tapas.r

# What did Barbara get up to in Spain?
#   TL;DR: 11 Gelato, 68 Drinks and 36 Tapas

library(XML)
library(ggplot2)
library(reshape2)

# parse page into XML DOM document
parse_doc = function (page) {

  # get html text so that bad tags can be stripped
  u = url(page)
  html = paste(readLines(u, warn=FALSE), collapse='\n')
  close(u)

  # remove spurious tag with no declared namespace (b:*)
  html = gsub('b:[^ ]+ ', '', html)

  # remove script tags with unparsable CDATA
  html = gsub('<script[^>]*>.*?</script>', '', html)

  return(xmlRoot(htmlTreeParse(html, useInt=TRUE, asText=TRUE)))

}

# get the next page
get_next_page = function (doc) {

  return(xpathSApply(doc, '//a[@class="blog-pager-newer-link"]/@href'))

}

# get consumption summary
get_consumption = function (doc) {

  # find a line in the post-body starting with Gelati
  # trim leading spaces
  return(gsub('^ +', '',
    xpathSApply(doc, '//div[@class="post-body entry-content"]/text()[contains(., "Gelati")]', xmlValue, trim=TRUE)
  ))

}

# parse consumption statistics
parse_consumption = function (consumption, synonyms=NULL) {

  # regular expressions to identify:
  #  - keys (like Gelati:)
  #  - values (integers)
  key_regex = '\\w+(?=: )'  # word characters (one or more) followed by a colon and a space
  value_regex = '\\d+'  # integers of one or more digits

  # use perl regexes to enable zero-width lookahead
  key_matches = gregexpr(key_regex, consumption, perl=TRUE)

  # process values as the inversion of the key matches
  values = sapply(unlist(regmatches(consumption, key_matches, invert=TRUE))[-1], function(x) {
    sum(as.integer(unlist(regmatches(x, gregexpr(value_regex, x)))))
  })

  # extract keys and rename synonyms
  keys = unlist(regmatches(consumption, key_matches))
  keys = ifelse(keys %in% names(synonyms), synonyms[keys], keys)

  names(values) = keys

  return(values)

}

# set up synonyms to map to standardised labels
synonyms = c('Beer'='Drink', 'Wine'='Drink')

# start page
page = 'http://barbara-foodwinepilgrimage.blogspot.com.au/2013/03/day-1-palma-de-mallorca-gelati-1-beer-2.html'

# loop until there are no more pages
l = list()

while (!is.null(page)) {

  doc = parse_doc(page)

  stats = parse_consumption(get_consumption(doc), synonyms)

  l = append(l, list(stats))

  page = get_next_page(doc)

}

# summarise the results
consumption = as.data.frame(do.call(rbind, l))
consumption$Day = 1:nrow(consumption)
d = melt(consumption, id='Day', variable.name='Type', value.name='Amount')

p = ggplot(data=d, aes(x=Day, y=Amount)) +
    geom_bar(aes(colour=Type, fill=Type), stat='identity') +
    facet_grid(Type ~ .)
plot(p)

p = ggplot(data=d, aes(x=Amount)) +
    geom_bar(aes(colour=Type, fill=Type)) +
    facet_grid(Type ~ .)
plot(p)