<?php
require_once '../../../classes/CreateDocx.inc';
$options = array('paragraph' => true, 'list' => true, 'table' => true, 'footnote' => true, 'endnote' => true, 'chart' => 0);
CreateDocx::DOCX2TXT('../../files/Text.docx', 'document_1.txt', $options);
function extract_text($file, $file0, $source_type, $url, $chrSet)
{
global $db_con, $tmp_dir, $pdftotext_path, $catdoc_path, $xls2csv_path, $op_system, $mb, $debug;
global $catppt_path, $home_charset, $command_line, $no_log, $clear, $converter_dir, $cl, $index_xmeta;
$result = array();
$home_charset1 = str_ireplace('iso-', '', $home_charset);
$charset_int = str_ireplace('iso', '', $home_charset1);
$temp_file = "tmp_file";
$filename = $tmp_dir . "/" . $temp_file;
if ($source_type == 'ods') {
$filename .= "." . $source_type . "";
}
if ($source_type == 'doc') {
$filename .= "." . $source_type . "";
}
if ($source_type == 'docx') {
$filename .= "." . $source_type . "";
}
if ($source_type == 'xlsx') {
$filename .= "." . $source_type . "";
}
if (!($handle = fopen($filename, 'w'))) {
die("Cannot open file {$filename} in temp folder");
}
mysqltest();
if (fwrite($handle, $file) === FALSE) {
die("Cannot write to file {$filename} in temp folder");
}
fclose($handle);
mysqltest();
// for PDF documents enter here
if ($source_type == 'pdf') {
/*
include('../converter/pdf2text.php');
$a = new PDF2Text();
$a->setFilename($filename);
$a->decodePDF();
echo $a->output();
*/
if (!($handle = fopen($pdftotext_path, 'rb'))) {
printStandardReport('errorNoPDFConv', $command_line);
$result[] = 'ERROR';
} else {
// prepare command line for PDF converter
if ($op_system != 'win') {
$command = "" . $pdftotext_path . " -enc UTF-8 " . $filename . "";
} else {
$command = "" . $pdftotext_path . " -cfg xpdfrc " . $filename . " -";
}
$a = exec($command, $result, $retval);
// convert the PDF document
if ($retval != '0') {
// error handler for PDF file converter
if ($retval == '1' || $retval == '3' || $retval == '127') {
if ($retval == '1') {
printStandardReport('errorOpenPDF', $command_line);
}
if ($retval == '3') {
printStandardReport('permissionError', $command_line);
}
if ($retval == '127') {
printStandardReport('noConverter', $command_line);
}
} else {
printStandardReport('ufoError', $command_line);
}
$result[] = 'ERROR';
}
$result = implode(' ', $result);
}
// for DOC and RTF files enter here
} else {
if ($source_type == 'doc' || $source_type == 'rtf') {
/*
echo "\r\n\r\n<br /> op_system: '$op_system'<br />\r\n";
echo "\r\n\r\n<br /> catdoc_path: '$catdoc_path'<br />\r\n";
echo "\r\n\r\n<br /> charset_int: '$charset_int'<br />\r\n";
echo "\r\n\r\n<br /> filename: '$filename'<br />\r\n";
*/
if ($op_system == 'win') {
$command = "" . $catdoc_path . " -s " . $charset_int . " -d utf-8 -x " . $filename . "";
$a = exec($command, $result, $retval);
if (stristr($result[0], "catdoc.exe")) {
printDocReport($result[0], $cl);
}
} else {
$message = " Indexing of .doc and .rtf documents is currently not supported on LINUX OS.";
printDocReport($message, $cl);
/*
$retval = '';
$catdoc_path = str_ireplace("catdoc.exe", "catdoc.lin", $catdoc_path);
//$command = "".$catdoc_path." -cfg xpdfrc ".$filename." -";
$command = "".$catdoc_path." -s ".$charset_int." -d utf-8 -w -x ".$filename."";
$a = exec($command, $result, $retval); // convert the DOC document
//echo "\r\n\r\n<br /> retval: '$retval'<br />\r\n";
//echo "\r\n\r\n<br>result Array:<br><pre>";print_r($result);echo "</pre>\r\n";
if ($retval) {
$result = 'ERROR';
//echo "\r\n\r\n<br /> retval: '$retval'<br />\r\n";
if($retval == '2') {
$message = " File to be converted not found";
printDocReport($message, $cl);
}
else if($retval == '3') {
$message = " Path to file not found.";
printDocReport($message, $cl);
}
else if($retval == '11') {
$message = " The executable is corrupted.";
printDocReport($message, $cl);
}
