nextcloud-memento/lib/Controller

<?php
namespace OCA\Memento\Controller;

use \DOMDocument;
use \DateTime;
use OCP\Share;
use OCP\Files\FileInfo;

// Finds HTML files that claim to be a snapshot of the given URL;
// Returns an array of mementos, sorted by datetime, with each memento represented by an array:
// [
//   'mementoUrl' => URL of the file, relative to the nextcloud instance
//   'originalUrls' => original URLs, usually just one.
//   'datetime' => snapshot datetime as a unix timestamp
// ]
//
// Each mementoUrl is hardcoded to /apps/raw/..., thus relying on the 'raw' app to serve the files.

trait MementoFinder {
	function findSingleUserMementosForUrl($userId, $url) {
		// Get the user's public mementos.
		$foundMementos = findPublicMementos($this->serverContainer->getShareManager(), $userId);

		// If logged in, and asking for one's own mementos, get private mementos too.
		if ($this->loggedInUserId === $userId) {
			$userFolder = $this->serverContainer->getUserFolder($this->loggedInUserId);
			$moreMementos = findPrivateMementos($userFolder, $this->loggedInUserId);
			$foundMementos = array_merge($foundMementos, $moreMementos);
		}

		return listMementosMatchingUrl($foundMementos, $url);
	}

	function findAllUsersMementosForUrl($url) {
		$foundMementos = [];

		// Get the public mementos of every user.
		$allUserIds = [];
		$this->serverContainer->getUserManager()->callForAllUsers(
			function ($user) use (&$allUserIds) { $allUserIds[] = $user->getUID(); }
		);
		$shareManager = $this->serverContainer->getShareManager();
		foreach ($allUserIds as $userId) {
			$moreMementos = findPublicMementos($shareManager, $userId);
			$foundMementos = array_merge($foundMementos, $moreMementos);
		}

		// If logged in, get current user's private mementos too.
		if ($this->loggedInUserId) {
			$userFolder = $this->serverContainer->getUserFolder($this->loggedInUserId);
			$moreMementos = findPrivateMementos($userFolder, $this->loggedInUserId);
			$foundMementos = array_merge($foundMementos, $moreMementos);
		}

		return listMementosMatchingUrl($foundMementos, $url);
	}
}

function findPrivateMementos($userFolder, $userId) {
	$urlForFile = function ($file) use ($userFolder, $userId) {
		$absoluteFilePath = $file->getPath();
		$relativeFilePath = $userFolder->getRelativePath($absoluteFilePath);
		$rawFileUrl = joinPaths("/apps/raw/u/$userId", $relativeFilePath); // XXX hardcoded dependency
		return $rawFileUrl;
	};

	// Peek into each HTML file the user owns, and return those that are mementos.
	$files = $userFolder->searchByMime('text/html');
	$foundMementos = [];
	foreach ($files as $file) {
		$mementoInfo = extractMementoInfo($file);
		if ($mementoInfo) {
			$mementoInfo['mementoUrl'] = $urlForFile($file);
			$foundMementos[] = $mementoInfo;
		}
	}
	return $foundMementos;
}

function findPublicMementos($shareManager, $userId) {
	$shares = $shareManager->getSharesBy(
		$userId,
		Share::SHARE_TYPE_LINK,
		null, /* path */
		true, /* include reshares */
		-1 /* no limit */
	);

	$urlForSharedFile = function ($share) {
		return "/apps/raw/s/" . $share->getToken(); // XXX hardcoded dependency
	};
	$urlForFileInsideSharedFolder = function ($share, $folder, $file) {
		$absoluteFilePath = $file->getPath();
		$relativeFilePath = $folder->getRelativePath($absoluteFilePath);
		return joinPaths("/apps/raw/s/{$share->getToken()}", $relativeFilePath);
	};

