Parsing a HTML page for getting a list of Twitter screen names

gistfile1.r

########################################################################################
#
#  Make a list of Twitter screen names from 
#  http://yle.fi/extrem/artikel/musiknoje/44988-Har-ar-svenskfinlands-basta-twittrare
#  
#  AFAIK there is no Twitter list, so we need HTML parsing
#  
#  Tuija Sonkkila
#  2012-07-05

library(RCurl)
library(RJSONIO)
library(XML)

url <- "http://yle.fi/extrem/artikel/musiknoje/44988-Har-ar-svenskfinlands-basta-twittrare"
d <- getURL(url)
doc <- htmlParse(d)

# Store into a list (src) all href attributes of those link elements which refer to Twitter
# (there are no extra, unrelated links on the page)
src <- xpathApply(doc, "//a[starts-with(@href, 'https://twitter.com/')]", xmlGetAttr, "href")

# Apply a find&replace function over all list elements, to get the screen names only
ppl <- lapply(src, function(x) gsub("https://twitter.com/", "", x))

# During the text mining phase later on, I got errors that I assumed where related to
# character encoding of the corpus (latin1). This conversion helped:
# 
# tweets.utf8 <- iconv(tweets, "latin1", "UTF-8")