PHP parser for Ville de Montréal rink conditions PDFs

rinks.php

<?php
// Original by: Philippe Dagenais-Pérusse

// Get the files to parse
if ($argc) { // Command-line interface
  $files = array_slice($argv, 1);
}
else {
  $files = array();
  if ($handle = opendir($path)) {
    while (($file = readdir($handle)) !== FALSE) {
      if ($file != '.' && $file != '..' && strstr($file, '.pdf')) {
        $files[$file] = $_SERVER['DOCUMENT_ROOT'] . '/glaces/lib/files/' . $file;
      }
    }
    closedir($handle);
  }
}

$headers = array(
  1 => 'OO', // Ouverte: oui
  2 => 'ON', // Ouverte: non
  3 => 'DO', // Déblayée: oui
  4 => 'DN', // Déblayée: non
  5 => 'AO', // Arrosée: oui
  6 => 'AN', // Arrosée: non
  7 => 'CE', // Condition: excellente
  8 => 'CB', // Condition: bonne
  9 => 'CM', // Condition: mauvaise
);

// The first text line on which a rink name appears
$first = 13;

// If the status line is neither empty nor "X", parsing has failed
function parsable($value) {
  return $value == '' || $value == 'X';
}

$boroughs = array();
foreach ($files as $basename => $file) {
  // Collect all text lines from the PDF
  $lines = array();
  foreach (file($file) as $line_no => $line) {
    if (preg_match('/Tj/', $line)) {
      $lines[] = trim(substr(str_replace(') Tj', '', stripslashes($line)), 1));
    }
  }

  // Get the borough data
  $borough = array(
    'Fichier'     => $basename,        // File
    'MAJ-Fichier' => filemtime($file), // Mise à jour du fichier / File updated at
    'Region'      => $lines[8],        // Arrondissement / Borough
    'MAJ'         => $lines[0],        // Mise à jour / Updated at
    'Remarques'   => '',               // Remarks
    'Patinoires'  => array(),          // Rinks
    'INTEGRITE'   => 'Oui',            // Parseable
  );

  // Get the borough remarks
  $index = array_search($file == 'L29_79.pdf' ? 'Remarques' : 'Oui', $lines);
  for ($i = 1; $i <= 5; $i++) {
    if ($lines[$index - $i] == '' || $lines[$index - $i] == 'X' || is_numeric($lines[$index - $i])) break;
    $borough['Remarques'] = $lines[$index - $i] . ' ' . $borough['Remarques'];
  }

  // Nombre total de patinoires / Total number of rinks
  $total = $lines[array_search('Entretenues par les citoyens', $lines) - 1];

  // Get the rinks data
  for ($i = 0; $i < $total; $i++) {
    $index = array_search('Grand total', $lines);
    $rink = array(
      'Nom' => $lines[$first + $i],
      'RO'  => $lines[$index + $i + 1],          // Resurfacée: oui
      'RN'  => $lines[$index + $i + 1 + $total], // Resurfacée: non
    );
    if (!parsable($lines[$index + $i + 1]) || !parsable($lines[$index + $i + 1 + $total])) {
      $borough['INTEGRITE'] = 'Non';
    }

    foreach ($headers as $offset => $key) {
      $index = $first + $i + ($total * $offset);
      $rink[$key] = $lines[$index];
      if (!parsable($lines[$index])) {
        $borough['INTEGRITE'] = 'Non';
      }
    }

    $borough['Patinoires'][] = $rink;
  }

  $boroughs[] = $borough;
}

function cmp($a, $b) {
  return strcmp($a['Region'], $b['Region']);
}
usort($boroughs, 'cmp'); // Order by borough