Tryag File Manager

/home/cluster1/data/bu01/1121861/html/jlex/php4/db_modifier.php4

<?
include_once("schema_loader.php4");

/**
 * db_modifier produces a modified version of an xml document. 
 *
 * Modifications include:
 * <UL>
 * <LI> cDATA symbol substitution: substituting one type of symbol for another, e.g. 'a' for 'á'. The new content 
 * is placed within a new tag with a '_s', e.g. <field>á</field> becomes <field_s>a</field_s>.</LI>
 * <LI> Field removal: specified xml tags and associated content can be discluded in the modified xml document.</LI>
 * <LI> alphabetization tag: A new tag is created <alpha> containing a value which will be used to sort query results.
 * The value will be determined by a set of conditions supplied by the user.</LI>
 * </UL>
 * 
 * The parsing ability of this class actually provides two seperate but related parses.
 * <UL>
 * <LI>The first function produces a list of all the field types that exist within the xml document.</LI>
 * <LI>The second function does the actual modifications and produces a new XML document.</LI>
 * </UL>
 * I chose to use the same class to provide both functions because it seemed a waste to include an entire new 
 * class to provide the functionality of creating a list of fields. In addition the results from this aggregation
 * are used by the modification function indirectly. The search functionality reports back not only which fields
 * exist in the XML document but also which fields contain characters contained in the $symbols and $first_chars arrays.
 * This information is generally published to the user, allowing him to select which fields he wants to include in the
 * MySQL version of his XML file and which fields he may want to add substituted versions of.
 * Perhaps in a future version this class should be split into two. 
 * NOTE: In PHP4, you do not need to first decode cdata from utf8. PHP5 assumes the source is in utf8, so you must 
 * first decode the cdata.
 *
 * @package JLex
 * @author Jonathan Dick
 * @date Dec. 30, 2005
 */
class db_modifier {

/**#@+          
     * @access public          
     */    
    var $parser;

/**
     * The file pointer for writing to disk.
     * @var resource
     */
    var $out;

/**
     * The current cDATA.
     * @var string
     */
    var $cur_data;

/**
     * The symbols to be substituted for. Note that I put these symbols and the substitutes in seperate arrays in order
     * to utilize the php function str_replace which takes in a symbols array and a substitutes array.
     * @var array of strings
     */
    var $symbols;

/**
     * An array of the substitue symbols.
     * @var array of strings
     */
    var $substitutes;

/**
     * The user may supply substitution rules which only apply to the first character. This array contains
     * the symbols which will be substituted for if found in the first character position.
     * @var array of strings
     */
    var $first_chars;

/**
     * This is a regular expression composed of the values from the $symbols and $first_chars array.
     *
     * It is used to determine if the content of a particular field meets the conditions specified by 
     * the $symbols and $first_chars arrays, i.e. if it contains symbols found in those arrays.
     * @var string
     */
    var $pattern;

/**
     * The set of symbols to be replaced from the alpha field.
     * @var array of strings
     */
    var $alpha_symbols;

/**
     * The set of substitutes for the $alpha_symbols.
     * @var array of strings
     */
    var $alpha_substitutes;

/**
     * The set of first characters which should be substituted for in the alpha field.
     * @var array of strings
     */
    var $alpha_first_chars;

/**
     * The regular expression representing the conditions set forth in the $alpha_symbols and $alpha_first_chars arrays.
     * @var string
     */
    var $alpha_pattern;

/**
     * The fields which will be used to create the alpha column. 
     * 
     * The first field will be used in the alpha column if it exists. If that field does not exist
     * then the second field will be used and so on. The rules specified by the $alpha_symbols and
     * $alpha_first_chars will then be applied to the value of the field selected.
     * @var array of strings
     */
    var $sort_fields;

/**
     * A boolean variable which is true when sort fields are provided by the user.
     *
     */
    var $has_sort_fields;

/**
     * The current value of the alpha column.
     *
     * Because the document is parsed in a linear fashion, we don't know if a field exists 
     * until we parse that field. Consequently, if any sort field is encountered we must save
     * the value and only discard once we reach a sort field of higher precedence.
     */
    var $alpha;

