#!/usr/bin/env ruby

# =comment_spam_stats.rb
#
# Simple stats tool for my weblog comment spam workflow, helps me figure out 
# which articles receive how much 'successful' spam.
#
# Uses Spotlight's mdfind commandline tool to search for a specific subset of 
# emails, then extracts a string pattern from each mail, and builds a ranking 
# of the captured strings.
#
# See http://dekstop.de/weblog/2006/07/spotlight_helps_fight_comment_spam/
#
# By Martin Dittus (martin_at_dekstop_dot_de), 2006-07-07
# -- last change: 2006-07-07

# =========
# = prefs =
# =========

@prefs = {
	
	# mdfind -onlyin parameter. Adjust to select a specific mail folder.
	:mail_dir => '~/Library/Mail/',
	
	# mdfind spotlight query. Adjust to filter against your specific criteria.
	:spotlight_query => "kMDItemTitle == '*New Comment*'",
	
	# string pattern we're extracting from each email.
	:url_pattern => 'http://dekstop.de(/weblog/2.*?/)index.html'
}

# ========
# = main =
# ========

urls = {}

# escape mail_dir for shell use
escaped_mail_dir = @prefs[:mail_dir].gsub(/([ @])/) { |s| '\\'+$1}

# get list of files
print "Searching..."
mdfind_output = `mdfind -onlyin #{escaped_mail_dir} \"#{@prefs[:spotlight_query]}\"`
mail_files = mdfind_output.split("\n")
puts " found #{mail_files.size} emails."

# extract URLs
puts "Parsing"
mail_files.each do |filename|
	text = File.read(filename.chomp)
	print '.'
	text.match(@prefs[:url_pattern]).captures.each do |url|
		urls[url] = (urls[url] || 0) + 1
	end
end
puts " done."

# output
puts "Ranking:"
urls.keys.sort_by { |k| -urls[k]}.each do |url|
	puts "#{urls[url]} times #{url}"
end