リモートホストごとの同時アクセス制御とメタデータ取得ポリシー制御

app/MetadataResolver/DisallowedByProviderException.php

@@ -0,0 +1,30 @@
+<?php
+
+namespace App\MetadataResolver;
+
+use RuntimeException;
+use Throwable;
+
+/**
+ * ContentProviderの提供するrobots.txtによってクロールが拒否された場合にスローされます。
+ */
+class DisallowedByProviderException extends RuntimeException
+{
+    private $url;
+
+    public function __construct(string $url, Throwable $previous = null)
+    {
+        parent::__construct("Access denied by robots.txt: $url", 0, $previous);
+        $this->url = $url;
+    }
+
+    public function getUrl(): string
+    {
+        return $this->url;
+    }
+
+    public function getHost(): string
+    {
+        return parse_url($this->url, PHP_URL_HOST);
+    }
+}

app/Services/MetadataResolveService.php

@@ -2,14 +2,20 @@
 
 namespace App\Services;
 
+use App\ContentProvider;
 use App\Metadata;
 use App\MetadataResolver\DeniedHostException;
+use App\MetadataResolver\DisallowedByProviderException;
 use App\MetadataResolver\MetadataResolver;
 use App\MetadataResolver\ResolverCircuitBreakException;
 use App\MetadataResolver\UncaughtResolverException;
 use App\Tag;
 use App\Utilities\Formatter;
+use Carbon\Carbon;
+use Carbon\CarbonInterface;
+use GuzzleHttp\Client;
 use Illuminate\Support\Facades\DB;
+use Illuminate\Support\Facades\Log;
 
 class MetadataResolveService
 {
@@ -44,13 +50,208 @@ class MetadataResolveService
             throw new DeniedHostException($url);
         }
 
-        DB::beginTransaction();
-        try {
         $metadata = Metadata::find($url);
 
