I have been working on a new package makeR to help manage Sweave projects where you wish to create multiple versions of documents that are based on a single source. For example, I create lots of monthly and quarterly reports using Sweave and the only differences between versions are a few variables. I have used GNU make and Apache ANT but wanted a 100% R solution. I will have more to write about that project in a few weeks. In the meantime I needed an example I could use publicly which led me to thinking about analyzing R-Bloggers. I wanted to use the RSS feed to get the frequency of posts and the tags and categories used. However, R-Bloggers, like most blogs, limits the RSS feeds to the latest few. Google Reader however keeps them all (or at least a lot more). The only downside is that you need to have a Google Reader account. The source code is hosted on Gist and provided below.
Setup: retrieve the RSS feed and save it.
source('https://raw.github.com/gist/1606595/269d61dfcc7930f5275a212e11f3c43771ab2591/GoogleReader.R')
rbloggers = getRSSFeed(feedURL="http://r-bloggers.com/feed",
email="GOOGLE READER EMAIL",
passwd="GOOGLE READER PASSWORD",
posts=5000)
entries = rbloggers[which(names(rbloggers) == "entry")]
length(entries)
saveXML(rbloggers, file='rbloggers.xml')
#This will create a data frame with some of the information from the RSS feed
posts = data.frame(title=character(0), author=character(0),
link=character(0), stringsAsFactors=FALSE)
posts[1:length(entries),1:ncol(posts)] = NA
posts$published = as.Date(NA)
posts.categories = list()
for(i in 1:length(entries)) {
entry = entries[[i]]
posts[i,]$title = unclass(xmlChildren(entry[['title']])$text)$value
posts[i,]$author = unclass(xmlChildren(entry[['author']][['name']])$text)$value
posts[i,]$link = xmlAttrs(entry[['link']])[['href']]
posts[i,]$published = as.Date(substr(unclass(
xmlChildren(entry[['published']])$text)$value, 1, 10))
categories = entry[which(names(entry) == "category")]
posts.categories[[i]] = list()
if(length(categories) > 1) { #Ignore the first category as it is used for Google Reader
l = list()
for(j in 2:length(categories)) {
l[(j-1)] = xmlAttrs(categories[[j]])[['term']]
}
posts.categories[[i]] = l
}
}
We’ll use Paul Bleicher’s calendarHeat function to visualize the number of posts per day
source('https://raw.github.com/tavisrudd/r_users_group_1/master/calendarHeat.R')
cal = as.data.frame(table(posts$published))
cal$Var1 = as.Date(cal$Var1)
calendarHeat(cal$Var1, cal$Freq, color="r2b", varname="Number of Posts on R-Bloggers.com")
Create a word cloud
require(wordcloud)
ctab = unlist(posts.categories)
ctab = unlist(strsplit(ctab, ' '))
ctab = as.data.frame(table(ctab))
ctab = ctab[-which(ctab$ctab == 'Uncategorized'),]
wordcloud(ctab$ctab, ctab$Freq, min.freq=10)
######The getRSSFeed function. Note that this function is included in the makeR package.
require(XML)
require(RCurl)
#' This function ruturns an XML tree of the RSS feed from the given URL.
#'
#' This function utilizes the (unofficial) Google Reader API to retrieve RSS
#' feeds. The advantage of access RSS feeds through the Google Reader API is that
#' you are not limited by the number of entries a website may included in their
#' feed. That is, Google maintains generally maintains a complete history of
#' entries from the RSS feed.
#'
#' Note that the contents of the results will be limited by what the website
#' provides in their feeds. That is, Google does not contain more information
#' per entry then what the website originally provided. If the initial feed
#' contained only excerpts of the article, the feed from Google will too only
#' contain excerpts. Be aware though that for sites that do provide the complete
#' contents of posts will result in potentially very large downloads.
#'
#' @param feedURL the full URL to the RSS feed.
#' @param email the email address for the Google Reader account
#' @param passwd the password for the Google Reader account
#' @param posts the number of posts to return
#' @return the root \code{XMLNode} for the RSS feed.
#' @seealso \code{\link{/xmlRoot}} for the format of the returned XML tree
#' @export
#' @example
#' \dontrun{
#' rbloggers = getRSSFeed(feedURL="http://r-bloggers.com/feed",
#' email="USERNAME@gmail.com", passwd="PASSWORD")
#' }
#' @author Jason Bryer <\email{jason@@bryer.org}x>
getRSSFeed <- function(feedURL, email, passwd, posts=1000) {
#Authenticate with Google
curlHandle = getCurlHandle(cookiefile="rcookies", ssl.verifyhost=FALSE, ssl.verifypeer=FALSE)
x = postForm("https://www.google.com/accounts/ClientLogin",
accountType="GOOGLE",
service="reader",
Email=email,
Passwd=passwd,
source="makeR",
curl = curlHandle)
gtoken = unlist(strsplit(x, "\n"))
parsed.gtoken <- unlist(strsplit(gtoken[3], "Auth="))
if (length(parsed.gtoken) >= 2) {
auth.token <- unlist(strsplit(gtoken[3], "Auth="))[[2]]
} else {
stop("Authentication failed.")
}
google.auth <- paste("GoogleLogin auth=", auth.token, sep='')
#Function to retrieve up to 1000 posts
getDoc <- function(n, c=NULL) {
feedURL = paste("http://www.google.com/reader/atom/feed/", feedURL, "?n=", n,
ifelse(is.null(c), "", paste("&c=", c, sep='')),
sep='')
feed = getURL(feedURL, .encoding = 'UTF-8', followlocation=TRUE,
httpheader=c("Authorization"=google.auth),
curl=curlHandle)
doc = xmlTreeParse(feed, asText=TRUE)
return(xmlRoot(doc))
}
root = NULL
continueValue = NULL
for(i in 1:ceiling(posts / 1000)) {
r = getDoc(n=ifelse(i == ceiling(posts / 1000), (posts-1) %% 1000 + 1, 1000),
c=continueValue)
if(is.null(root)) {
root = r
} else {
entries = which(xmlSApply(r, xmlName) == 'entry')
if(length(entries) > 0) {
root = addChildren(root, kids=r[entries])
}
}
if(is.null(r[['continuation']])) {
break #No more posts to retrieve
} else {
continueValue = unclass(xmlChildren(r[['continuation']])$text)$value
}
}
return(root)
}