1 | const { Request, Response } = require('minipass-fetch')
|
---|
2 | const Minipass = require('minipass')
|
---|
3 | const MinipassCollect = require('minipass-collect')
|
---|
4 | const MinipassFlush = require('minipass-flush')
|
---|
5 | const MinipassPipeline = require('minipass-pipeline')
|
---|
6 | const cacache = require('cacache')
|
---|
7 | const url = require('url')
|
---|
8 |
|
---|
9 | const CachePolicy = require('./policy.js')
|
---|
10 | const cacheKey = require('./key.js')
|
---|
11 | const remote = require('../remote.js')
|
---|
12 |
|
---|
13 | const hasOwnProperty = (obj, prop) => Object.prototype.hasOwnProperty.call(obj, prop)
|
---|
14 |
|
---|
15 | // maximum amount of data we will buffer into memory
|
---|
16 | // if we'll exceed this, we switch to streaming
|
---|
17 | const MAX_MEM_SIZE = 5 * 1024 * 1024 // 5MB
|
---|
18 |
|
---|
19 | // allow list for request headers that will be written to the cache index
|
---|
20 | // note: we will also store any request headers
|
---|
21 | // that are named in a response's vary header
|
---|
22 | const KEEP_REQUEST_HEADERS = [
|
---|
23 | 'accept-charset',
|
---|
24 | 'accept-encoding',
|
---|
25 | 'accept-language',
|
---|
26 | 'accept',
|
---|
27 | 'cache-control',
|
---|
28 | ]
|
---|
29 |
|
---|
30 | // allow list for response headers that will be written to the cache index
|
---|
31 | // note: we must not store the real response's age header, or when we load
|
---|
32 | // a cache policy based on the metadata it will think the cached response
|
---|
33 | // is always stale
|
---|
34 | const KEEP_RESPONSE_HEADERS = [
|
---|
35 | 'cache-control',
|
---|
36 | 'content-encoding',
|
---|
37 | 'content-language',
|
---|
38 | 'content-type',
|
---|
39 | 'date',
|
---|
40 | 'etag',
|
---|
41 | 'expires',
|
---|
42 | 'last-modified',
|
---|
43 | 'location',
|
---|
44 | 'pragma',
|
---|
45 | 'vary',
|
---|
46 | ]
|
---|
47 |
|
---|
48 | // return an object containing all metadata to be written to the index
|
---|
49 | const getMetadata = (request, response, options) => {
|
---|
50 | const metadata = {
|
---|
51 | time: Date.now(),
|
---|
52 | url: request.url,
|
---|
53 | reqHeaders: {},
|
---|
54 | resHeaders: {},
|
---|
55 | }
|
---|
56 |
|
---|
57 | // only save the status if it's not a 200 or 304
|
---|
58 | if (response.status !== 200 && response.status !== 304)
|
---|
59 | metadata.status = response.status
|
---|
60 |
|
---|
61 | for (const name of KEEP_REQUEST_HEADERS) {
|
---|
62 | if (request.headers.has(name))
|
---|
63 | metadata.reqHeaders[name] = request.headers.get(name)
|
---|
64 | }
|
---|
65 |
|
---|
66 | // if the request's host header differs from the host in the url
|
---|
67 | // we need to keep it, otherwise it's just noise and we ignore it
|
---|
68 | const host = request.headers.get('host')
|
---|
69 | const parsedUrl = new url.URL(request.url)
|
---|
70 | if (host && parsedUrl.host !== host)
|
---|
71 | metadata.reqHeaders.host = host
|
---|
72 |
|
---|
73 | // if the response has a vary header, make sure
|
---|
74 | // we store the relevant request headers too
|
---|
75 | if (response.headers.has('vary')) {
|
---|
76 | const vary = response.headers.get('vary')
|
---|
77 | // a vary of "*" means every header causes a different response.