         // 無かったら取得
         // TODO: ある程度古かったら再取得とかありだと思う
         if ($metadata == null || $metadata->needRefresh()) {
+            $hostWithPort = $this->getHostWithPortFromUrl($url);
+            $metadata = $this->hostLock($hostWithPort, function (?CarbonInterface $lastAccess) use ($url) {
+                // HostLockの解放待ちをしている間に、他のプロセスで取得完了しているかもしれない
+                $metadata = Metadata::find($url);
+                if ($metadata !== null && !$metadata->needRefresh()) {
+                    return $metadata;
+                }
+
+                $this->checkProviderPolicy($url, $lastAccess);
+
+                return $this->resolve($url, $metadata);
+            });
+        }
+
+        return $metadata;
+    }
+
+    /**
+     * URLからホスト部とポート部を抽出
+     * @param string $url
+     * @return string
+     */
+    private function getHostWithPortFromUrl(string $url): string
+    {
+        $parts = parse_url($url);
+        $host = $parts['host'];
+        if (isset($parts['port'])) {
+            $host .= ':' . $parts['port'];
+        }
+
+        return $host;
+    }
+
+    /**
+     * アクセス先ホスト単位の排他ロックを取って処理を実行
+     * @param string $host
+     * @param callable $fn
+     * @return mixed return of $fn
+     * @throws \RuntimeException いろいろな死に方をする
+     */
+    private function hostLock(string $host, callable $fn)
+    {
+        $lockDir = storage_path('content_providers_lock');
+        if (!file_exists($lockDir)) {
+            if (!mkdir($lockDir)) {
+                throw new \RuntimeException("Lock failed! Can't create lock directory.");
+            }
+        }
+
+        $lockFile = $lockDir . DIRECTORY_SEPARATOR . $host;
+        $fp = fopen($lockFile, 'c+b');
+        if ($fp === false) {
+            throw new \RuntimeException("Lock failed! Can't open lock file.");
+        }
+
+        try {
+            if (!flock($fp, LOCK_EX)) {
+                throw new \RuntimeException("Lock failed! Can't lock file.");
+            }
+
+            try {
+                $accessInfoText = stream_get_contents($fp);
+                if ($accessInfoText !== false) {
+                    $accessInfo = json_decode($accessInfoText, true);
+                }
+
+                $result = $fn(isset($accessInfo['time']) ? new Carbon($accessInfo['time']) : null);
+
+                $accessInfo = [
+                    'time' => now()->toIso8601String()
+                ];
+                fseek($fp, 0);
+                if (fwrite($fp, json_encode($accessInfo)) === false) {
+                    throw new \RuntimeException("I/O Error! Can't write to lock file.");
+                }
+
+                return $result;
+            } finally {
+                if (!flock($fp, LOCK_UN)) {
+                    throw new \RuntimeException("Unlock failed! Can't unlock file.");
+                }
+            }
+        } finally {
+            if (!fclose($fp)) {
+                throw new \RuntimeException("Unlock failed! Can't close lock file.");
+            }
+        }
+    }
+
+    /**
+     * 指定したメタデータURLのホストが持つrobots.txtをダウンロードします。
+     * @param string $url メタデータのURL
+     * @return string
+     */
+    private function fetchRobotsTxt(string $url): ?string
+    {
+        $parts = parse_url($url);
+        $robotsUrl = http_build_url([
+            'scheme' => $parts['scheme'],
+            'host' => $parts['host'],
+            'port' => $parts['port'] ?? null,
+            'path' => '/robots.txt'
+        ]);
+
+        $client = app(Client::class);
+        try {
+            $res = $client->get($robotsUrl);
+
+            return (string) $res->getBody();
+        } catch (\Exception $e) {
+            Log::error("robots.txtの取得に失敗: {$e}");
