Duncan Temple Lang

University of California at Davis

Department of Statistics

library(XML)
dir = "~/Books/XMLTechnologies/XSL/examples/Earthquakes/"
doc = xmlParse(sprintf("%s%s", dir, "2004_Earthquakes_ALL.kml"))

We get all the <Placemark> nodes under the <Folder> nodes.

pl = getNodeSet(doc, "//kml:Folder/kml:Placemark", "kml")

Now we extract the longitude and latitude, the magnitude, etc.

Before we do this, let's just test one thing. Which is faster: getting the <Placemark> nodes once and then working on these to extract their children in R, or to use XPath many times to get each of the different elements in the document that we want, <coordinates>, <description>, <name> within in each <Placemark>. Let's time these

system.time({desc = sapply(pl, function(x) xmlValue(x[["description"]]))})
   user  system elapsed 
 18.411   0.972  35.986 

system.time({desc1 = xpathSApply(doc, "//kml:Folder/kml:Placemark/kml:description", xmlValue, namespaces = "kml")})
   user  system elapsed 
  0.711   0.052   1.262 

all(desc1 == desc)

So it is much faster (in this context) to use XPath many times rather than using R to access the children by name. Accessing by index may be different.

system.time({desc = sapply(pl, function(x) xmlValue(x[[2]]))})
   user  system elapsed 
 10.201   0.697  17.039

Reading one File

nm = xpathSApply(doc, "//kml:Folder/kml:Placemark/kml:name", xmlValue, namespaces = "kml")
#nm = sapply(pl, function(x) xmlValue(x[["name"]]))
mags = gsub("^M (None|[0-9.]+) -.*", "\\1", nm)
mags[mags == "None"] = "0"
magnitude = as.numeric(mags)
# Check the results
sum(is.na(magnitude)) == 0
table(floor(magnitude))

Now get the depth from the HTML content in the description.

desc = xpathSApply(doc, "//kml:Folder/kml:Placemark/kml:description", xmlValue, namespaces = "kml")
# desc = sapply(pl, function(x) xmlValue(x[["description"]]))
getDepthStringFromDesc = 
function(x) { 
  hdoc = htmlParse(x, asText = TRUE)
  xmlValue(getNodeSet(hdoc, "//tr[2]/td/font")[[1]])
}
str = sapply(desc, getDepthStringFromDesc)
depth = as.numeric(gsub(" .*", "", str)) # strip names
sum(is.na(depth))

Let's get the date

dates = unlist(getNodeSet(doc, "//kml:Folder/kml:Placemark/@id", "kml"))
dates = as.POSIXct(strptime(dates, "%Y %b %d %H:%M:%S %Z"))

Now the longitude and latitude

coords = xpathSApply(doc, "//kml:Folder/kml:Placemark/kml:Point/kml:coordinates", xmlValue, namespaces = "kml")
locations = t(sapply(strsplit(coords, ","), function(x) x[1:2]))

Now we assemble the data frame

eq = data.frame(dates = dates, depth = depth, magnitude = magnitude)
eq$longitude = as.numeric(locations[,1])
eq$latitude = as.numeric(locations[,2])

readKMZEarthquakes =
function(doc)
{
 if(is.character(doc)) {
   if(length(grep("^http.*kmz$", doc))) {
       doc = xmlParse(getKMZURL(doc), asText = TRUE)
   } else
       doc = xmlParse(doc)
 }

cat("parsed the XML file\n")
 
  <<magnitude>>
 
 
  <<dates>>
 
 
  <<description>>
 
 
  <<coordinates>>
 
 
  <<assemble>>
 
 eq
}

getKMZURL =
function(doc, readXML = TRUE)
{
  kmz = getURLContent('http://neic.usgs.gov/neis/epic/kml/2009_Earthquakes_ALL.kmz')
  ar = zipArchive(kmz)
  if(readXML)
    invisible(xmlParse(ar[[1]], asText = TRUE))
   else
     ar
}