
File containing an empty polygon geometry cause summarise to crash R

> bad_states %>% group_by(value) %>% summarize(num=sum(aland, na.rm = TRUE))

 *** caught segfault ***
address 0x8, cause 'memory not mapped'

Traceback:
 1: cpp_s2_unary_union(x, options)
 2: structure(x, class = c("s2_geography", "wk_vctr"))
 3: new_s2_geography(cpp_s2_unary_union(x, options))
 4: s2_union(x, options = options)
 5: cpp_s2_union_agg(s2_union(x, options = options), options, na.rm)
 6: structure(x, class = c("s2_geography", "wk_vctr"))
 7: new_s2_geography(cpp_s2_union_agg(s2_union(x, options = options),     options, na.rm))
 8: s2::s2_union_agg(x, ...)
 9: st_as_sfc(s2::s2_union_agg(x, ...), crs = st_crs(x))
10: st_union.sfc(geom[i == x], is_coverage = is_coverage)
11: st_union(geom[i == x], is_coverage = is_coverage)
12: FUN(X[[i]], ...)
13: lapply(sort(unique(i)), function(x) st_union(geom[i == x], is_coverage = is_coverage))
14: summarise.sf(., num = sum(aland, na.rm = TRUE))
15: summarize(., num = sum(aland, na.rm = TRUE))
16: bad_states %>% group_by(value) %>% summarize(num = sum(aland,     na.rm = TRUE))

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace

To Reproduce

library(tidyverse)

#   Grab state boundaries as convenient polygon set
states <- USAboundaries::us_states(resolution = "low")

#   Create a bogus state to join
state_abbv <- c(states$state_abbr, "??") 
new_state <- tibble::tibble(value=1:53, state_abbr=state_abbv)

#   Join to original state dataframe to create a bad polygon
bad_states <- left_join(new_state, states, by="state_abbr")
bad_states <- sf::st_as_sf(bad_states)

#   The summarize will crash the rstudio session
bad_states %>% 
  group_by(value) %>% 
    summarize(num=sum(aland, na.rm = TRUE)) 

If reporting a change from previous versions

Please read https://cran.r-project.org/web/packages/sf/news/news.html first.

Additional context
Ubuntu 20.04.5 LTS
R version 4.2.1 (2022-06-23)

Paste the output of your `sessionInfo()` and `sf::sf_extSoftVersion()`

sessionInfo()
R version 4.2.1 (2022-06-23)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.5 LTS

Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/liblapack.so.3

locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

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

loaded via a namespace (and not attached):
[1] compiler_4.2.1

sf::sf_extSoftVersion()
GEOS GDAL proj.4 GDAL_with_GEOS USE_PROJ_H
"3.8.0" "3.0.4" "6.3.1" "true" "true"
PROJ
"6.3.1"

[edzer]

Could you please report the output of sessionInfo() after loading tidyverse and sf?

[rsbivand]

I can see a segfault in s2 unary union. Am updating packages to get the freshest troubleverse versions, s2 as CRAN, sf local devel version. Will re-run with gdb later today.

[rsbivand]

library(tidyverse)
library(sf)
states <- USAboundaries::us_states(resolution = "low")
state_abbv <- c(states$state_abbr, "??") 
new_state <- tibble::tibble(value=1:53, state_abbr=state_abbv)
bad_states <- left_join(new_state, states, by="state_abbr")
bad_states <- st_as_sf(bad_states)
st_is_longlat(bad_states)
a <- group_by(bad_states, value)
summarize(a, num=sum(aland, na.rm = TRUE))

# Thread 1 "R" received signal SIGSEGV, Segmentation fault.

s2geography::s2_find_validation_error (geog=..., error=error@entry=0x7ffffffbdb60) at /usr/include/c++/12/bits/unique_ptr.h:191
191	      pointer    _M_ptr() const noexcept { return std::get<0>(_M_t); }