+
+            return null;
+        }
+    }
+
+    /**
+     * ContentProviderポリシー情報との照合を行い、アクセス可能かチェックします。アクセスできない場合は例外をスローします。
+     * @param string $url メタデータを取得したいURL
+     * @param CarbonInterface|null $lastAccess アクセス先ホストへの最終アクセス日時 (記録がある場合)
+     * @throws DeniedHostException アクセス先がTissue内のブラックリストに入っている場合にスロー
+     * @throws DisallowedByProviderException アクセス先のrobots.txtによって拒否されている場合にスロー
+     */
+    private function checkProviderPolicy(string $url, ?CarbonInterface $lastAccess): void
+    {
+        DB::beginTransaction();
+        try {
+            $hostWithPort = $this->getHostWithPortFromUrl($url);
+            $contentProvider = ContentProvider::sharedLock()->find($hostWithPort);
+            if ($contentProvider === null) {
+                $contentProvider = ContentProvider::create([
+                    'host' => $hostWithPort,
+                    'robots' => $this->fetchRobotsTxt($url),
+                    'robots_cached_at' => now(),
+                ]);
+            }
+
+            if ($contentProvider->is_blocked) {
+                throw new DeniedHostException($url);
+            }
+
+            // 連続アクセス制限
+            if ($lastAccess !== null) {
+                $elapsedSeconds = $lastAccess->diffInSeconds(now(), false);
+                if ($elapsedSeconds < $contentProvider->access_interval_sec) {
+                    if ($elapsedSeconds < 0) {
+                        $wait = abs($elapsedSeconds) + $contentProvider->access_interval_sec;
+                    } else {
+                        $wait = $contentProvider->access_interval_sec - $elapsedSeconds;
+                    }
+                    sleep($wait);
+                }
+            }
+
+            // Fetch robots.txt
+            if ($contentProvider->robots_cached_at->diffInDays(now()) >= 7) {
+                $contentProvider->update([
+                    'robots' => $this->fetchRobotsTxt($url),
+                    'robots_cached_at' => now(),
+                ]);
+            }
+
+            // Check robots.txt
+            $robotsParser = new \RobotsTxtParser($contentProvider->robots);
+            $robotsParser->setUserAgent('TissueBot');
+            $robotsDelay = $robotsParser->getDelay();
+            if ($robotsDelay !== 0 && $robotsDelay >= $contentProvider->access_interval_sec) {
+                $contentProvider->access_interval_sec = (int) $robotsDelay;
+                $contentProvider->save();
+            }
+            if ($robotsParser->isDisallowed(parse_url($url, PHP_URL_PATH))) {
+                throw new DisallowedByProviderException($url);
+            }
+
+            DB::commit();
+        } catch (DeniedHostException | DisallowedByProviderException $e) {
+            // ContentProviderのデータ更新は行うため
+            DB::commit();
+            throw $e;
+        } catch (\Exception $e) {
+            DB::rollBack();
+            throw $e;
+        }
+    }
+
+    /**
+     * メタデータをリモートサーバに問い合わせて取得します。
+     * @param string $url メタデータを取得したいURL
+     * @param Metadata|null $metadata キャッシュ済のメタデータ (存在する場合)
+     * @return Metadata 取得できたメタデータ
+     * @throws UncaughtResolverException Resolver内で例外が発生して取得できなかった場合にスロー
+     * @throws ResolverCircuitBreakException 規定回数以上の解決失敗により、メタデータの取得が不能となっている場合にスロー
+     */
+    private function resolve(string $url, ?Metadata $metadata): Metadata
+    {
+        DB::beginTransaction();
+        try {
             if ($metadata === null) {
                 $metadata = new Metadata(['url' => $url]);
             }
@@ -84,7 +285,6 @@ class MetadataResolveService
                 $tagIds[] = $tag->id;
             }
             $metadata->tags()->sync($tagIds);
-            }
 
