imgur-rescue-project/collect/collect.php

#!/usr/bin/php
<?php

error_reporting(E_ALL);
define('IMGUR_CLIENT_ID', '546c25a59c58ad7'); // extract from any web interface pageview

function is_png($data) {
	return (substr($data, 0, 4) === "\x89PNG");
}

function is_jpg($data) {
	if (substr($data, 0, 3) === "\xFF\xD8\xFF") {
		return true;
	}
	
	return str_contains(substr($data, 0, 64), "JFIF");
}

function is_gif($data) {
	return (substr($data, 0, 4) === "GIF8");
}

function is_mp4($data) {
	return str_contains(substr($data, 0, 64), "isom");
}

/**
 * @param $type 'media' or 'albums'
 */ 
function imgur_mediainfo(string $code, string $type) {

	// n.b. 'posts' is often a synonym for 'media' with the same result
	// There is also a /meta endpoint possible after {$code}

	$ret = file_get_contents("https://api.imgur.com/post/v1/{$type}/{$code}?client_id=" . IMGUR_CLIENT_ID . "&include=media");
	if ($ret === false) {
		throw new Exception("Failed API lookup for {$code} as {$type}");
	}
	
	if ($ret[0] !== '{') {
		throw new Exception("API result for {$code} as {$type} got unexpected body: {$ret}");
	}
	
	$obj = json_decode($ret, true);
	if (isset($obj['errors'])) {
		throw new Exception("API result for {$code} as {$type} got unexpected body: {$ret}");
	}
	
	if ( ($type === 'albums') !== $obj['is_album'] ) {
		throw new Exception("Unexpected type mismatch from API");
	}

	return $obj;
}

function imgur_download_single(string $code, string $type) {
	echo "Downloading {$code}...\n";
	
	if (file_exists("errors.{$type}/".$code)) {
		echo "skipping (known error)\n";
		return false;
	}
	
	if (! file_exists("metadata.{$type}/".$code)) {
		try {
			$metadata = imgur_mediainfo($code, $type);
			
		} catch (Exception $ex) {
			echo "WARNING: metadata download failed\n";
			echo (string)$ex . "\n";
			
			file_put_contents("errors.{$type}/".$code, json_encode([
				"message" => $ex->getMessage(),
				"failure_time" => date(DATE_RSS),
			]));
			
			return false;
		}
		
		file_put_contents("metadata.{$type}/".$code, json_encode($metadata));
		
	} else {
		$metadata = json_decode(file_get_contents("metadata.{$type}/".$code), true);
	}
	
	echo "code {$code} (type={$type}) contains ".$metadata['image_count']." media entries\n";
	if ($metadata['image_count'] == 0) {
		echo "WARNING: weird album with no images!\n";
		return true;
	}
	
	foreach($metadata['media'] as $single) {
		
		echo "- entry ".$single['id']."\n";
		
		if (file_exists("images/".$single['id'])) {
			echo "already exists (OK)\n";
			continue;
		}
		
		// Download whole URL
		$ret = file_get_contents($single['url']);
		if ($ret === false) {
			echo "download failed (WARNING)\n";
			continue;
		}
		
		if (! is_png($ret) && ! is_jpg($ret) && ! is_gif($ret) && ! is_mp4($ret)) {
			echo "unexpected result, not jpg/gif/png/mp4 (WARNING)\n";
			file_put_contents("images/".$single['id'].".INVALID-NOT-AN-IMAGE", $ret);	
			continue;
		}
		
		if (strlen($ret) === 503 && md5($ret) == "d835884373f4d6c8f24742ceabe74946") {
			echo "fake image result for image not found (WARNING)\n";
			continue;
		}
		
		file_put_contents("images/".$single['id'], $ret);
	}
	
	// all done
	return true;
}

function main() {
	
	foreach(['images', 'metadata.albums', 'metadata.media', 'errors.albums', 'errors.media'] as $dirname) {
		if (! is_dir($dirname)) {
			mkdir($dirname);
		}
	}
	
	$urls = explode("\n", file_get_contents("all-urls.txt"));
	$matches = [];

	foreach($urls as $url) {
		
		if (preg_match("~^https://(i\\.|m\\.|img\\.|www\.)?imgur\\.com/([0-9a-zA-Z,]+)\\.?(jpg|jpeg|gif|webm|png|gifv|mp4)?(#.+)?$~", $url, $matches)) {	
			
			foreach(explode(',', $matches[2]) as $single_id) {
				imgur_download_single($single_id, 'media');
			}
		
		} else if (preg_match("~^https://(i\\.|m\\.|img\\.|www\.)?imgur\\.com/(a/|gallery/|(?:r|t|topic)/[a-zA-Z0-9_]+/)?([0-9a-zA-Z,]+)\\.?(jpg|jpeg|gif|webm|png|gifv|mp4)?(#.+)?$~", $url, $matches)) {
			
			imgur_download_single($matches[3], 'albums');
		
		} else {
			echo "WARNING: Unsupported URL: {$url}\n";
			
		}
		
	}

}

main();