parse_habr_career.php

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

error_reporting(E_ALL);
set_time_limit(0);

function get_sitemap_urls($url) {
  preg_match_all("#<loc>(.+?)</loc>#", gzdecode(file_get_contents($url)), $matches);
  return $matches[1];
}

$sitemap_urls = get_sitemap_urls("https://career.habr.com/assets/sitemap.xml.gz");

foreach ($sitemap_urls as $sitemap_url) {
  foreach (get_sitemap_urls($sitemap_url) as $company_url) {
    if (preg_match('#/companies/[^/]+$#', $company_url)) {
      fwrite(STDERR, "Check $company_url\n");
      $contents = @file_get_contents($company_url);
      if (preg_match('#<div class="company_site"><a rel="nofollow" href="([^"]+)#', $contents, $matches)) {
        fwrite(STDERR, "Found " . $matches[1] . "\n");
        echo $matches[1] . "\n";
      }
    }
  }
}