Convert a PDF to text, HTML and XML with various tools

pdf-reader.rb

#!/usr/bin/env ruby

require 'pdf-reader'

infile=ARGV[0]
outfile=ARGV[1]

reader = PDF::Reader.new(infile)
text = ""
reader.pages.each do |page|
  text << page.text
end

File.open(outfile, 'w') do |file|
  file.write(text)
end

script.sh

#!/bin/sh

tdocsplit="/usr/local/Cellar/poppler/0.20.0/bin/pdftotext"
txpdf="/usr/local/Cellar/xpdf/3.03/bin/pdftotext"
tpdfreader="./pdf-reader.rb"
hpdftohtml="/usr/local/Cellar/pdftohtml/0.40a/bin/pdftohtml"
hdocsplit="/usr/local/Cellar/poppler/0.20.0/bin/pdftohtml"

rm_blank_lines() {
    _infile=$1
    _outfile=$2
    sed '/^[ \t]*$/d' $_infile > $_outfile
}

totext() {
    _command=$1
    _cmdname=`echo $_command | cut -f 5 -d "/"`
    if [ -z $_cmdname ]; then
        _cmdname="pdfreader"
    fi
    _infile=$INFILE
    _infilename=`echo $_infile | cut -f 1 -d .`
    _filename="${_infilename}_${_cmdname}.txt"
    _tmpfile="$TEMP_DIR/${_filename}"
    _outfile="$OUTPUT_DIR/${_filename}"
    $_command $_infile $_tmpfile
    rm_blank_lines $_tmpfile $_outfile
}

tohtml() {
    _command="$1 -noframes -i -c"
    _cmdname=`echo $1 | cut -f 5 -d "/"`
    _infile=$INFILE
    _infilename=`echo $_infile | cut -f 1 -d .`
    _filename="${_infilename}_${_cmdname}.html"
    _tmpfile="$TEMP_DIR/${_filename}"
    _outfile="$OUTPUT_DIR/${_filename}"
    $_command $_infile $_tmpfile
    rm_blank_lines $_tmpfile $_outfile
}

toxml() {
    _command="$1 -xml -i"
    _cmdname=`echo $1 | cut -f 5 -d "/"`
    _infile=$INFILE
    _infilename=`echo $_infile | cut -f 1 -d .`
    _filename="${_infilename}_${_cmdname}.xml"
    if [ $_cmdname == "pdftohtml" ]; then
        _filename="${_infilename}_${_cmdname}"
    fi
    _tmpfile="$TEMP_DIR/${_filename}"
    _outfile="$OUTPUT_DIR/${_filename}"
    $_command $_infile $_tmpfile
    if [ $_cmdname == "pdftohtml" ]; then
        rm_blank_lines "${_tmpfile}.xml" "${_outfile}.xml"
    else
        rm_blank_lines $_tmpfile $_outfile
    fi
}

INFILE=$1
OUTPUT_DIR="compare"
TEMP_DIR="${OUTPUT_DIR}/tmp"

if [ -z "$INFILE" ]; then
    echo "Usage: $0 file.pdf"
    exit 1
fi

if [ ! -d "$COMPARE_DIR" ]; then
    mkdir -p "$TEMP_DIR"
fi


totext $tdocsplit
totext $txpdf
totext $tpdfreader

tohtml $hpdftohtml
tohtml $hdocsplit

toxml $hpdftohtml
toxml $hdocsplit