             DB::commit();
 

composer.json

@@ -33,7 +33,8 @@
         "sentry/sentry-laravel": "1.8.0",
         "staudenmeir/eloquent-eager-limit": "^1.0",
         "symfony/css-selector": "^4.3",
-        "symfony/dom-crawler": "^4.3"
+        "symfony/dom-crawler": "^4.3",
+        "t1gor/robots-txt-parser": "^0.2.4"
     },
     "require-dev": {
         "barryvdh/laravel-debugbar": "^3.1",

composer.lock

@@ -4,7 +4,7 @@
         "Read more about it at https://getcomposer.org/doc/01-basic-usage.md#installing-dependencies",
         "This file is @generated automatically"
     ],
-    "content-hash": "1bba68b609be6a0dcdaf05d72e8eb759",
+    "content-hash": "bbb184ff943ae3a938a8370d94b6afb2",
     "packages": [
         {
             "name": "anhskohbo/no-captcha",
@@ -547,20 +547,6 @@
                 "uppercase",
                 "words"
             ],
-            "funding": [
-                {
-                    "url": "https://www.doctrine-project.org/sponsorship.html",
-                    "type": "custom"
-                },
-                {
-                    "url": "https://www.patreon.com/phpdoctrine",
-                    "type": "patreon"
-                },
-                {
-                    "url": "https://tidelift.com/funding/github/packagist/doctrine%2Finflector",
-                    "type": "tidelift"
-                }
-            ],
             "time": "2020-05-29T15:13:26+00:00"
         },
         {
@@ -623,20 +609,6 @@
                 "parser",
                 "php"
             ],
-            "funding": [
-                {
-                    "url": "https://www.doctrine-project.org/sponsorship.html",
-                    "type": "custom"
-                },
-                {
-                    "url": "https://www.patreon.com/phpdoctrine",
-                    "type": "patreon"
-                },
-                {
-                    "url": "https://tidelift.com/funding/github/packagist/doctrine%2Flexer",
-                    "type": "tidelift"
-                }
-            ],
             "time": "2020-05-25T17:44:05+00:00"
         },
         {
@@ -5649,6 +5621,63 @@
             ],
             "time": "2020-05-30T20:06:45+00:00"
         },
+        {
+            "name": "t1gor/robots-txt-parser",
+            "version": "v0.2.4",
+            "source": {
+                "type": "git",
+                "url": "https://github.com/t1gor/Robots.txt-Parser-Class.git",
+                "reference": "7ff08da5625fb4f72d17b1528c60aadb184e9e68"
+            },
+            "dist": {
+                "type": "zip",
+                "url": "https://api.github.com/repos/t1gor/Robots.txt-Parser-Class/zipball/7ff08da5625fb4f72d17b1528c60aadb184e9e68",
+                "reference": "7ff08da5625fb4f72d17b1528c60aadb184e9e68",
+                "shasum": ""
+            },
+            "require": {
+                "ext-mbstring": "*",
+                "php": ">=5.5.0",
+                "vipnytt/useragentparser": "^1.0"
+            },
+            "require-dev": {
+                "codeclimate/php-test-reporter": ">=0.2",
+                "phpunit/phpunit": "~3.7"
+            },
+            "type": "library",
+            "autoload": {
+                "classmap": [
+                    "source/robotstxtparser.php"
+                ]
+            },
+            "notification-url": "https://packagist.org/downloads/",
+            "license": [
+                "MIT"
+            ],
+            "authors": [
+                {
+                    "name": "Igor Timoshenkov",
+                    "email": "igor.timoshenkov@gmail.com",
+                    "role": "creator"
+                },
+                {
+                    "name": "Jan-Petter Gundersen",
+                    "email": "jpg@vipnytt.no",
+                    "role": "contributor"
+                }
+            ],
+            "description": "PHP class to parse robots.txt rules according to Google, Yandex, W3C and The Web Robots Pages specifications.",
+            "homepage": "https://github.com/t1gor/Robots.txt-Parser-Class",
+            "keywords": [
+                "The Web Robots Pages",
+                "W3C",
+                "google",
+                "parser",
+                "robots.txt",
+                "yandex"
+            ],
+            "time": "2018-07-21T20:01:19+00:00"
+        },
         {
             "name": "tijsverkoyen/css-to-inline-styles",
             "version": "2.2.3",
@@ -5698,6 +5727,64 @@
             "homepage": "https://github.com/tijsverkoyen/CssToInlineStyles",
             "time": "2020-07-13T06:12:54+00:00"
         },
+        {
+            "name": "vipnytt/useragentparser",
+            "version": "v1.0.4",
+            "source": {
+                "type": "git",
+                "url": "https://github.com/VIPnytt/UserAgentParser.git",
+                "reference": "c5a6718a57088e0d45c2e36f09efabc4e008bd8c"
+            },
+            "dist": {
+                "type": "zip",
+                "url": "https://api.github.com/repos/VIPnytt/UserAgentParser/zipball/c5a6718a57088e0d45c2e36f09efabc4e008bd8c",
+                "reference": "c5a6718a57088e0d45c2e36f09efabc4e008bd8c",
+                "shasum": ""
+            },
+            "require": {
+                "php": "^5.5 || ^7.0"
+            },
+            "require-dev": {
+                "phpunit/phpunit": "^4.8.35 || ^5.7 || ^6.5"
+            },
+            "type": "library",
+            "autoload": {
+                "psr-4": {
+                    "vipnytt\\": "src/"
+                }
+            },
+            "notification-url": "https://packagist.org/downloads/",
+            "license": [
+                "MIT"
+            ],
+            "authors": [
+                {
+                    "name": "VIP nytt AS",
+                    "email": "support@vipnytt.no",
+                    "role": "Owner"
+                },
+                {
+                    "name": "Jan-Petter Gundersen",
+                    "email": "jpg@vipnytt.no",
+                    "role": "Developer"
+                }
+            ],
+            "description": "User-Agent parser for robot rule sets",
+            "homepage": "https://github.com/VIPnytt/UserAgentParser",
+            "keywords": [
+                "REP",
+                "Robots Exclusion Protocol",
+                "Robots meta tag",
+                "crawler",
+                "robot",
+                "robots.txt",
+                "spider",
+                "user-agent",
+                "useragent",
+                "x-robots-tag"
+            ],
+            "time": "2017-12-17T14:23:27+00:00"
+        },
         {
             "name": "vlucas/phpdotenv",
             "version": "v3.6.7",

storage/content_providers_lock/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore