#!/usr/bin/ruby -Ke
WDCNT_VERSION = "2000-01-23"

=begin
= wdcnt -- word counter for English/Japanese text file. 

== SYNOPSIS

((%wdcnt [-r] [-p|-z] [-e] ((|files|)) ...%))

((%wdcnt [-r] [-p|-z] [-e] < ((|file|))%))

((%wdcnt -v%))

== DESCRIPTION

(('wdcnt')) counts reports English or Japanese words in files or
standard input. (('wdcnt')) ignores punctuation, digits, quote signs
or HTML tags.  The output is sorted in the order of the occurrence
frequency and can be plotted directly by (('gnuplot(1)')) as follows.

  gnuplot> set log xy
  gnuplot> plot "< wdcnt file"

== OPTIONS

: -p 
  Reports probability instead of number of occurrences. 
  Each frequency is normalized by 1.0. 

: -z
  Reports relative frequency instead of number of occurrences. 
  1.0 for the most occurring word. 

: -r
  Puts rank to the head of each line. 

: -e
  Does not use KAKASI.  This option is NOT useful to Japanese documents.

: -v, -h
  Prints usage and version then exit. 

== HISTORY

For English document, a traditional one-liner is known: 

  % tr -s '\040' '\012' files ... | sort -n | uniq -c | sort -n -r

== SEE ALSO

(('Ruby/KAKASI')) ((<URL:http://www.ruby-lang.org/en/raa.html#Ruby%2FKAKASI>)),
(('ruby(1)')) ((<URL:http://www.ruby-lang.org/>)), 
(('kakasi(1)')) ((<URL:http://kakasi.namazu.org/>)), 
(('gnuplot(1)')), (('tr(1)')), (('sort(1)')), (('uniq(1)'))

== BUGS

Word separation is not accurate. 

== AUTHOR

Gotoken ((<URL:mailto:gotoken@notwork.org>))

=end
require "getopts"
getopts("epzvhr")
ME = File.basename($0)

if $OPT_v || $OPT_h || ($OPT_p && $OPT_z)
  puts "#{ME} - version #{WDCNT_VERSION}"
  puts <<EOS
USAGE:
  #{ME} [-r] [-p|-z] [-e] files ...
  #{ME} [-r] [-p|-z] [-e] < file
  #{ME} -v
OPTIONS:
  -p: reports as probability.
  -z: reports relative frequency (1.0 for most word).
  -e: does not use KAKASI.
  -v: prints this message. 
EOS
  exit
end

if $OPT_e
  def kakasi(opt, src); src end
else
  require "kakasi"
  include Kakasi
end

files = ARGV.dup

# delete html-tag or symbol
txt = kakasi("-oeuc -Ea", $<.read || "")
txt.gsub!(/<[^>]*>/p, " ")
txt.gsub!(/^\.\S+\s+/, "")
txt.gsub!(/[\/<>\s\n\r\d%\(\),:;~"`]+/p, " ") # "
txt.gsub!(/(?!\w)[-'\.]\W/, " ") # '
words = kakasi("-ieuc -w", txt).split.sort!

n = 0.0
w0, c, wc = words.shift, 1, []
w0 or (STDERR.puts "#{ME}: nodata??"; exit)

# counting words
words.each{|w|
  if w != w0
    wc.push [c,w0]
    n += c
    c = 1
  else
    c += 1
  end
  w0 = w
}
wc.push [c, w0]
n += c

# report
puts "# creator: #{ME} - version #{WDCNT_VERSION}"
puts "# date   : #{Time.now}"
puts "# source : #{files.join(',')}"
puts "# comment: KAKASI was not used." if $OPT_e
puts "# summary: #{wc.size} different words in #{n.to_i} words."

wc.sort!.reverse!.each_with_index{|d,i| 
  fmt, data = if $OPT_r
		["%5d ", [i+1]]
	      else
		["", []]
	      end
  if $OPT_p
    puts "#{fmt}%.5f # %s" % (data + [d[0]/n, d[1]])
  elsif $OPT_z
    puts "#{fmt}%.5f # %s" % (data + [d[0]/wc[0][0].to_f, d[1]])
  else
    puts "#{fmt}%4d # %s" % (data + d)
  end
}