DBLP dump parser for generating multiple networks

dblp.rb

#!/usr/bin/env ruby

#
# dblp.rb --- Parse dblp.xml to generate graphs
#
# Usage:
#   dblp.rb author <dblp.xml     --> co-author network
#   dblp.rb key cite <dblp.xml   --> citation network
#   dblp.rb key author <dblp.xml --> paper-author network
#   dblp.rb key venue <dblp.xml  --> paper-venue network
#

$type = nil
$date = nil
$key = nil

$venues = []
$authors = []
$cites = []

def getit(s)
  if s == "key"
    return [$key]
  elsif s == "venue"
    return $venues
  elsif s == "author"
    return $authors
  elsif s == "cite"
    return $cites
  else
    throw "What is #{s}???"
  end
end

def output
  if ARGV.length == 1
    as = getit(ARGV[0])
    as.each_with_index do |a, i|
      as[0, i].each do |b|
        puts "#{$date}\t#{b}\t#{a}"
      end
    end
  else
    ls = getit(ARGV[0])
    rs = getit(ARGV[1])

    ls.each do |l|
      rs.each do |r|
        puts "#{$date}\t#{l}\t#{r}"
      end
    end
  end
end

if $0 == __FILE__
  while line = $stdin.gets
    if /^<(\S*) mdate=\"([\d\-]*)\" key=\"([^\"]*)\"/ =~ line   # Beginning of a different item
      output

      $type = $1
      $date = $2
      $key = $3

      $venues = []
      $authors = []
      $cites = []
    elsif /^<journal>(.*)<\/journal>$/ =~ line || /^<booktitle>(.*)<\/booktitle>$/ =~ line
      $venues.push($1)
    elsif /^<author>(.*)<\/author>$/ =~ line
      # $authors.push($1.gsub(/\s/, "_"))
      $authors.push($1)
    elsif /^<cite>(.*)<\/cite>$/ =~ line && $1 != "..."
      $cites.push($1)
    end
  end

  output  # For the last item
end