/**
     * The index within the array of the field whose content currently exists in the $alpha variable.
     *
     * When a field contained in the $sort_fields array is parsed, it's index is retrieved from the array
     * and compared to the value in $cur_alpha_index. If it is lower, the value of this field will replace
     * the content currently stored in $alpha.
     *
     * @var integer
     */
    var $cur_alpha_index;

/**
     * The XML tag marking the beginning of an entry.
     * @var string
     */
    var $head_tag;

/**
     * This boolean variable is true when the function being performed by this parse is modification.
     * @var boolean
     */
    var $produce_stripped_version;

/**
     * The set of fields in the XML document which contain cDATA obeying the rules found in the
     * $symbols and $first_chars arrays.
     *
     * This array is filled when parsing for field types, i.e. the first functionality discussed above.
     */
    var $strippable_fields;

/**
     * This variable holds two types of fieldsets depending on the function being implemented. If function 1
     * (field search) is being used, this variable contains the growing list of fields found within entries in
     * the XML document. If function 2 (modification) is being used, this contains the set of fields to be 
     * included in the modified version of the XML document.
     * @var array
     */
    var $fields;

/**
     * The set of fields for which modified versions should be created based on the rules in the 
     * $symbols and $first_chars array.
     * @var array
     */
    var $fields_to_strip;

/**
     * Boolean variable which is set to true one a start $head_tag is found and false when an end $head_tag is parsed.
     * @var boolean
     */
    var $in_entry;

/**#@-*/

function db_modifier() 
    {
	$this->fields_to_strip = array();
	$this->check_all_fields = false;
	$this->cur_data = "";
	$this->symbols = array();
	$this->substitutes = array();
	$this->pattern = "";

$this->alpha_symbols = array();
	$this->alpha_substitutes = array();
	$this->alpha_first_chars = array();
	$this->alpha_pattern = "";

$this->sort_fields = array();
	$this->has_sort_fields = false;
	$this->alpha = "";
	$this->cur_alpha_index = 10;

$this->strippable_fields = array();
	$this->fields = array();
	$this->fields_to_strip = array();
    }

/**
     * Extracts strings from space-seperated string and inputs them into the $sort_fields array.
     * @param string $sort_fields A space seperated list of fields contained in the XML document.
     * @return void The fields are inputted into the global variable $sort_fields.
     */
    function get_sort_fields($sort_fields) {
	$fields = explode(" ",$sort_fields);
	foreach($fields as $field) {
	    $this->sort_fields[] = $field;
	}
    }

/**
     * Extracts fields, one per line, from a file and returns them in an array.
     * 
     * This function is no longer used.
     * @param string $fields The filename containing the fields to be extracted.
     * @return array The fields contained within the file.
     */
    function get_fields_to_strip($fields) {
	$in = fopen($fields,"r");
	while($line = fgets($in)) {
	    $field = trim($line);
	    $fields_to_strip[] = $field;
	}
	return $fields_to_strip;
    }

/**
     * This function takes user input specifying substitution rules contained in a textarea within an HTML form
     * and inputs the symbols into the $alpha_symbols array and their substitutes into the $alpha_substitutes array.
     *
     * Each line of the textarea should (if properly formatted by the user) a substitution rule in the form "X = Y".
     * Additionally, the '' is used to specify that a symbol will be deleted rather than subsituted for, e.g. "X = ''". 
     * A user may specify multiple symbols to be replaed by a single substitute, e.g. "WXY = Z". 
     * Lastly, all the symbols which need to be subsituted for are concatenated and made into a regular expression
     *
     * @param string $textarea A textarea from an HTML form.
     * @return void Only global variables are affected, no results are returned.
     */
    function make_alpha_symbols_table($textarea) {
	$textarea = stripslashes($textarea);
	$textarea = trim(utf8_decode($textarea));
	$lines = explode("\n",$textarea);	
	foreach($lines as $line) {
	    $vals = explode(" = ",$line);
	    $substitute = trim($vals[1]);
	    if($substitute == "''") {
		$substitute = "";
	    }
	    $symbols = preg_split('//', $vals[0], -1, PREG_SPLIT_NO_EMPTY);
	    foreach($symbols as $symbol) {
		$this->alpha_symbols[] = $symbol;
		$this->alpha_substitutes[] = $substitute;
	    }
	}
	$symbols_string = implode("",$this->alpha_symbols);
	if(substr($symbols_string,-1,1) == "'") {
	    $symbols_string = "'".substr($symbols_string,0,-1);
	}
	if($symbols_string != "") {
	    $this->alpha_pattern = "[".$symbols_string."]";
	}
    }