|
---|
78 | // in that scenario, we do not include any additional headers
|
---|
79 | // as the freshness check will always fail anyway and we don't
|
---|
80 | // want to bloat the cache indexes
|
---|
81 | if (vary !== '*') {
|
---|
82 | // copy any other request headers that will vary the response
|
---|
83 | const varyHeaders = vary.trim().toLowerCase().split(/\s*,\s*/)
|
---|
84 | for (const name of varyHeaders) {
|
---|
85 | // explicitly ignore accept-encoding here
|
---|
86 | if (name !== 'accept-encoding' && request.headers.has(name))
|
---|
87 | metadata.reqHeaders[name] = request.headers.get(name)
|
---|
88 | }
|
---|
89 | }
|
---|
90 | }
|
---|
91 |
|
---|
92 | for (const name of KEEP_RESPONSE_HEADERS) {
|
---|
93 | if (response.headers.has(name))
|
---|
94 | metadata.resHeaders[name] = response.headers.get(name)
|
---|
95 | }
|
---|
96 |
|
---|
97 | // we only store accept-encoding and content-encoding if the user
|
---|
98 | // has disabled automatic compression and decompression in minipass-fetch
|
---|
99 | // since if it's enabled (the default) then the content will have
|
---|
100 | // already been decompressed making the header a lie
|
---|
101 | if (options.compress === false) {
|
---|
102 | metadata.reqHeaders['accept-encoding'] = request.headers.get('accept-encoding')
|
---|
103 | metadata.resHeaders['content-encoding'] = response.headers.get('content-encoding')
|
---|
104 | }
|
---|
105 |
|
---|
106 | return metadata
|
---|
107 | }
|
---|
108 |
|
---|
109 | // symbols used to hide objects that may be lazily evaluated in a getter
|
---|
110 | const _request = Symbol('request')
|
---|
111 | const _response = Symbol('response')
|
---|
112 | const _policy = Symbol('policy')
|
---|
113 |
|
---|
114 | class CacheEntry {
|
---|
115 | constructor ({ entry, request, response, options }) {
|
---|
116 | if (entry) {
|
---|
117 | this.key = entry.key
|
---|
118 | this.entry = entry
|
---|
119 | // previous versions of this module didn't write an explicit timestamp in
|
---|
120 | // the metadata, so fall back to the entry's timestamp. we can't use the
|
---|
121 | // entry timestamp to determine staleness because cacache will update it
|
---|
122 | // when it verifies its data
|
---|
123 | this.entry.metadata.time = this.entry.metadata.time || this.entry.time
|
---|
124 | } else
|
---|
125 | this.key = cacheKey(request)
|
---|
126 |
|
---|
127 | this.options = options
|
---|
128 |
|
---|
129 | // these properties are behind getters that lazily evaluate
|
---|
130 | this[_request] = request
|
---|
131 | this[_response] = response
|
---|
132 | this[_policy] = null
|
---|
133 | }
|
---|
134 |
|
---|
135 | // returns a CacheEntry instance that satisfies the given request
|
---|
136 | // or undefined if no existing entry satisfies
|
---|
137 | static async find (request, options) {
|
---|
138 | try {
|
---|
139 | // compacts the index and returns an array of unique entries
|
---|
140 | var matches = await cacache.index.compact(options.cachePath, cacheKey(request), (A, B) => {
|
---|
141 | const entryA = new CacheEntry({ entry: A, options })
|
---|
142 | const entryB = new CacheEntry({ entry: B, options })
|
---|
143 | return entryA.policy.satisfies(entryB.request)
|
---|
144 | }, {
|
---|
145 | validateEntry: (entry) => {
|
---|
146 | // if an integrity is null, it needs to have a status specified
|
---|
147 | if (entry.integrity === null)
|
---|
148 | return !!(entry.metadata && entry.metadata.status)
|
---|
149 |
|
---|
150 | return true
|
---|
151 | },
|
---|
152 | })
|
---|
153 | } catch (err) {
|
---|
154 | // if the compact request fails, ignore the error and return
|
---|
155 | return
|
---|
156 | }
|
---|
157 |
|
---|
158 | // a cache mode of 'reload' means to behave as though we have no cache
|
---|
159 | // on the way to the network. return undefined to allow cacheFetch to
|
---|
160 | // create a brand new request no matter what.
|
---|
161 | if (options.cache === 'reload')
|
---|
162 | return
|
---|
163 |
|
---|
164 | // find the specific entry that satisfies the request
|
---|
165 | let match
|
---|
166 | for (const entry of matches) {
|
---|
167 | const _entry = new CacheEntry({
|
---|
168 | entry,
|
---|
169 | options,
|
---|
170 | })
|
---|
171 |
|
---|
172 | if (_entry.policy.satisfies(request)) {
|
---|
173 | match = _entry
|
---|
174 | break
|
---|
175 | }
|
---|
176 | }
|
---|
177 |
|
---|
178 | return match
|
---|
179 | }
|
---|
180 |
|
---|
181 | // if the user made a PUT/POST/PATCH then we invalidate our
|
---|
182 | // cache for the same url by deleting the index entirely
|
---|
183 | static async invalidate (request, options) {
|
---|
184 | const key = cacheKey(request)
|
---|
185 | try {
|
---|
186 | await cacache.rm.entry(options.cachePath, key, { removeFully: true })
|
---|
187 | } catch (err) {
|
---|
188 | // ignore errors
|
---|
189 | }
|
---|
190 | }
|
---|
191 |
|
---|
192 | get request () {
|
---|
193 | if (!this[_request]) {
|
---|
194 | this[_request] = new Request(this.entry.metadata.url, {
|
---|
195 | method: 'GET',
|
---|
196 | headers: this.entry.metadata.reqHeaders,
|
---|
197 | })
|
---|
198 | }
|
---|
199 |
|
---|
200 | return this[_request]
|
---|
201 | }
|
---|
202 |
|
---|
203 | get response () {
|
---|
204 | if (!this[_response]) {
|
---|
205 | this[_response] = new Response(null, {
|
---|
206 | url: this.entry.metadata.url,
|
---|
207 | counter: this.options.counter,
|
---|
208 | status: this.entry.metadata.status || 200,
|
---|
209 | headers: {
|
---|
210 | ...this.entry.metadata.resHeaders,
|
---|
211 | 'content-length': this.entry.size,
|
---|
212 | },
|
---|
213 | })
|
---|
214 | }
|
---|
215 |
|
---|
216 | return this[_response]
|
---|
217 | }
|
---|
218 |
|
---|
219 | get policy () {
|
---|
220 | if (!this[_policy]) {
|
---|
221 | this[_policy] = new CachePolicy({
|
---|
222 | entry: this.entry,
|
---|
223 | request: this.request,
|
---|
224 | response: this.response,
|
---|
225 | options: this.options,
|
---|
226 | })
|
---|
227 | }
|
---|
228 |
|
---|
229 | return this[_policy]
|
---|
230 | }
|
---|
231 |
|
---|
232 | // wraps the response in a pipeline that stores the data
|
---|
233 | // in the cache while the user consumes it
|
---|
234 | async store (status) {
|
---|
235 | // if we got a status other than 200, 301, or 308,
|
---|
236 | // or the CachePolicy forbid storage, append the
|
---|
237 | // cache status header and return it untouched
|
---|
238 | if (this.request.method !== 'GET' || ![200, 301, 308].includes(this.response.status) || !this.policy.storable()) {
|
---|
239 | this.response.headers.set('x-local-cache-status', 'skip')
|
---|
240 | return this.response
|
---|
241 | }
|
---|
242 |
|
---|
243 | const size = this.response.headers.get('content-length')
|
---|
244 | const fitsInMemory = !!size && Number(size) < MAX_MEM_SIZE
|
---|
245 | const shouldBuffer = this.options.memoize !== false && fitsInMemory
|
---|
246 | const cacheOpts = {
|
---|
247 | algorithms: this.options.algorithms,
|
---|
248 | metadata: getMetadata(this.request, this.response, this.options),
|
---|
249 | size,
|
---|
250 | memoize: fitsInMemory && this.options.memoize,
|
---|
251 | }
|
---|
252 |
|
---|
253 | let body = null
|
---|
