subtitle-cleanup

#!/bin/bash

HOMEBASE=$(pwd)
__exists () {
      which $1 >/dev/null 2>&1;
}
__exists wget && getCommand="wget" && getArgs="-O"
__exists fetch && getCommand="fetch" && getArgs="-o"

grab() {
  $getCommand $getArgs "$1" "$2" || exit $?
}

utf8() {
  iconv -f latin1 -t UTF-8 -o "$1"
}

find -name '*html' | while read file; do
  id=`echo $file | sed -e 's/.*\/\([-0-9]*\)*.html/\1/'`
  dir=`dirname $file`
  xml="$id.xml"
  cd $dir
  grab - "http://video.google.com/videotranscript?docid=$id&type=list" | utf8 "$xml"
  cat -- "$xml" | tr ">" "\n" | grep "^<track " | while read line; do
    tid=`echo $line | sed -n -e 's/.*id="\([^"]*\)".*/\1/p'`
    name=`echo $line | sed -n -e 's/.*name="\([^"]*\)".*/\1/p'`
    lang=`echo $line | sed -n -e 's/.*lang_code="\([^"]*\)".*/\1/p'`
    lang_name=`echo $line | sed -n -e 's/.*lang_translated="\([^"]*\)".*/\1/p'`
    echo "Found $lang_name subtitle track $tid named $name..."
    tfile="$id.$tid.xml"
    grab - "http://video.google.com/videotranscript?docid=$id&type=track&name=$name&lang=$lang" | utf8 "$tfile"
  done
  cd $HOMEBASE
  sleep 1
done