(gdb) bt
#0  s2geography::s2_find_validation_error (geog=..., 
    error=error@entry=0x7ffffffbdb60)
    at /usr/include/c++/12/bits/unique_ptr.h:191
#1  0x00007fffdb75ca1d in s2geography::s2_find_validation_error (geog=..., 
    error=error@entry=0x7ffffffbdb60)
    at /usr/include/c++/12/bits/unique_ptr.h:191
#2  0x00007fffdb76077c in s2geography::s2_unary_union (geog=..., options=...)
    at s2geography/build.cc:202
#3  0x00007fffdb738051 in Op::processFeature (this=<optimized out>, 
    feature=..., i=<optimized out>) at s2-transformers.cpp:295
#4  0x00007fffdb7237b3 in UnaryGeographyOperator<Rcpp::Vector<19, Rcpp::PreserveStorage>, SEXPREC*>::processVector (this=0x7ffffffbded0, geog=...)
    at /tmp/RtmpxIxiha/R.INSTALL38cc872f3842/s2/src/geography-operator.h:35
#5  0x00007fffdb737f84 in cpp_s2_unary_union (geog=..., s2options=...)
    at s2-transformers.cpp:306
#6  0x00007fffdb747f9c in _s2_cpp_s2_unary_union (geogSEXP=0x80c6d68, 
    s2optionsSEXP=<optimized out>) at RcppExports.cpp:1293
#7  0x00007ffff7afd09c in R_doDotCall (
    ofun=ofun@entry=0x7fffdb747f10 <_s2_cpp_s2_unary_union(SEXP, SEXP)>, 
    nargs=nargs@entry=2, cargs=cargs@entry=0x7ffffffc0830, 
    call=call@entry=0x8ad5d30) at ../../../src/main/dotcode.c:604
#8  0x00007ffff7b3e2f3 in bcEval (body=<optimized out>, rho=<optimized out>, 
    useCache=<optimized out>) at ../../../src/main/eval.c:7692
#9  0x00007ffff7b53e90 in Rf_eval (e=0x8ad5dd8, rho=rho@entry=0xa2782c0)
    at ../../../src/main/eval.c:748
#10 0x00007ffff7b5598a in R_execClosure (call=call@entry=0x89d3d48, 
    newrho=newrho@entry=0xa2782c0, sysparent=<optimized out>, 
    rho=rho@entry=0xa1c8950, arglist=arglist@entry=0xa2783d8, 
    op=op@entry=0x8ad6190) at ../../../src/main/eval.c:1918
#11 0x00007ffff7b5684e in Rf_applyClosure (call=call@entry=0x89d3d48, 
    op=op@entry=0x8ad6190, arglist=arglist@entry=0xa2783d8, 
    rho=rho@entry=0xa1c8950, suppliedvars=<optimized out>)
    at ../../../src/main/eval.c:1844
#12 0x00007ffff7b419fe in bcEval (body=<optimized out>, rho=<optimized out>, 
    useCache=<optimized out>) at ../../../src/main/eval.c:7104
#13 0x00007ffff7b53e90 in Rf_eval (e=0x89d3680, rho=0xa1c8950)
    at ../../../src/main/eval.c:748
#14 0x00007ffff7b54812 in forcePromise (e=e@entry=0xa2748b8)

No problem with sf_use_s2(FALSE). @paleolimbot any idea why an empty geometry causes havoc here?

[paleolimbot]

Off the top of my head, no, but I remember that the unary union was something I had to implement "by hand" (i.e., it's not in S2 proper) so it's likely an error of mine somewhere. I'll step through the example in the debugger later today 🙂

[fmark]

I get a similar error trying to do a summarise on a different data set with an empty geometry using a different dataset.

[rsbivand]

@fmark Did you try to install the development version of s2 in which the problem was fixed? You give no details of your data set - GEOGCRS not PROJCRS, for example?

[paleolimbot]

It will be a few days until there are binaries, but the s2 version that fixes this is now on CRAN!

[alankjackson]

Awesome! Thank you.