	// Look into every shared file to see if it is a memento.
	$foundMementos = [];
	foreach ($shares as $share) {
		$node = $share->getNode();
		if ($node->getType() === FileInfo::TYPE_FILE) {
			$mementoInfo = extractMementoInfo($node);
			if ($mementoInfo) {
				$mementoInfo['mementoUrl'] = $urlForSharedFile($share);
				$foundMementos[] = $mementoInfo;
			}
		} else {
			// Share is a folder: Go through all html files inside the shared folder.
			$folder = $node;
			$files = $folder->searchByMime('text/html');
			foreach ($files as $file) {
				$mementoInfo = extractMementoInfo($file);
				if ($mementoInfo) {
					$mementoInfo['mementoUrl'] = $urlForFileInsideSharedFolder($share, $folder, $file);
					$foundMementos[] = $mementoInfo;
				}
			}
		}
	}
	return $foundMementos;
}

function listMementosMatchingUrl($foundMementos, $url) {
	// Filter those that match the requested URL
	$matchingMementos = filterMementosByUrl($foundMementos, $url);
	// Deduplicate (as a file may be accessible both through a public and a private URL)
	$matchingMementos = deduplicateMementos($matchingMementos);
	// Sort them by date.
	$matchingMementos = sortMementos($matchingMementos);
	return $matchingMementos;
}

function filterMementosByUrl($mementos, $url) {
	$matchingMementos = array_filter($mementos, function ($mementoInfo) use ($url) {
		return matchesUrl($mementoInfo, $url);
	});
	return $matchingMementos;
}

function matchesUrl($mementoInfo, $url) {
	$originalUrls = $mementoInfo['originalUrls'];
	foreach ($originalUrls as $originalUrl) {
		if (normaliseUrl($originalUrl) === normaliseUrl($url)) {
			return true;
		}
	}
	return false;
}

function normaliseUrl($url) {
	// Ignore trailing slashes. Because everybody does.
	$url = rtrim($url, '/');
	return $url;
}

function deduplicateMementos($mementos) {
	$deduped = [];
	$seenIds = [];
	foreach ($mementos as $memento) {
		if (!array_key_exists($memento['id'], $seenIds)) {
			$deduped[] = $memento;
			$seenIds[$memento['id']] = null;
		}
	}
	return $deduped;
}

// Sort an array of mementos by their datetime. Oldest first.
function sortMementos($mementos) {
	usort($mementos, function ($m1, $m2) { return $m1['datetime'] <=> $m2['datetime']; });
	return $mementos;
}

function joinPaths($piece1, $piece2) {
	$left = rtrim($piece1, '/');
	$right = ltrim($piece2, '/');
	return "$left/$right";
}

function extractMementoInfo($file) {
	$content = $file->getContent();
	$DOM = new DOMDocument;
	$DOM->loadHTML($content);
	$headElement = $DOM->documentElement->getElementsByTagName('head')[0];
	if (!$headElement) return null; // possibly $content was not HTML at all.
	$originalUrls = getOriginalUrls($headElement);
	$datetime = getDatetime($headElement);
	return [
		'originalUrls' => $originalUrls,
		'datetime' => $datetime,
		'id' => $file->getFileInfo()->getId() // for deduplication
	];
}

// Reads hrefs from any <link> with relation type "original".
// (note the plural: we also accept pages that claim to correspond to multiple original URLs)
function getOriginalUrls($headElement) {
	$originalUrls = [];
	$links = $headElement->getElementsByTagName('link');
	foreach ($links as $link) {
		$rels = explode(' ', $link->getAttribute('rel'));
		if (in_array('original', $rels)) {
			$href = $link->getAttribute('href');
			$href = filter_var($href, FILTER_VALIDATE_URL, FILTER_FLAG_SCHEME_REQUIRED);
			if ($href) {
				$originalUrls[] = $href;
			}
		}
	}
	return $originalUrls;
}

// Read the content of the first <meta http-equiv="Memento-Datetime">, if any.
function getDatetime($headElement) {
	$metas = $headElement->getElementsByTagName('meta');
	foreach($metas as $meta) {
		// Let's match case-insensitively, I guess?
		if (strtolower($meta->getAttribute('http-equiv')) === 'memento-datetime') {
			$datetime = $meta->getAttribute('content');
			$datetime = DateTime::createFromFormat(DateTime::RFC1123, $datetime)->getTimestamp();
			return $datetime; // Return directly at the first match
		}
	}
	return null;
}