read a .kml file with the KML driver: does not load the data columns?

[MatthieuStigler]

Hi

I am reading a kml file with st_read, and have the issue that it only reads the geometry field, as well as two empty fields Name and Description, but not the rest. Interestingly, this only happens on one machine that uses the driver KML (st_read gives 2 fields), but not on another machine that uses LIBKML (st_read gives 24 fields).

Should I actually put specify options when using KML instead of LIBKML? Or try to have the LIBKML driver on the machine, hoping that would temporarily solve the issue?

Thanks!

library(sf)

nc = st_read(system.file("shape/nc.shp", package="sf"), quiet=TRUE) 
st_write(nc, "test.kml")
st_read("test.kml")

Output on machine that has only driver KML:

Reading layer test' from data source /home/stigler/GIS/test.kml' using driver `KML'
Simple feature collection with 100 features and 2 fields

Output on the other machine using driver LIBKML

Reading layer test' from data source /home/matifou/test.kml' using driver `LIBKML'
Simple feature collection with 100 features and 24 fields

[edzer]

My gdal installation has both,

> st_drivers()[19:20,]
         name                        long_name write  copy is_raster is_vector
LIBKML LIBKML Keyhole Markup Language (LIBKML)  TRUE FALSE     FALSE      TRUE
KML       KML    Keyhole Markup Language (KML)  TRUE FALSE     FALSE      TRUE

and apparently libkml is the better one and automatically chosen. st_read doesn't let you choose a driver, so I can't try the other.

My guess is that your machine with only driver KML runs windows or OSX, and comes with a gdal installation that doesn't have libkml. You'd need to recompile gdal on that machine, linked to libkml, and then install sf from source, linking against that.

[MatthieuStigler]

Thanks for checking this! So I guess the problem is likely at the gdal level, or even at the KML library level?

The machine with only KML is actually a Linux red hat 7.4 server , I'll have to speak to the admin to see what he can do.

Thanks!

[edzer]

Yes, you can solve it by installing a better equipped gdal.

[senickel]

I hope it is fine if still add into this discussion. I had a similar problem and tried to install LibKML on Windows 10 which did not work. My work around was to access the ExtendedData in the KML separately and attach it to the simple feature. I posted a similar answer for rgdal here: https://stackoverflow.com/questions/45989198/how-to-load-all-fields-extendeddata-not-just-name-and-description-from-kml/51657844#51657844

readKML<-function(file,...) {
  sp_obj<-st_read(file,...)
  xml1<-read_xml(file)

    # extract name and type of variables
  variable_names<-xml_find_first(xml1,".//d1:Schema[@name]") %>% 
    xml2::xml_contents() %>%
    xml2::xml_attrs() %>% 
    do.call(rbind,.)
  
  # return sp_obj if no ExtendedData is present
  if (is.null(variable_names)) return(sp_obj)
  
  data<-lapply(seq(nrow(variable_names)),function(x) {
    x1<-xml_find_all(xml1,paste0(".//d1:SimpleData[@name=\"",variable_names[x,1],"\"]")) %>% 
      xml_contents()  
    if (variable_names[x,2]=="string") return(xml_text(x1))
    xml_double(x1)
  }) %>% 
    do.call(cbind,.) %>% 
    as.data.frame()
  
  colnames(data)<-variable_names[,1]
  
    sp_obj[,variable_names[,1]]<-data
  sp_obj
}

[markbneal]

I think this is still an issue that kml driver is not reading and making available the relevant info. I did not have any luck with @senickel s code.

Using online web converters from kml to shp or geojson, then reading in by st_read() did make all relevant fields available, but not very scalable!

I have access to RStudio on a server running linux, so I'm reading my kml files there for the moment.