抛弃Cloudflare Page Rule，拥抱Workers–自定义CDN缓存策略

Cloudflare一直是我非常喜欢的厂商，免费的套餐就可以提供不错的服务，只不过现在被玩坏了，国内访问速度大不如前，但是总体来说有还是比没有强。

不过免费版套餐的Cloudflare CDN有个非常致命的缺点，就是不能根据cookie来区分访客，从而针对性提供访问内容（比如不为登陆、评论过的用户提供缓存）。并且还会把已登录或已评论用户的信息缓存上去，这点难以接受，也是我抛弃Cloudflare的重要原因之一。

不过在Cloudflare Workers面世后，这个问题就迎刃而解了，我们可以使用Workers来绕过这个限制，自己编写缓存策略。

介绍

关于Cloudflare Workers的介绍可以自行搜索，通过搜索引擎看到本文人或许应该已经有所了解。

我简单阐述一下，本文要讲的内容，就是通过Cloudflare官方曾编写的一款脚本，实现自定义缓存策略。

Edge Cache HTML

Cloudflare官方曾编写的一款WordPress插件，用以搭配edge-cache-html，不过这个项目已经快两年没更新了，应该是觉得这个项目会影响到他们付费套餐的出售。

https://github.com/cloudflare/worker-examples/tree/master/examples/edge-cache-html

目前这个脚本直接使用，是无法在WordPress下生效的，并且缺失URL路径规则。我稍作修改，增加了路径排除和支持在WordPress下是缓存（去除header cache判断）。

前期操作请参考 [WordPress]利用 Cloudflare Workers 来缓存博客的 HTML 网页 ，写的非常详细，只不过这位博主没有没有缓存成功。

以下是修改后的脚本

// IMPORTANT: Either A Key/Value Namespace must be bound to this worker script
// using the variable name EDGE_CACHE. or the API parameters below should be
// configured. KV is recommended if possible since it can purge just the HTML
// instead of the full cache.


// Default cookie prefixes for bypass
const DEFAULT_BYPASS_COOKIES = [
"wp-",
"wordpress",
"comment_",
"woocommerce_"
];

// URL paths to bypass the cache (each pattern is a regex)
const BYPASS_URL_PATTERNS = [
/\/wp-admin\/.*/,
/\/wp-adminlogin\/.*/
];

/**
     * Main worker entry point. 
     */
addEventListener("fetch", event => {
const request = event.request;
let upstreamCache = request.headers.get('x-HTML-Edge-Cache');

// Only process requests if KV store is set up and there is no
// HTML edge cache in front of this worker (only the outermost cache
// should handle HTML caching in case there are varying levels of support).
let configured = false;
if (typeof EDGE_CACHE !== 'undefined') {
configured = true;
} else if (CLOUDFLARE_API.email.length && CLOUDFLARE_API.key.length && CLOUDFLARE_API.zone.length) {
configured = true;
}

// Bypass processing of image requests (for everything except Firefox which doesn't use image/*)
const accept = request.headers.get('Accept');
let isImage = false;
if (accept && (accept.indexOf('image/*') !== -1)) {
isImage = true;
}

if (configured && !isImage && upstreamCache === null) {
event.passThroughOnException();
event.respondWith(processRequest(request, event));
}
});

/**
     * Process every request coming through to add the edge-cache header,
     * watch for purge responses and possibly cache HTML GET requests.
     * 
     * @param {Request} originalRequest - Original request
     * @param {Event} event - Original event (for additional async waiting)
     */
async function processRequest(originalRequest, event) {
let cfCacheStatus = null;
const accept = originalRequest.headers.get('Accept');
const isHTML = (accept && accept.indexOf('text/html') >= 0);
let {response, cacheVer, status, bypassCache} = await getCachedResponse(originalRequest);

if (response === null) {
// Clone the request, add the edge-cache header and send it through.
let request = new Request(originalRequest);
request.headers.set('x-HTML-Edge-Cache', 'supports=cache|purgeall|bypass-cookies');
response = await fetch(request);

if (response) {
const options = getResponseOptions(response);
if (options && options.purge) {
await purgeCache(cacheVer, event);
status += ', Purged';
}
bypassCache = bypassCache || shouldBypassEdgeCache(request, response);
if ((!options || options.cache) && isHTML &&
originalRequest.method === 'GET' && response.status === 200 &&
!bypassCache) {
status += await cacheResponse(cacheVer, originalRequest, response, event);
}
}
} else {
// If the origin didn't send the control header we will send the cached response but update
// the cached copy asynchronously (stale-while-revalidate). This commonly happens with
// a server-side disk cache that serves the HTML directly from disk.
cfCacheStatus = 'HIT';
if (originalRequest.method === 'GET' && response.status === 200 && isHTML) {
bypassCache = bypassCache || shouldBypassEdgeCache(originalRequest, response);
if (!bypassCache) {
const options = getResponseOptions(response);
if (!options) {
status += ', Refreshed';
event.waitUntil(updateCache(originalRequest, cacheVer, event));
}
}
}
}

if (response && status !== null && originalRequest.method === 'GET' && response.status === 200 && isHTML) {
response = new Response(response.body, response);
response.headers.set('x-HTML-Edge-Cache-Status', status);
if (cacheVer !== null) {
response.headers.set('x-HTML-Edge-Cache-Version', cacheVer.toString());
}
if (cfCacheStatus) {
response.headers.set('CF-Cache-Status', cfCacheStatus);
}
}

return response;
}

/**
     * Determine if the cache should be bypassed for the given request/response pair.
     * Specifically, if the request includes a cookie that the response flags for bypass.
     * Can be used on cache lookups to determine if the request needs to go to the origin and
     * origin responses to determine if they should be written to cache.
     * @param {Request} request - Request
     * @param {Response} response - Response
     * @returns {bool} true if the cache should be bypassed
     */
function shouldBypassEdgeCache(request, response) {
let bypassCache = false;

// Bypass the cache for all requests to a URL that matches any of the URL path bypass patterns
const url = new URL(request.url);
const path = url.pathname + url.search;
if (BYPASS_URL_PATTERNS.length) {
for (let pattern of BYPASS_URL_PATTERNS) {
if (path.match(pattern)) {
bypassCache = true;
break;
}
}
}

if (request && response) {
const options = getResponseOptions(response);
const cookieHeader = request.headers.get('cookie');
let bypassCookies = DEFAULT_BYPASS_COOKIES;
if (options) {
bypassCookies = options.bypassCookies;
}
if (cookieHeader && cookieHeader.length && bypassCookies.length) {
const cookies = cookieHeader.split(';');
for (let cookie of cookies) {
// See if the cookie starts with any of the logged-in user prefixes
for (let prefix of bypassCookies) {
if (cookie.trim().startsWith(prefix)) {
bypassCache = true;
break;
}
}
if (bypassCache) {
break;
}
}
}
}

return bypassCache;
}

const CACHE_HEADERS = ['Cache-Control', 'Expires', 'Pragma'];

/**
     * Check for cached HTML GET requests.
     * 
     * @param {Request} request - Original request
     */
async function getCachedResponse(request) {
let response = null;
let cacheVer = null;
let bypassCache = false;
let status = 'Miss';

// Only check for HTML GET requests (saves on reading from KV unnecessarily)
// and not when there are cache-control headers on the request (refresh)
const accept = request.headers.get('Accept');
const cacheControl = request.headers.get('Cache-Control');
let noCache = false;
// if (cacheControl && cacheControl.indexOf('no-cache') !== -1) {
//   noCache = true;
//   status = 'Bypass for Reload';
// }
if (!noCache && request.method === 'GET' && accept && accept.indexOf('text/html') >= 0) {
// Build the versioned URL for checking the cache
cacheVer = await GetCurrentCacheVersion(cacheVer);
const cacheKeyRequest = GenerateCacheRequest(request, cacheVer);

// See if there is a request match in the cache
try {
let cache = caches.default;
let cachedResponse = await cache.match(cacheKeyRequest);
if (cachedResponse) {
// Copy Response object so that we can edit headers.
cachedResponse = new Response(cachedResponse.body, cachedResponse);

// Check to see if the response needs to be bypassed because of a cookie
bypassCache = shouldBypassEdgeCache(request, cachedResponse);

// Copy the original cache headers back and clean up any control headers
if (bypassCache) {
status = 'Bypass Cookie';
} else {
status = 'Hit';
cachedResponse.headers.delete('Cache-Control');
cachedResponse.headers.delete('x-HTML-Edge-Cache-Status');
for (header of CACHE_HEADERS) {
let value = cachedResponse.headers.get('x-HTML-Edge-Cache-Header-' + header);
if (value) {
cachedResponse.headers.delete('x-HTML-Edge-Cache-Header-' + header);
cachedResponse.headers.set(header, value);
}
}
response = cachedResponse;
}
} else {
status = 'Miss';
}