/**
 * This function takes user input specifying substitution rules contained in a textarea within an HTML form 
 * and inputs the results into $alpha_first chars, an associative array indexed by the symbol to be replaced 
 * and containing the substitute symbol.
 *
 * Note that in this case, two arrays are not used to store the symbols. This is because only the first character
 * should be substituted for, not a set of characters. 
 * Each line of the textarea should (if properly formatted by the user) a substitution rule in the form "X = Y".
 * Additionally, the '' is used to specify that a symbol will be deleted rather than subsituted for, e.g. "X = ''". 
 * A user may specify multiple symbols to be replaed by a single substitute, e.g. "WXY = Z". 
 * Lastly, all the symbols which need to be subsituted for are concatenated and made into a regular expression
 *
 * @param string $textarea A textarea from an HTML form.
 * @return void Only global variables are affected, no results are returned.
 */
 function make_alpha_first_chars_table($textarea) {
	$textarea = trim(stripslashes($textarea));
	$textarea = utf8_decode($textarea);
	$lines = explode("\n",$textarea);
	foreach($lines as $line) {
	 $vals = explode(" = ",$line);
	 $substitute = trim($vals[1]);
	 if($substitute == "''") {
		$substitute = "";
	 }
	 $symbols = preg_split('//', $vals[0], -1, PREG_SPLIT_NO_EMPTY);
	 foreach($symbols as $symbol) {
		$this->alpha_first_chars[$symbol] = $substitute;
	 }
	}
	$first_chars_string = trim(implode("",array_keys($this->alpha_first_chars)));
	if(substr($first_chars_string,-1,1) == "'") {
	 $first_chars_string = "'".substr($first_chars_string,0,-1);
	}
	if($first_chars_string != "") {
	 if($this->alpha_pattern != "") {
		$this->alpha_pattern .= "|";
	 }
	 $this->alpha_pattern .= "^[".$first_chars_string."]";
	}
 }

/**
     * This function takes user input specifying substitution rules contained in a textarea within an HTML form  
     * and inputs the symbols into the $symbols array and their substitutes into the $substitutes array.
     *
     * Each line of the textarea should (if properly formatted by the user) a substitution rule in the form "X = Y".
     * Additionally, the '' is used to specify that a symbol will be deleted rather than subsituted for, e.g. "X = ''".
     * A user may specify multiple symbols to be replaed by a single substitute, e.g. "WXY = Z". 
     * Lastly, all the symbols which need to be subsituted for are concatenated and made into a regular expression
     *
     * @param string $textarea A textarea from an HTML form.
     * @return void Only global variables are affected, no results are returned.
     */
    function make_symbols_table($textarea) {
	$textarea = stripslashes($textarea);
	$textarea = trim(utf8_decode($textarea));
	$lines = explode("\n",$textarea);	
	foreach($lines as $line) {
	    $vals = explode(" = ",$line);
	    $substitute = trim($vals[1]);
	    if($substitute == "''") {
		$substitute = "";
	    }
	    $symbols = preg_split('//', $vals[0], -1, PREG_SPLIT_NO_EMPTY);
	    foreach($symbols as $symbol) {
		$this->symbols[] = $symbol;
		$this->substitutes[] = $substitute;
	    }
	}
	$symbols_string = implode("",$this->symbols);
	if(substr($symbols_string,-1,1) == "'") {
	    $symbols_string = "'".substr($symbols_string,0,-1);
	}
	if($symbols_string != "") {
	    $this->pattern = "[".$symbols_string."]";
	}
    }

/**
 * This function takes user input specifying substitution rules contained in a textarea within an HTML form 
 * and inputs the results into $first chars, an associative array indexed by the symbol to be replaced 
 * and containing the substitute symbol.
 *
 * Note that in this case, two arrays are not used to store the symbols. This is because only the first character
 * should be substituted for, not a set of characters. 
 * Each line of the textarea should (if properly formatted by the user) a substitution rule in the form "X = Y".
 * Additionally, the '' is used to specify that a symbol will be deleted rather than subsituted for, e.g. "X = ''". 
 * A user may specify multiple symbols to be replaed by a single substitute, e.g. "WXY = Z". 
 * Lastly, all the symbols which need to be subsituted for are concatenated and made into a regular expression
 *
 * @param string $textarea A textarea from an HTML form.
 * @return void Only global variables are affected, no results are returned.
 */
 function make_first_chars_table($textarea) {
	$textarea = trim(stripslashes($textarea));
	$textarea = utf8_decode($textarea);
	$lines = explode("\n",$textarea);
	foreach($lines as $line) {
	 $vals = explode(" = ",$line);
	 $substitute = trim($vals[1]);
	 if($substitute == "''") {
		$substitute = "";
	 }
	 $symbols = preg_split('//', $vals[0], -1, PREG_SPLIT_NO_EMPTY);
	 foreach($symbols as $symbol) {
		$this->first_chars[$symbol] = $substitute;
	 }
	}
	$first_chars_string = trim(implode("",array_keys($this->first_chars)));
	if(substr($first_chars_string,-1,1) == "'") {
	 $first_chars_string = "'".substr($first_chars_string,0,-1);
	}
	if($first_chars_string != "") {
	 if($this->pattern != "") {
		$this->pattern .= "|";
	 }
	 $this->pattern .= "^[".$first_chars_string."]";
	}
 }

/**
     * Replace the specified $value with the stripped version according to the conditions
     * specified by $alpha_symbols and $alpha_first_chars.
     *
     * @param string $value The content to be modified.
     * @return string The modified content.
     */
    function alpha_strip($value) {
	$stripped = str_replace($this->alpha_symbols,$this->alpha_substitutes,$value);
	$first_char = substr($stripped,0,1);
	while(@array_key_exists($first_char,$this->alpha_first_chars)) {
	    $stripped = $this->alpha_first_chars[$first_char].substr($stripped,1);
	    $first_char = substr($stripped,0,1);
	}
	return $stripped;
    }

/**
     * Replace the specified $value with the stripped version according to the conditions
     * specified by $alpha_symbols and $alpha_first_chars.
     *
     * @param string $value The content to be modified.
     * @return string The modified content.
     */
    function strip($value) {
	$stripped = str_replace($this->symbols,$this->substitutes,$value);
	$first_char = substr($stripped,0,1);
	while(@array_key_exists($first_char,$this->first_chars)) {
	    $stripped = $this->first_chars[$first_char].substr($stripped,1);
	    $first_char = substr($stripped,0,1);
	}
	return $stripped;
    }

/**
 * The function which handles a new start tag during XML parsing.
 *
 * If function 1 (field searching) is used, do nothing. 
 * If function 2 (modification) is used, write the element to the new, modified XML document if:
 * <UL>
 * <LI> The field exists withing $fields.</LI>
 * <LI> The field is outside of an entry.</LI>
 * </UL>
 */
 function startHandler($xp, $element, $attribs) {	
	$element = strtolower($element);
	if($element == $this->head_tag) {
	 $this->in_entry = true;	 
	}
	if($this->produce_stripped_version) {
	 if($this->in_entry) {
		if(in_array($element,$this->fields)) {
		 fwrite($this->out,"<$element>");
		}
	 }
	 else {
		fwrite($this->out,"<$element>\n");
	 }
	}
 }