254 | // we only set a body if the status is a 200, redirects are
|
---|
255 | // stored as metadata only
|
---|
256 | if (this.response.status === 200) {
|
---|
257 | let cacheWriteResolve, cacheWriteReject
|
---|
258 | const cacheWritePromise = new Promise((resolve, reject) => {
|
---|
259 | cacheWriteResolve = resolve
|
---|
260 | cacheWriteReject = reject
|
---|
261 | })
|
---|
262 |
|
---|
263 | body = new MinipassPipeline(new MinipassFlush({
|
---|
264 | flush () {
|
---|
265 | return cacheWritePromise
|
---|
266 | },
|
---|
267 | }))
|
---|
268 |
|
---|
269 | let abortStream, onResume
|
---|
270 | if (shouldBuffer) {
|
---|
271 | // if the result fits in memory, use a collect stream to gather
|
---|
272 | // the response and write it to cacache while also passing it through
|
---|
273 | // to the user
|
---|
274 | onResume = () => {
|
---|
275 | const collector = new MinipassCollect.PassThrough()
|
---|
276 | abortStream = collector
|
---|
277 | collector.on('collect', (data) => {
|
---|
278 | // TODO if the cache write fails, log a warning but return the response anyway
|
---|
279 | cacache.put(this.options.cachePath, this.key, data, cacheOpts).then(cacheWriteResolve, cacheWriteReject)
|
---|
280 | })
|
---|
281 | body.unshift(collector)
|
---|
282 | body.unshift(this.response.body)
|
---|
283 | }
|
---|
284 | } else {
|
---|
285 | // if it does not fit in memory, create a tee stream and use
|
---|
286 | // that to pipe to both the cache and the user simultaneously
|
---|
287 | onResume = () => {
|
---|
288 | const tee = new Minipass()
|
---|
289 | const cacheStream = cacache.put.stream(this.options.cachePath, this.key, cacheOpts)
|
---|
290 | abortStream = cacheStream
|
---|
291 | tee.pipe(cacheStream)
|
---|
292 | // TODO if the cache write fails, log a warning but return the response anyway
|
---|
293 | cacheStream.promise().then(cacheWriteResolve, cacheWriteReject)
|
---|
294 | body.unshift(tee)
|
---|
295 | body.unshift(this.response.body)
|
---|
296 | }
|
---|
297 | }
|
---|
298 |
|
---|
299 | body.once('resume', onResume)
|
---|
300 | body.once('end', () => body.removeListener('resume', onResume))
|
---|
301 | this.response.body.on('error', (err) => {
|
---|
302 | // the abortStream will either be a MinipassCollect if we buffer
|
---|
303 | // or a cacache write stream, either way be sure to listen for
|
---|
304 | // errors from the actual response and avoid writing data that we
|
---|
305 | // know to be invalid to the cache
|
---|
306 | abortStream.destroy(err)
|
---|
307 | })
|
---|
308 | } else
|
---|
309 | await cacache.index.insert(this.options.cachePath, this.key, null, cacheOpts)
|
---|
310 |
|
---|
311 | // note: we do not set the x-local-cache-hash header because we do not know
|
---|
312 | // the hash value until after the write to the cache completes, which doesn't
|
---|
313 | // happen until after the response has been sent and it's too late to write
|
---|
314 | // the header anyway
|
---|
315 | this.response.headers.set('x-local-cache', encodeURIComponent(this.options.cachePath))
|
---|
316 | this.response.headers.set('x-local-cache-key', encodeURIComponent(this.key))
|
---|
317 | this.response.headers.set('x-local-cache-mode', shouldBuffer ? 'buffer' : 'stream')
|
---|
318 | this.response.headers.set('x-local-cache-status', status)
|
---|
