[R-sig-Geo] st_union crashes RStudio - bug?

Roger Bivand Roger@Biv@nd @ending from nhh@no
Fri Sep 14 22:20:35 CEST 2018 

The in-flight analysis shows that the precision model interacts with the
geojson text representation of coordinates. I tried to use st_precision
without success. Using:

library(rgdal)
library(rgeos)
test1 <- readOGR("Australia.geojson")
setScale(4e+4)
crash1 <- gUnaryUnion(test1)
plot(crash1)
length(slot(slot(crash1, "polygons")[[1]], "Polygons"))
areas <- sapply(slot(slot(crash1, "polygons")[[1]], "Polygons"), slot,
"area")
head(sort(areas), n=10)
which(areas < 5e-8)
Pols <- slot(slot(crash1, "polygons")[[1]], "Polygons")
Pols1 <- Pols[-which(areas < 5e-8)]
slot(slot(crash1, "polygons")[[1]], "Polygons") <- Pols1
plot(crash1)

and changing scale to 4e+4 in rgeos notation, most of the remnant boundary
artefacts are removed. You can even remove the three dots in south-east
Australia. So: if you want to operations other than visualisation, avoid
geojson. If you can't avoid geojson, don't be surprised by topological
issues. Further, do use a vector representation suited to your purpose -
this one has far too much detail.

Roger

PS. Not tried on Windows - can someone please check that this doesn't
misbehave on the OP's platform?

On Fri, 14 Sep 2018, Roger Bivand wrote:

> (from an airport)
>
>> crash <- st_union(noACT)
>
> completed without error, test also passes. Your problem is not general
>
> There is absolutely no need to use ggplot to display the geometries, by 
> the way.
>
> plot(st_geometry(crash))
>
> shows errant partial boundaries, probably caused by the default 
> precision model and geojson's character representation of numbers.
>
> Roger
>
> Roger Bivand
> Norwegian School of Economics
> Bergen, Norway
>
>
>
>
> On Fri, Sep 14, 2018 at 7:00 PM +0200, "Roger Bivand" <Roger.Bivand using nhh.no<mailto:Roger.Bivand using nhh.no>> wrote:
>
>
> On Fri, 14 Sep 2018, Roman Luštrik wrote:
>
>> Crashes R or Rstudio?
>
> I guess the OP will get up sometime soon; of course one should always run
> R in a console, not RStudio when reporting issues of this kind. Further,
> one should run R -d gdb (if possible - see RW FAQ) to trap any important
> information. I can see that crash <- st_union(noACT) is long-running, with
> one thread at 100%, but no obvious memory issues (Fedora 28, sf 0.6-3,
> GEOS 3.7.0). It would further make sense to go the sp/rgeos route to see
> if it is data+GEOS or that the behavious is R-implementation dependent.
>
> Roger
>
>>
>> Cheers,
>> Roman
>>
>> On Fri, Sep 14, 2018 at 8:33 AM Bertram Ostendorf <
>> bertram.ostendorf using adelaide.edu.au> wrote:
>>
>>> I am trying to dissolve internal boundaries using sf.  This crashes R in
>>> the example geojson map below, a simple map of Australia’s states and
>>> territories. I am using a polygon layer that processes fine in other GIS
>>> and also works fine using sp (casting to sp and using gUnaryUnion).  I am
>>> not after a workaround but I'd like to explore what's different in sf and
>>> why this particular layer causes problems.
>>>
>>> My questions: Is this a bug in sf? Have others experienced similar issues
>>> using simple features?
>>>
>>> Thanks
>>>
>>> Bertram Ostendorf
>>> Chair for Spatial Environmental Sciences
>>> Director: Spatial Information Group
>>> The University of Adelaide, AUSTRALIA 5005
>>>
>>>
>>> # Here's my example layer and code:
>>> # Note: Neither ArcGIS ‘repair geometry’ nor  ‘st_is_valid’ show any
>>> issues.
>>> library(sf)
>>> library(ggplot2)
>>> # read Australia.geojson from dropbox, 15.5Mb
>>> test <- st_read("
>>> https://www.dropbox.com/s/060c6lfijyx1e5v/Australia.geojson?dl=1")
>>> st_is_valid(test)
>>>
>>> # The next lines work fine, I added them to illustrate what works
>>> ggplot(test) + geom_sf(aes(fill=STATE))
>>> ACT <- test[test$STATE == 'ACT', ]
>>> NSW <- test[test$STATE == 'NSW', ]
>>> ACTandNSW <- test[test$STATE %in% c('NSW','ACT'), ]
>>> noACT <- test[test$STATE != 'ACT', ]
>>> noNSW <- test[test$STATE != 'NSW', ]
>>> noACTandNSW <- test[!test$STATE %in% c('NSW','ACT'), ]
>>> # st_union works ok for most subsets. But note that some internal
>>> boundaries are not removed.
>>> ok <- st_union(ACT)
>>> # donut polygons work fine
>>> ok <- st_union(NSW)
>>> # donut polygons (NSW) filled (ACT) work fine
>>> ok <- st_union(ACTandNSW)
>>> # all states except NSW work fine
>>> ok <- st_union(noNSW)
>>> # ok if both ACT and NSW are removed
>>> ok <- st_union(noACTandNSW)
>>> ggplot(ok) + geom_sf()
>>>
>>> # Thing go pear-shaped if the hole is filled (ACT in NSW), but only if
>>> neighbours of the filled donut polygon are present.
>>> # The two lines below stall R and eventually crash RStudio
>>> crash <- st_union(test)
>>> # Australia without its governing territory kills R
>>> crash <- st_union(noACT)
>>>
>>>
>>>> sessionInfo()
>>> R version 3.5.1 (2018-07-02)
>>> Platform: x86_64-w64-mingw32/x64 (64-bit)
>>> Running under: Windows >= 8 x64 (build 9200)
>>>
>>> Matrix products: default
>>>
>>> locale:
>>>   [1] LC_COLLATE=English_Australia.1252  LC_CTYPE=English_Australia.1252
>>>   LC_MONETARY=English_Australia.1252
>>> [4] LC_NUMERIC=C                       LC_TIME=English_Australia.1252
>>>
>>> attached base packages:
>>>   [1] stats     graphics  grDevices utils     datasets  methods   base
>>>
>>> other attached packages:
>>>   [1] ggplot2_3.0.0 sf_0.6-3
>>>
>>> loaded via a namespace (and not attached):
>>>   [1] Rcpp_0.12.18     rstudioapi_0.7   bindr_0.1.1      magrittr_1.5
>>>  units_0.6-0      tidyselect_0.2.4 munsell_0.5.0
>>> [8] colorspace_1.3-2 R6_2.2.2         rlang_0.2.2      plyr_1.8.4
>>>  dplyr_0.7.6      tools_3.5.1      grid_3.5.1
>>> [15] gtable_0.2.0     e1071_1.7-0      DBI_1.0.0        withr_2.1.2
>>> class_7.3-14     digest_0.6.16    yaml_2.2.0
>>> [22] lazyeval_0.2.1   assertthat_0.2.0 tibble_1.4.2     crayon_1.3.4
>>>  bindrcpp_0.2.2   spData_0.2.9.3   purrr_0.2.5
>>> [29] glue_1.3.0       compiler_3.5.1   pillar_1.3.0     scales_1.0.0
>>>  classInt_0.2-3   pkgconfig_2.0.2
>>>
>>>
>>>
>>>
>>>
>>>         [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo using r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>
>>
>>
>>
>
> --
> Roger Bivand
> Department of Economics, Norwegian School of Economics,
> Helleveien 30, N-5045 Bergen, Norway.
> voice: +47 55 95 93 55; e-mail: Roger.Bivand using nhh.no
> http://orcid.org/0000-0003-2392-6140
> https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: Roger.Bivand using nhh.no
http://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en

More information about the R-sig-Geo mailing list