Fix image paths in an archive of sites.middlebury.edu/middmag-old/ produced by HTTrack

fix-image-paths.php

#!/usr/bin/env php
<?php

$file = $argv[1];
if (!preg_match('#\.html$#', $file)) {
  throw new Exception("$file must be a .html file.");
}

if (preg_match('#(.+)sites.middlebury.edu/middmag-old/(.+)#', $file, $pathMatches)) {
  $directory = dirname($pathMatches[2]);
  if ($directory == '.') {
    $sitePrefix = '';
  } else {
    $path = explode("/", $directory);
    $depth = count($path);
    $sitePrefix = implode("/", array_fill(0, $depth, '..')).'/';
  }
} else {
  throw new Exception("File must be within sites.middlebury.edu/middmag-old/, $file given.");
}

var_dump($directory, $path, $depth, $sitePrefix);

$html = $origHtml = file_get_contents($file);
$patterns = [];

######################################################
# Replacement patterns and functions.
######################################################

# Images under sites.middlebury.edu/middmag/
$patterns['# (src|href)="((?:\.\./)+)\.\./middmag/(files/(?:[^/"]+/)*([^"/]+))"#'] = 'fix_middmag_files';
function fix_middmag_files($file, $html, $sitePrefix, $m) {
  $srcAttra = $m[0];
  $relativePrefix = $m[2];
  $filePath = $m[3];
  $newAttr = ' '.$m[1].'="'.$relativePrefix.$filePath.'"';
  fwrite(STDOUT, "\n".__FUNCTION__."\n$srcAttra\n$newAttr\n");
  return str_replace($srcAttra, $newAttr, $html);
}

# Images under middleburymagazine.com/
$patterns['# (src|href)="((?:\.\./)+)\.\./\.\./middleburymagazine\.com/(files/(?:[^/"]+/)*([^"/]+))"#'] = 'fix_middleburymagazine_com_files';
function fix_middleburymagazine_com_files($file, $html, $sitePrefix, $m) {
  $srcAttra = $m[0];
  $relativePrefix = $m[2];
  $path = dirname($m[3]);
  $fileName = basename($m[3]);
  // Look for a png, jpg or real path.
  if (preg_match('/\.html$/', $fileName)) {
    $realPath = realpath(dirname($file).'/'.$relativePrefix.$path);
    $fileNameBase = pathinfo($fileName, PATHINFO_FILENAME);
    foreach (scandir($realPath) as $f) {
      // Strip the extension and compare.
      $base = pathinfo($f, PATHINFO_FILENAME);
      if ($base == $fileNameBase) {
        $fileName = $f;
        break;
      }
    }
  }
  $newAttr = ' '.$m[1].'="'.$relativePrefix.$path.'/'.$fileName.'"';
  fwrite(STDOUT, "\n".__FUNCTION__."\n$srcAttra\n$newAttr\n");
  return str_replace($srcAttra, $newAttr, $html);
}

# Links to sites.middlebury.edu/middmag/
$patterns['# (href|src)="https?://(?:sites\.middlebury\.edu/middmag|middmag\.com)/([^"]+)"#'] = 'fix_middmag_links';
function fix_middmag_links($file, $html, $sitePrefix, $m) {
  $srcAttra = $m[0];
  $u = parse_url($m[2]);
  if (empty($u['path'])) {
    $pagePath = 'index.html';
  } else {
    $pagePath = $u['path'];
  }

  // Trailing slash as expected
  if (preg_match('#/$#', $pagePath)) {
    // Ending with a trailing slash as expected for directories.
    $pagePath = $pagePath.'index.html';
  }
  // Ending with a filename.
  elseif (preg_match('#^.+/([^/]+\.\w+)$#', $pagePath)) {
    // Nothing to do.
  }
  // Odd special case seen.
  elseif($pagePath == '2014/02/26/uncle-donnie-t…s-on-the-world') {
    $pagePath = '2014/02/26/uncle-donnie-takes-on-the-world/index.html';
  }
  // ending with no file extension or trailing slash.
  else {
    $pagePath = $pagePath.'/index.html';
  }
  # Add query string.
  if (!empty($u['query'])) {
    $pagePath .= '?'.$u['query'];
  }
  # Add fragments
  if (!empty($u['fragment'])) {
    $pagePath .= '#'.$u['fragment'];
  }

  $newAttr = ' '.$m[1].'="'.$sitePrefix.$pagePath.'"';
  fwrite(STDOUT, "\n".__FUNCTION__."\n$srcAttra\n$newAttr\n");
  return str_replace($srcAttra, $newAttr, $html);
}

# Remove onclick from branding.
$patterns['# onclick="location.href=\'http://sites.middlebury.edu/middmag-old\';"#'] = 'remove_branding_onclick';
function remove_branding_onclick($file, $html, $sitePrefix, $m) {
  $srcAttra = $m[0];
  $newAttr = 'onclick="location.href=\''.$sitePrefix.'index.html\';"';
  fwrite(STDOUT, "\n".__FUNCTION__."\n$srcAttra\n$newAttr\n");
  return str_replace($srcAttra, $newAttr, $html);
}

# Fix srcset urls.
$patterns['# srcset="([^"]+)"#'] = 'fix_srcset_urls';
function fix_srcset_urls($file, $html, $sitePrefix, $m) {
  $srcAttra = $m[0];
  $contents = $m[1];
  $contents = str_replace('https://sites.middlebury.edu/middmag-old/', $sitePrefix, $contents);
  $newAttr = 'srcset="'. $contents .'"';
  fwrite(STDOUT, "\n".__FUNCTION__."\n$srcAttra\n$newAttr\n");
  return str_replace($srcAttra, $newAttr, $html);
}



######################################################
# Loop through our patterns and make changes.
######################################################
foreach ($patterns as $regex => $callback) {
  if (preg_match_all($regex, $html, $matches, PREG_SET_ORDER)) {
    foreach ($matches as $match) {
      $html = call_user_func($callback, $file, $html, $sitePrefix, $match);
    }
  }
}

if ($html != $origHtml) {
  file_put_contents($file, $html);
}