319 | this.response.headers.set('x-local-cache-time', new Date().toISOString())
|
---|
320 | const newResponse = new Response(body, {
|
---|
321 | url: this.response.url,
|
---|
322 | status: this.response.status,
|
---|
323 | headers: this.response.headers,
|
---|
324 | counter: this.options.counter,
|
---|
325 | })
|
---|
326 | return newResponse
|
---|
327 | }
|
---|
328 |
|
---|
329 | // use the cached data to create a response and return it
|
---|
330 | async respond (method, options, status) {
|
---|
331 | let response
|
---|
332 | const size = Number(this.response.headers.get('content-length'))
|
---|
333 | const fitsInMemory = !!size && size < MAX_MEM_SIZE
|
---|
334 | const shouldBuffer = this.options.memoize !== false && fitsInMemory
|
---|
335 | if (method === 'HEAD' || [301, 308].includes(this.response.status)) {
|
---|
336 | // if the request is a HEAD, or the response is a redirect,
|
---|
337 | // then the metadata in the entry already includes everything
|
---|
338 | // we need to build a response
|
---|
339 | response = this.response
|
---|
340 | } else {
|
---|
341 | // we're responding with a full cached response, so create a body
|
---|
342 | // that reads from cacache and attach it to a new Response
|
---|
343 | const body = new Minipass()
|
---|
344 | const removeOnResume = () => body.removeListener('resume', onResume)
|
---|
345 | let onResume
|
---|
346 | if (shouldBuffer) {
|
---|
347 | onResume = async () => {
|
---|
348 | removeOnResume()
|
---|
349 | try {
|
---|
350 | const content = await cacache.get.byDigest(this.options.cachePath, this.entry.integrity, { memoize: this.options.memoize })
|
---|
351 | body.end(content)
|
---|
352 | } catch (err) {
|
---|
353 | if (err.code === 'EINTEGRITY')
|
---|
354 | await cacache.rm.content(this.options.cachePath, this.entry.integrity, { memoize: this.options.memoize })
|
---|
355 | if (err.code === 'ENOENT' || err.code === 'EINTEGRITY')
|
---|
356 | await CacheEntry.invalidate(this.request, this.options)
|
---|
357 | body.emit('error', err)
|
---|
358 | }
|
---|
359 | }
|
---|
360 | } else {
|
---|
361 | onResume = () => {
|
---|
362 | const cacheStream = cacache.get.stream.byDigest(this.options.cachePath, this.entry.integrity, { memoize: this.options.memoize })
|
---|
363 | cacheStream.on('error', async (err) => {
|
---|
364 | cacheStream.pause()
|
---|
365 | if (err.code === 'EINTEGRITY')
|
---|
366 | await cacache.rm.content(this.options.cachePath, this.entry.integrity, { memoize: this.options.memoize })
|
---|
367 | if (err.code === 'ENOENT' || err.code === 'EINTEGRITY')
|
---|
368 | await CacheEntry.invalidate(this.request, this.options)
|
---|
369 | body.emit('error', err)
|
---|
370 | cacheStream.resume()
|
---|
371 | })
|
---|
372 | cacheStream.pipe(body)
|
---|
373 | }
|
---|
374 | }
|
---|
375 |
|
---|
376 | body.once('resume', onResume)
|
---|
377 | body.once('end', removeOnResume)
|
---|
378 | response = new Response(body, {
|
---|
379 | url: this.entry.metadata.url,
|
---|
380 | counter: options.counter,
|
---|
381 | status: 200,
|
---|
382 | headers: {
|
---|
383 | ...this.policy.responseHeaders(),
|
---|
384 | },
|
---|
385 | })
|
---|
386 | }
|
---|
387 |
|
---|
388 | response.headers.set('x-local-cache', encodeURIComponent(this.options.cachePath))
|
---|
389 | response.headers.set('x-local-cache-hash', encodeURIComponent(this.entry.integrity))