else if($retval == '12') {
$message = " Out of memory execution.";
printDocReport($message, $cl);
}
else if($retval == '22') {
$message = " dll error";
printDocReport($message, $cl);
}
else if($retval == '31') {
$message = " The association is missing, use Shell to try the OpenWith dialog.";
printDocReport($message, $cl);
}
else if($retval == '32') {
$message = " File could not be opened.";
printDocReport($message, $cl);
}
else if($retval == '126') {
$message = " Command invoked cannot execute (Permission problem or command is not an executable).";
printDocReport($message, $cl);
}
else if($retval == 127) {
$message = " Command not found.";
printDocReport($message, $cl);
}
else if($retval == 128) {
$message = " Invalid argument to exit. Exit takes only integer range 0 – 255.";
printDocReport($message, $cl);
}
else if($retval > 128 && $retval < 255) {
$message = " Fatal error code $retval";
printDocReport($message, $cl);
}
else if($retval == 255) {
$message = " Exit status out of range. Exit takes only integer range 0 – 255.";
printDocReport($message, $cl);
} else {
$message = " Unknown error code $retval.";
printDocReport($message, $cl);
}
}
*/
}
// for PPT files enter here
} else {
if ($source_type == 'ppt') {
// currently unsupported,as a failure was encountered for large PowerPoint presentations
$a = '';
/*
$command = $catppt_path." -s $charset_int -d utf-8 $filename";
$a = exec($command, $result, $retval);
*/
// for XLS spreadsheets enter here
} else {
if ($source_type == 'xls') {
$error = '';
require_once "" . $converter_dir . "/xls_reader.php";
$data = new Spreadsheet_Excel_Reader();
if ($mb == '1') {
// if extention exists, change 'iconv' to mb_convert_encoding:
$data->setUTFEncoder('mb');
}
// set output encoding.
$data->setOutputEncoding('UTF-8');
// read this document
$data->read($filename);
$error = $data->_ole->error;
if ($error == '1') {
printStandardReport('xlsError', $command_line, $no_log);
$result = 'ERROR';
} else {
$result = '';
$boundsheets = array();
$sheets = array();
$boundsheets = $data->boundsheets;
// get all tables in this file
$sheets = $data->sheets;
// get content of all sheets in all tables
if ($boundsheets) {
foreach ($boundsheets as &$bs) {
$result .= "" . $bs['name'] . ", ";
// collect all table names in this file
}
if ($sheets) {
foreach ($sheets as &$sheet) {
$cells = $sheet['cells'];
if ($cells) {
// ignore all empty cells
foreach ($cells as &$cell) {
foreach ($cell as &$content) {
$result .= "" . $content . ", ";
// collect content of all cells
}
}
}
}
}
if (strtoupper($home_charset) == 'ISO-8859-1') {
$result = utf8_encode($result);
}
}
}
// for ODS spreadsheets enter here
} else {
if ($source_type == 'ods') {
require_once "" . $converter_dir . "/ods_reader.php";
$reader = ods_reader::reader($filename);
$sheets = $reader->read($filename);
if ($sheets) {
$result = '';
foreach ($sheets as &$sheet) {
if ($sheet) {
foreach ($sheet as &$cell) {
if ($cell) {
// ignore all empty cells
foreach ($cell as &$content) {
$result .= "" . $content . " ";
// collect content of all cells
}
}
}
}
}
} else {
$result = 'ERROR';
}
// for ODT documents enter here
} else {
if ($source_type == 'odt') {
require_once "" . $converter_dir . "/odt_reader.php";
$x = new odt_reader();
// Unzip the document
$u = $x->odt_unzip($filename, false);
// read the document
$result = $x->odt_read($u[0], 2);
// create some blanks around the <div> tags
$result = str_replace("<", " <", $result);
$result = str_replace(">", "> ", $result);
//echo "\r\n\r\n<br /> odt result: $result<br />\r\n";
// for DOCX files enter here
} else {
if ($source_type == 'docx') {
// converter class supplied by http://www.phpdocx.com
$options = array('paragraph' => false, 'list' => false, 'table' => false, 'footnote' => false, 'endnote' => false, 'chart' => 0);
$docx_file = "docx.txt";
$result = '';