/**
 * The function for handling an end tag during an XML parse.
 *
 * If function 2 (modification) is being used: 
 * <UL>
 * <LI>Check to see if the field is a sort field and make the changes as necessary.</LI>
 * <LI>If the element is a head tag, we are at then end of an entry. So, write the alpha column
 * into the modified XML document, set $alpha to the empty string and set $cur_alpha_index
 * to the highest number it can be (this ensures that the first sort field found will be used).</LI>
 * <LI>If the element is in $fields, write $cur_data to disk and a close tag for this field.</LI>
 * <LI>If $in_entry is false, write the cDATA and a close tag. Nothing should be removed from the original
 * XML document that is not explicitly specified by the modification rules.</LI>
 * </UL>
 * If function 1 (field search) is being used: if the element is not already there, add it to $fields. 
 * Lastly, check to see if the field meets the conditions set in $pattern. If there is a pattern match:
 * <UL> 
 * <LI>If in function 1, add the field (if it doesn't already exist) to $strippable_fields.</LI>
 * <LI>If in function 2: if the modified field exists in $fields_to_strip, write the modified field 
 * to the new XML document.</LI>
 * </UL>
 */
 function endHandler($xp, $element) {	
	$element = strtolower($element);
	if($this->produce_stripped_version) {
	 if($this->has_sort_fields && in_array($element, $this->sort_fields)) {
		$index = array_search($element,$this->sort_fields);
		if($index < $this->cur_alpha_index) {
		 $this->alpha = $this->alpha_strip($this->cur_data);
		 $this->cur_alpha_index = $index;
		}
	 }
	 if(($element == $this->head_tag) && $this->has_sort_fields) {
		fwrite($this->out,"<alpha>$this->alpha</alpha>\n");
		$this->alpha = "";
		$this->cur_alpha_index = count($this->sort_fields);	 
	 }
	 if($this->in_entry) {
		if(in_array($element,$this->fields)) {
		 fwrite($this->out,"$this->cur_data</$element>\n");
		}
	 }
	 else {
		fwrite($this->out,"$this->cur_data</$element>\n");
	 }
	}
	else {
	 if($this->in_entry) {
		if(!in_array($element,$this->fields)) {
		 $this->fields[] = $element;
		}
	 }
	}

if($this->pattern != "") {
	 if(ereg($this->pattern,$this->cur_data)) {
		if($this->produce_stripped_version) {
		 if(in_array($element,$this->fields_to_strip)) {
			$stripped_value = $this->strip($this->cur_data);
			fwrite($this->out,"<".$element."_s>$stripped_value</".$element."_s>\n");
		 }
		}
		else {
		 if(!in_array($element,$this->strippable_fields)) {
			$this->strippable_fields[] = $element;
		 }
		}
	 }
	}
	 
	if($element == $this->head_tag) {
	 $this->in_entry = false;
	}

$this->cur_data = "";

}

/** 
 * Function for handling cDATA during parsing. If the cDATA is not an empty string, append it to then end 
 * of $cur_data.
 */
 function cDataHandler($xp, $data) {
	$data = trim($data);
	if($data != "")
	{
	 $data = ereg_replace("&","&amp;",$data);
	 $data = ereg_replace("<","&lt;",$data);
	 $data = ereg_replace(">","&gt;",$data);
	 $this->cur_data .= $data;
	}
 }

/**
     * A function to remove html encodings of special characters.
     */ 
    function unhtmlentities ($string)  {
	$trans_tbl = get_html_translation_table (HTML_ENTITIES);
	$trans_tbl = array_flip ($trans_tbl);
	$ret = strtr ($string, $trans_tbl);
	return  preg_replace('/\&\#([0-9]+)\;/me',
			     "chr('\\1')",$ret);
    }

/**
     * Function for parsing an XML document and collecting the set of all fields used within entries 
     * and collecting the set of all fields containing characters meeting the $pattern conditions.
     *
     * As noted above, this is the first function, which only searches but produces no new XML document.
     * 
     * @param string $xml The filename of the XML document to be parsed.
     * @param string $head_tag The tag marking the beginning of an entry.
     * @param $symbols string The textarea containing the rules for symbol substitition
     * @param $first_chars The textarea containing the rules for first character substitution.
     * @return void The fields are sorted in the global variables $fields and $strippable fields.
     */
    function get_fields($xml,$head_tag,$symbols,$first_chars) {
	$this->parser = xml_parser_create();
	xml_set_object($this->parser,$this);
	xml_set_element_handler($this->parser,"startHandler","endHandler");
	xml_set_character_data_handler($this->parser,"cDataHandler");

$this->head_tag = $head_tag;

if(trim($symbols) != "") {
	    $this->make_symbols_table($symbols);
	}
	if(trim($first_chars) != "") {
	    $this->make_first_chars_table($first_chars);
	}
	$this->produce_stripped_version = false;