|
---|
390 | response.headers.set('x-local-cache-key', encodeURIComponent(this.key))
|
---|
391 | response.headers.set('x-local-cache-mode', shouldBuffer ? 'buffer' : 'stream')
|
---|
392 | response.headers.set('x-local-cache-status', status)
|
---|
393 | response.headers.set('x-local-cache-time', new Date(this.entry.metadata.time).toUTCString())
|
---|
394 | return response
|
---|
395 | }
|
---|
396 |
|
---|
397 | // use the provided request along with this cache entry to
|
---|
398 | // revalidate the stored response. returns a response, either
|
---|
399 | // from the cache or from the update
|
---|
400 | async revalidate (request, options) {
|
---|
401 | const revalidateRequest = new Request(request, {
|
---|
402 | headers: this.policy.revalidationHeaders(request),
|
---|
403 | })
|
---|
404 |
|
---|
405 | try {
|
---|
406 | // NOTE: be sure to remove the headers property from the
|
---|
407 | // user supplied options, since we have already defined
|
---|
408 | // them on the new request object. if they're still in the
|
---|
409 | // options then those will overwrite the ones from the policy
|
---|
410 | var response = await remote(revalidateRequest, {
|
---|
411 | ...options,
|
---|
412 | headers: undefined,
|
---|
413 | })
|
---|
414 | } catch (err) {
|
---|
415 | // if the network fetch fails, return the stale
|
---|
416 | // cached response unless it has a cache-control
|
---|
417 | // of 'must-revalidate'
|
---|
418 | if (!this.policy.mustRevalidate)
|
---|
419 | return this.respond(request.method, options, 'stale')
|
---|
420 |
|
---|
421 | throw err
|
---|
422 | }
|
---|
423 |
|
---|
424 | if (this.policy.revalidated(revalidateRequest, response)) {
|
---|
425 | // we got a 304, write a new index to the cache and respond from cache
|
---|
426 | const metadata = getMetadata(request, response, options)
|
---|
427 | // 304 responses do not include headers that are specific to the response data
|
---|
428 | // since they do not include a body, so we copy values for headers that were
|
---|
429 | // in the old cache entry to the new one, if the new metadata does not already
|
---|
430 | // include that header
|
---|
431 | for (const name of KEEP_RESPONSE_HEADERS) {
|
---|
432 | if (!hasOwnProperty(metadata.resHeaders, name) && hasOwnProperty(this.entry.metadata.resHeaders, name))
|
---|
433 | metadata.resHeaders[name] = this.entry.metadata.resHeaders[name]
|
---|
434 | }
|
---|
435 |
|
---|
436 | try {
|
---|
437 | await cacache.index.insert(options.cachePath, this.key, this.entry.integrity, {
|
---|
438 | size: this.entry.size,
|
---|
439 | metadata,
|
---|
440 | })
|
---|
441 | } catch (err) {
|
---|
442 | // if updating the cache index fails, we ignore it and
|
---|
443 | // respond anyway
|
---|
444 | }
|
---|
445 | return this.respond(request.method, options, 'revalidated')
|
---|
446 | }
|
---|
447 |
|
---|
448 | // if we got a modified response, create a new entry based on it
|
---|
449 | const newEntry = new CacheEntry({
|
---|
450 | request,
|
---|
451 | response,
|
---|
452 | options,
|
---|
453 | })
|
---|
454 |
|
---|
455 | // respond with the new entry while writing it to the cache
|
---|
456 | return newEntry.store('updated')
|
---|
457 | }
|
---|
458 | }
|
---|
459 |
|
---|
460 | module.exports = CacheEntry
|
---|