st_is_valid: return logical vector?

[GreenStat]

related tot this item #290 :

when I have:

library(tidyverse)
library(sf) # lates dev version
# toy df
df <- data.frame(
id=c("A","A","B","B","B","C"),
x=c(1,2,4,3,3,5),
y=c(1,2,4,3,5,5))

# note C !

I like to make a 'MULTIPLELINESTRING' object based on id and calculate length

df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% 
 summarize() %>% 
 st_cast() %>%  # this needs to be here, makes multipoint 
 st_cast('MULTILINESTRING') %>%
 ggplot(aes(color=id)) + geom_sf()

this looks good (disregard the legend):

I need length of line segments but this obvouisly fails because "C" is a multilinestring geometry with
only one point

df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
 st_cast() %>%  st_length()
#Error in p[1, ] : subscript out of bounds

checking geometry with st_is_valid and st_make_valid reveals some unexpected results ....

# st_is_valid
df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
 st_cast() %>%  st_is_valid()
[1] TRUE TRUE   NA
#
# st_is_valid should return a logical vector (i guess)
# I expected FALSE on the third item.....


#st_make_valid
df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
 st_cast() %>%  st_make_valid()
Error in CPL_make_valid(x) : 
  st_make_valid requires compilation against liblwgeom

# how do I get liblwgeom?

# also st_is_simple chokes on this ....

#st_is_simple
df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
 st_cast() %>%  st_is_simple()
Error in CPL_geos_is_simple(st_geometry(x)) : 
  IllegalArgumentException: point array must contain 0 or >1 elements

# 

> sessionInfo()
R version 3.3.1 (2016-06-21)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7600)

locale:
[1] LC_COLLATE=Dutch_Netherlands.1252  LC_CTYPE=Dutch_Netherlands.1252   
[3] LC_MONETARY=Dutch_Netherlands.1252 LC_NUMERIC=C                      
[5] LC_TIME=Dutch_Netherlands.1252    

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

other attached packages:
 [1] data.table_1.10.4  sf_0.4-2           mapview_1.2.0     
 [4] leaflet_1.0.1      sp_1.2-4           dplyr_0.5.0       
 [7] purrr_0.2.2        readr_1.0.0        tidyr_0.6.0       
[10] tibble_1.2         ggplot2_2.2.1.9000 tidyverse_1.1.1   

loaded via a namespace (and not attached):
 [1] reshape2_1.4.2      haven_1.0.0         lattice_0.20-34    
 [4] colorspace_1.3-1    htmltools_0.3.5     stats4_3.3.1       
 [7] viridisLite_0.1.3   R.oo_1.21.0         foreign_0.8-67     
[10] DBI_0.6-1           R.utils_2.5.0       RColorBrewer_1.1-2 
[13] modelr_0.1.0        readxl_0.1.1        foreach_1.4.3      
[16] plyr_1.8.4          stringr_1.2.0       munsell_0.4.3      
[19] gtable_0.2.0        rvest_0.3.2         raster_2.5-8       
[22] R.methodsS3_1.7.1   htmlwidgets_0.8     codetools_0.2-15   
[25] psych_1.6.12        latticeExtra_0.6-28 forcats_0.2.0      
[28] parallel_3.3.1      gdalUtils_2.0.1.7   broom_0.4.1        
[31] Rcpp_0.12.10        geosphere_1.5-5     udunits2_0.13      
[34] scales_0.4.1        satellite_0.2.0     webshot_0.4.0      
[37] jsonlite_1.3        mnormt_1.5-5        hms_0.3            
[40] png_0.1-7           digest_0.6.12       stringi_1.1.2      
[43] grid_3.3.1          rgdal_1.2-5         tools_3.3.1        
[46] magrittr_1.5        lazyeval_0.2.0      xml2_1.1.1         
[49] lubridate_1.6.0     assertthat_0.1      httr_1.2.1         
[52] iterators_1.0.8     R6_2.2.0            units_0.4-3        
[55] nlme_3.1-128  

[edzer]

You get liblwgeom by

sudo apt-get install liblwgeom-dev

but that might not work on every operating system.

Also st_make_valid will not make a linestring from your single point (C). Maybe filter it out using st_dimension?

The NA returned by st_is_valid is a logical NA. I'd be careful and not interpret it as TRUE: MULTILINESTRING((5 5))

[GreenStat]

Oke
no, sudo is not working for win os, but I'll find it.

st_dimension does not 'solve' it either,

df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
 group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
 st_cast() %>% st_dimension()
[1] 1 1 1
# (1 is line)

however, using summarize to count the # points gives an other handle to filter:

df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
group_by(id) %>% summarize(len=n()) %>%  filter(len>1) %>%
st_cast() %>%  st_cast('MULTILINESTRING') %>% 
ggplot(aes(color=id,size=2)) + geom_sf()

and now st_length also works

df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
group_by(id) %>% summarize(len=n()) %>%  filter(len>1) %>%
st_cast() %>%  st_cast('MULTILINESTRING') %>% st_length()
Units: m
[1] 156876.1 377825.2

somehow I expected a 'proper' multilinestring geometry to have at least 2 points ....
anyway problem solved, case closed :-)

[edzer]

I now remember why it is the way it currently is; see also this blog. If you run

 x <- df %>% st_as_sf(coords=c('x','y'),crs=4326) %>%
  group_by(id) %>% summarize() %>% st_cast() %>% st_cast('MULTILINESTRING') %>%
  st_cast()
x
Simple feature collection with 3 features and 1 field
geometry type:  MULTILINESTRING
dimension:      XY
bbox:           xmin: 1 ymin: 1 xmax: 5 ymax: 5
epsg (SRID):    4326
proj4string:    +proj=longlat +datum=WGS84 +no_defs
  id                       geometry
1  A    MULTILINESTRING((1 1, 2 2))
2  B MULTILINESTRING((3 3, 3 5, ...
3  C         MULTILINESTRING((5 5))
> st_is_valid(x)
[1] TRUE TRUE   NA
> st_is_valid(x, NA_on_exception=FALSE)
Error in CPL_geos_is_valid(st_geometry(x), as.logical(NA_on_exception)) : 
  IllegalArgumentException: point array must contain 0 or >1 elements
> st_is_valid(x, reason = TRUE)
[1] "Valid Geometry" "Valid Geometry" NA              

I turned off NA_on_exceptoin=FALSE as default, since you typically have many feature geometries, in which case a single error message does not tell you which geometry causes this error is, and how many of them do. Returning NA_logical_ lets you filter these. This still needs proper documentation though.