How to parse XML to R data frame

Data in XML format are rarely organized in a way that would allow the xmlToDataFrame function to work. You're better off extracting everything in lists and then binding the lists together in a data frame:

require(XML)
data <- xmlParse("http://forecast.weather.gov/MapClick.php?lat=29.803&lon=-82.411&FcstType=digitalDWML")

xml_data <- xmlToList(data)

In the case of your example data, getting location and start time is fairly straightforward:

location <- as.list(xml_data[["data"]][["location"]][["point"]])

start_time <- unlist(xml_data[["data"]][["time-layout"]][
    names(xml_data[["data"]][["time-layout"]]) == "start-valid-time"])

Temperature data is a bit more complicated. First you need to get to the node that contains the temperature lists. Then you need extract both the lists, look within each one, and pick the one that has "hourly" as one of its values. Then you need to select only that list but only keep the values that have the "value" label:

temps <- xml_data[["data"]][["parameters"]]
temps <- temps[names(temps) == "temperature"]
temps <- temps[sapply(temps, function(x) any(unlist(x) == "hourly"))]
temps <- unlist(temps[[1]][sapply(temps, names) == "value"])

out <- data.frame(
  as.list(location),
  "start_valid_time" = start_time,
  "hourly_temperature" = temps)

head(out)
  latitude longitude          start_valid_time hourly_temperature
1    29.81    -82.42 2013-06-19T16:00:00-04:00                 91
2    29.81    -82.42 2013-06-19T17:00:00-04:00                 90
3    29.81    -82.42 2013-06-19T18:00:00-04:00                 89
4    29.81    -82.42 2013-06-19T19:00:00-04:00                 85
5    29.81    -82.42 2013-06-19T20:00:00-04:00                 83
6    29.81    -82.42 2013-06-19T21:00:00-04:00                 80

Use xpath more directly for both performance and clarity.

time_path <- "//start-valid-time"
temp_path <- "//temperature[@type='hourly']/value"

df <- data.frame(
    latitude=data[["number(//point/@latitude)"]],
    longitude=data[["number(//point/@longitude)"]],
    start_valid_time=sapply(data[time_path], xmlValue),
    hourly_temperature=as.integer(sapply(data[temp_path], as, "integer"))

leading to

> head(df, 2)
  latitude longitude          start_valid_time hourly_temperature
1    29.81    -82.42 2014-02-14T18:00:00-05:00                 60
2    29.81    -82.42 2014-02-14T19:00:00-05:00                 55

Here's a partial solution using xml2. Breaking the solution up into smaller pieces generally makes it easier to ensure everything is lined up:

library(xml2)
data <- read_xml("http://forecast.weather.gov/MapClick.php?lat=29.803&lon=-82.411&FcstType=digitalDWML")

# Point locations
point <- data %>% xml_find_all("//point")
point %>% xml_attr("latitude") %>% as.numeric()
point %>% xml_attr("longitude") %>% as.numeric()

# Start time
data %>% 
  xml_find_all("//start-valid-time") %>% 
  xml_text()

# Temperature
data %>% 
  xml_find_all("//temperature[@type='hourly']/value") %>% 
  xml_text() %>% 
  as.integer()

Tags:

Xml

R