require_once "" . $converter_dir . "/docx/CreateDocx.inc";
CreateDocx::DOCX2TXT($filename, $tmp_dir . "/" . $docx_file, $options);
if ($file = @file_get_contents($tmp_dir . "/" . $docx_file)) {
$result = "{$file} ";
}
if ($index_xmeta) {
require_once "" . $converter_dir . "/xmeta_converter.php";
$docxmeta = new x_metadata();
$docxmeta->setDocument($filename);
/*
echo "Title : " . $docxmeta->getTitle() . "<br>";
echo "Subject : " . $docxmeta->getSubject() . "<br>";
echo "Creator : " . $docxmeta->getCreator() . "<br>";
echo "Keywords : " . $docxmeta->getKeywords() . "<br>";
echo "Description : " . $docxmeta->getDescription() . "<br>";
echo "Last Modified By : " . $docxmeta->getLastModifiedBy() . "<br>";
echo "Revision : " . $docxmeta->getRevision() . "<br>";
echo "Date Created : " . $docxmeta->getDateCreated() . "<br>";
echo "Date Modified : " . $docxmeta->getDateModified() . "<br>";
*/
$result .= $docxmeta->getTitle() . $docxmeta->getSubject() . $docxmeta->getCreator() . $docxmeta->getKeywords() . $docxmeta->getDescription() . $docxmeta->getLastModifiedBy() . $docxmeta->getRevision() . $docxmeta->getDateCreated() . $docxmeta->getDateModified();
}
@unlink($tmp_dir . "/" . $docx_file);
/*
if($result && $chrSet != "UTF-8") {
$result = @mb_convert_encoding($result, "UTF-8", $chrSet);
}
*/
// for XLSX spreadsheets enter here
} else {
if ($source_type == 'xlsx') {
$result = '';
$i = 1;
$name = '';
$finished = false;
$names = array();
require_once "" . $converter_dir . "/xlsx_reader.php";
$xlsx = new SimpleXLSX($filename);
$names = $xlsx->sheetNames();
//echo "\r\n\r\n<br>names array:<br><pre>";print_r($names);echo "</pre>\r\n";
if ($debug == 2 && $names) {
printXLSXreport(count($names), $cl);
}
foreach ($names as $my_name) {
$result .= $my_name . " ";
if ($debug == 2) {
printActKeyword($my_name);
}
}
while (!$finished) {
// get all sheets
if ($rows = $xlsx->rows($i)) {
foreach ($rows as $key) {
foreach ($key as $val) {
if ($val) {
$result .= " " . $val;
// add value of each cell
}
}
}
} else {
$finished = true;
// no more sheets found
}
//$my_name = $xlsx->sheetName($i);
//echo "\r\n\r\n<br /> sheet name $i: '$my_name'<br />\r\n";
$i++;
// try to get next sheet
}
if ($index_xmeta) {
require_once "" . $converter_dir . "/xmeta_converter.php";
$xlscxmeta = new x_metadata();
$xlscxmeta->setDocument($filename);
/*
echo "Title : " . $xlscxmeta->getTitle() . "<br>";
echo "Subject : " . $xlscxmeta->getSubject() . "<br>";
echo "Creator : " . $xlscxmeta->getCreator() . "<br>";
echo "Keywords : " . $xlscxmeta->getKeywords() . "<br>";
echo "Description : " . $xlscxmeta->getDescription() . "<br>";
echo "Last Modified By : " . $xlscxmeta->getLastModifiedBy() . "<br>";
echo "Revision : " . $xlscxmeta->getRevision() . "<br>";
echo "Date Created : " . $xlscxmeta->getDateCreated() . "<br>";
echo "Date Modified : " . $xlscxmeta->getDateModified() . "<br>";
*/
$result .= $xlscxmeta->getTitle() . $xlscxmeta->getSubject() . $xlscxmeta->getCreator() . $xlscxmeta->getKeywords() . $xlscxmeta->getDescription() . $xlscxmeta->getLastModifiedBy() . $xlscxmeta->getRevision() . $xlscxmeta->getDateCreated() . $xlscxmeta->getDateModified();
}
/*
if($result && $chrSet != "UTF-8") {
$result = @mb_convert_encoding($result, "UTF-8", $chrSet);
}
*/
// for JavaScript enter here
} else {
if ($source_type == 'js') {
$result = extract_js($file);
}
}
}
}
}
}
}
}
}
if ($result != 'ERROR') {
if (is_array($result)) {
$result = implode(" ", $result);
}
$count = strlen($result);
if ($count == '0') {
// if there was not one word found, print warning message
if ($source_type == 'js') {
printStandardReport('jsEmpty', $command_line, $no_log);
} else {
printStandardReport('nothingFound', $command_line, $no_log);
}
$result = 'ERROR';
}
}
unlink($filename);
mysqltest();
if ($clear == 1) {
unset($command, $retval, $a, $file, $count);
}
return $result;
}