$in = fopen($xml,"r");
	while($line = fgets($in)) {
	    xml_parse($this->parser,$line,false);
	}

fclose($in);
	sort($this->strippable_fields);	
	sort($this->fields);
    }

/**
 * This function produces a modified version of an XML document based on conditions set forth in the required
 * parameters.
 *
 * First, set $produce_stripped_version to true. This tells the parsing functions which function is being
 * utilized. 
 * Second, enter all the user conditions into the relevant global variables. 
 * Third, parse the document and produce the modified XML document.
 *
 * @param string $xml The name of the $xml file to be modified.
 * @param string $new_xml_name The name of the new modified XML file. This is used in case the user wants to save
 * then new XML file.
 * @param string $head_tag The name of tag enclosing an entry.
 * @param string $sort_fields A space seperated list of fields which will be used to create the alpha field.
 * @param string $alpha_symbols The textarea containing the substitution rules for the alpha field.
 * @param string $alpha_firs_chars The textarea containing the substitution rules for the first character of the
 * alpha field.
 * @param string $symbols The textarea containing the substitution rules for the value of fields in $fields.
 * @param string $first_chars The textarea containing the substitution rules for the first character of fields
 * in $fields.
 * @param array $fields An array used to store a list of all fields within entries in the XML document.
 * @param array $fields_to_strip An array of the fields on which the substitution rules will be applied.
 */
 function modify($xml,$new_xml_name, $head_tag, $sort_fields,
		 $alpha_symbols, $alpha_first_chars,
		 $symbols, $first_chars, $fields, 
		 $fields_to_strip) {
	
	$this->produce_stripped_version = true;

$this->parser = xml_parser_create();
	xml_set_object($this->parser,$this);
	xml_set_element_handler($this->parser,"startHandler","endHandler");
	xml_set_character_data_handler($this->parser,"cDataHandler");

$this->fields = $fields;
	$this->fields_to_strip = $fields_to_strip;

$this->head_tag = $head_tag;
	if($sort_fields != "") {
	    $this->get_sort_fields($sort_fields);
	    $this->has_sort_fields = true;
	    $this->fields[] = "alpha";
	}
	else {
	    $this->has_sort_fields = false;
	}
	
	if(trim($symbols) != "") {
	    $this->make_symbols_table($symbols);
	}
	if(trim($first_chars) != "") {
	    $this->make_first_chars_table($first_chars);
	}

if(trim($alpha_symbols) != "") {
	 $this->make_alpha_symbols_table($alpha_symbols);
	}
	if(trim($alpha_first_chars) != "") {
	 $this->make_alpha_first_chars_table($alpha_first_chars);
	}
	$this->out = fopen($new_xml_name,"w");
	fwrite($this->out,"<?xml version=\"1.0\" encoding=\"ISO-8859-1\"?>\n");
	$in = fopen($xml,"r");
	while($line = fgets($in)) {
	 xml_parse($this->parser,$line,false);
	}
	fclose($in);
	fclose($this->out);

xml_parser_free($this->parser);
    }
}

/*
//NOTE: you must remove utf8_encode from the make_symbol_table functions when testing from the command line
set_time_limit(0);
$converter = new db_modifier();
$sort_fields = "lxa lxo";
$symbols = "á = a\né = e\ní = i\nó = o\nú = u\nÁ = A\nÉ = E\nÍ = I\nÓ = O\nÚ = U\nŃ = N\nń = n\nü = u";
$first_chars = "-' = ''";
$alpha_symbols = "á = a";
$alpha_first_chars = "-' = ''";
$fields_not_to_strip = array("compound","disamb","enc_note","fl","fla","flao","flo","grm","grmx","irregv","lex_aff","mod","nae","ncol","nde","nfe","nse","nss","pea","peo","pl_com","plo","qry","qry_fon","qry_h","root_notes","se","sea","seao","sem","seo","spko","src","subadj");

$converter->get_fields("test.xml",$symbols,$first_chars);
foreach($converter->strippable_fields as $field) {
    echo "$field \n";
}

$fields_to_strip = array_diff($converter->strippable_fields,$fields_not_to_strip);
$converter->modify("ActiveNahuatl_2005_august_final.xml","with_stripped.xml","refgroup",
		   $sort_fields,$alpha_symbols,$alpha_first_chars,$symbols,$first_chars,
		   $converter->fields,$fields_to_strip);
*/