Jump to content

Wikipedia:WikiProject Dates/Parse script

From Wikipedia, the free encyclopedia

Usage

[edit]
cat enwiki-20080724-pages-articles.xml | php parse.php > data

Script: parse.php

[edit]
<?php
while($line=fgets(STDIN)) {
       if(preg_match('/^\s*<title>(.*)<\/title>\s*$/', $line, $matches)) {
               $title = strtr($matches[1], ' ', '_');
       } elseif($buffer) {
               if(preg_match('/(.*)<\/text>\s*$/', $line, $matches)) {
                       process_page($title, $buffer . ' ' . $matches[1]);
                       $buffer = null;
               } else {
                       $buffer .= ' ' . rtrim($line);
               }
       } elseif(preg_match('/^\s*<text[^>]*>(.*)<\/text>\s*$/', $line, $matches)) {
               process_page($title, rtrim($matches[1]));
       } elseif(preg_match('/^\s*<text[^>]*>(.*)/', $line, $matches)) {
               $buffer = ' ' . rtrim($matches[1]);
       }
}
function process_page($title, $body) {
       print $title;
       while(preg_match('/(.*)\{\{([^\{\}]*)\}\}(.*)/', $body, $matches)) {
               $body = $matches[1] . ' ' . $matches[3];
               $dates = tally_dates($matches[2], $dates);
       }
       pretty_print($dates, 'template');
       $dates = null;
       while(preg_match('/(.*)\<ref[^\&]*\>(.*?)\<\/ref[^\&]*\>(.*)/i', $body, $matches)) {
               $body = $matches[1] . ' ' . $matches[3];
               $dates = tally_dates($matches[2], $dates);
       }
       pretty_print($dates, 'references');
       $dates = null;
       $dates = tally_dates($body, $dates);
       pretty_print($dates);
       print "\n";
}
function tally_dates($string, $dates) {
       $month_regex = '(january|february|march|april|may|june|july|august|september|october|november|december)';
       $regexTrail = '(.*)/iu';
       $prxDM = "\[\[(\d{1,2})[ _]{$month_regex}]]";
       $prxMD = "\[\[{$month_regex}[ _](\d{1,2})]]";
       $prxY = "\[\[(\d{1,4}([ _]BC|))]]";
       $prxISO1 = "\[\[(-?\d{4})]]-\[\[(\d{2})-(\d{2})]]";
       $prxISO2 = "\[\[(-?\d{4})-(\d{2})-(\d{2})]]";
       $DMY_linked = "/(.*){$prxDM} *,? *{$prxY}{$regexTrail}";
       $DMY_raw = "/(.*)(\d{1,2})[ _]{$month_regex} *,? *(\d{1,4}([ _]BC|)){$regexTrail}";
       $YDM_linked = "/(.*){$prxY} *,? *{$prxDM}{$regexTrail}";
       $YDM_raw = "/(.*)(\d{1,4}([ _]BC|)) *,? +(\d{1,2})[ _]{$month_regex}{$regexTrail}";
       $MDY_linked = "/(.*){$prxMD} *,? *{$prxY}{$regexTrail}";
       $MDY_raw = "/(.*){$month_regex} +(\d{1,2}) *,? +(\d{1,4}([ _]BC|)){$regexTrail}";
       $YMD_linked = "/(.*){$prxY} *,? *{$prxMD}{$regexTrail}";
       $YMD_raw = "/(.*)(\d{1,4}([ _]BC|)) *,? +{$month_regex} +(\d{1,2}){$regexTrail}";
       $DM_linked = "/(.*){$prxDM}{$regexTrail}";
       $MD_linked = "/(.*){$prxMD}{$regexTrail}";
       $ISO1_linked = "/(.*){$prxISO1}{$regexTrail}";
       $ISO2_linked = "/(.*){$prxISO2}{$regexTrail}";
       $ISO_raw = "/(.*)(-?\d{4})-(\d{2})-(\d{2}){$regexTrail}";
       while(preg_match($DMY_linked, $string, $matches)) {
               $dates['DMY_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($MDY_linked, $string, $matches)) {
               $dates['MDY_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($YDM_linked, $string, $matches)) {
               $dates['YDM_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($YMD_linked, $string, $matches)) {
               $dates['YMD_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($MD_linked, $string, $matches)) {
               $dates['MD_linked']++;
               $string = $matches[1] . ' ' . $matches[4];
       }
       while(preg_match($DM_linked, $string, $matches)) {
               $dates['DM_linked']++;
               $string = $matches[1] . ' ' . $matches[4];
       }
       while(preg_match($DMY_raw, $string, $matches)) {
               $dates['DMY_raw']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($MDY_raw, $string, $matches)) {
               $dates['MDY_raw']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($YDM_raw, $string, $matches)) {
               $dates['YDM_raw']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($ISO1_linked, $string, $matches)) {
               $dates['ISO1_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($ISO2_linked, $string, $matches)) {
               $dates['ISO2_linked']++;
               $string = $matches[1] . ' ' . $matches[6];
       }
       while(preg_match($ISO_raw, $string, $matches)) {
               $dates['ISO_raw']++;
               $string = $matches[1] . ' ' . $matches[5];
       }
       return $dates;
}
function pretty_print($dates, $type = null) {
       if(!is_array($dates)) {
               return;
       }
       if($type == 'template') {
               print ' {';
       } elseif($type == 'references') {
               print ' <';
       } else {
               print ' ';
       }
       foreach($dates as $format => $count) {
               print $maybe_comma . $format . ':' . $count;
               $maybe_comma = ',';
       }
       if($type == 'template') {
               print '}';
       } elseif($type == 'references') {
               print '>';
       }
}
?>