#!/usr/bin/perl
# clean Word HTML
# by mitcho, 20091230
# http://mitcho.com

$_ = `cat ${ARGV[0]}`;

s!\n!\r!gm;

s! class=Section\d+!!gm;
s! class=Mso\w+!!gm;
s! class=SpellE!!gm;

s!mso-[^:]+:[^;'"]+;?!!gm;
s!mso-[^:]+:[^;]+;!!gm;
s!font-weight:\s*normal;?!!gm;
s!font-style:\s*normal;?!!gm;
s!font-family:\s*"Times New Roman";?!!gm;
s!font-family:\s*TimesNewRomanPSMT;?!!gm;
s!\s*style=''!!gm;
s!\s*style=""!!gm;

# there are a bunch of tags with o or v namespaces. get rid of them!
s!</?[ov]:[^>]+>!!gm;

# omg overlapping id's for style!!
s!id=!class=!gm;

# remove IE/MS style conditional comments
s?<!\[if [^\]]+\]>??gm;
s?<!\[endif\]>??gm;

# don't use blank paragraphs for spacing!
s!\r<p>(&nbsp;|\s)*</p>\r!\r!gm;

# remove spans with nothing in them anymore.
s!<span\s*>([^<]*)</span>!$1!gm;
s!<span\s*>([^<]*)</span>!$1!gm; #twice for good measure

# remove newlines within paragraphs
s!(\w)\s*\r(\w)!$1 $2!gm;

#m!<body[^>]*>(.*)</body>!m;
#$_ = $1;

print;