diff options
author | Satya Rohith <me@satyarohith.com> | 2022-09-28 17:41:12 +0530 |
---|---|---|
committer | GitHub <noreply@github.com> | 2022-09-28 17:41:12 +0530 |
commit | b312279e58e51520a38e51cca317a09cdadd7cb4 (patch) | |
tree | a0c6f432042ba25b569c151bbe59f1e721788d0c /ext/cache | |
parent | 1156f726a92d3d3985e591327c7526cd3e2b0473 (diff) |
feat: implement Web Cache API (#15829)
Diffstat (limited to 'ext/cache')
-rw-r--r-- | ext/cache/01_cache.js | 287 | ||||
-rw-r--r-- | ext/cache/Cargo.toml | 22 | ||||
-rw-r--r-- | ext/cache/README.md | 24 | ||||
-rw-r--r-- | ext/cache/lib.deno_cache.d.ts | 72 | ||||
-rw-r--r-- | ext/cache/lib.rs | 214 | ||||
-rw-r--r-- | ext/cache/sqlite.rs | 503 |
6 files changed, 1122 insertions, 0 deletions
diff --git a/ext/cache/01_cache.js b/ext/cache/01_cache.js new file mode 100644 index 000000000..9c624b5d7 --- /dev/null +++ b/ext/cache/01_cache.js @@ -0,0 +1,287 @@ +// Copyright 2018-2022 the Deno authors. All rights reserved. MIT license. +"use strict"; + +((window) => { + const core = window.__bootstrap.core; + const webidl = window.__bootstrap.webidl; + const { + Symbol, + TypeError, + ObjectPrototypeIsPrototypeOf, + } = window.__bootstrap.primordials; + const { + Request, + toInnerResponse, + toInnerRequest, + } = window.__bootstrap.fetch; + const { URLPrototype } = window.__bootstrap.url; + const RequestPrototype = Request.prototype; + const { getHeader } = window.__bootstrap.headers; + const { readableStreamForRid } = window.__bootstrap.streams; + + class CacheStorage { + constructor() { + webidl.illegalConstructor(); + } + + async open(cacheName) { + webidl.assertBranded(this, CacheStoragePrototype); + const prefix = "Failed to execute 'open' on 'CacheStorage'"; + webidl.requiredArguments(arguments.length, 1, { prefix }); + cacheName = webidl.converters["DOMString"](cacheName, { + prefix, + context: "Argument 1", + }); + const cacheId = await core.opAsync("op_cache_storage_open", cacheName); + return new Cache(cacheId); + } + + async has(cacheName) { + webidl.assertBranded(this, CacheStoragePrototype); + const prefix = "Failed to execute 'has' on 'CacheStorage'"; + webidl.requiredArguments(arguments.length, 1, { prefix }); + cacheName = webidl.converters["DOMString"](cacheName, { + prefix, + context: "Argument 1", + }); + return await core.opAsync("op_cache_storage_has", cacheName); + } + + async delete(cacheName) { + webidl.assertBranded(this, CacheStoragePrototype); + const prefix = "Failed to execute 'delete' on 'CacheStorage'"; + webidl.requiredArguments(arguments.length, 1, { prefix }); + cacheName = webidl.converters["DOMString"](cacheName, { + prefix, + context: "Argument 1", + }); + return await core.opAsync("op_cache_storage_delete", cacheName); + } + } + + const _id = Symbol("id"); + + class Cache { + /** @type {number} */ + [_id]; + + constructor(cacheId) { + this[_id] = cacheId; + } + + /** See https://w3c.github.io/ServiceWorker/#dom-cache-put */ + async put(request, response) { + const prefix = "Failed to execute 'put' on 'Cache'"; + webidl.requiredArguments(arguments.length, 2, { prefix }); + request = webidl.converters["RequestInfo_DOMString"](request, { + prefix, + context: "Argument 1", + }); + response = webidl.converters["Response"](response, { + prefix, + context: "Argument 2", + }); + // Step 1. + let innerRequest = null; + // Step 2. + if (ObjectPrototypeIsPrototypeOf(RequestPrototype, request)) { + innerRequest = toInnerRequest(request); + } else { + // Step 3. + innerRequest = toInnerRequest(new Request(request)); + } + // Step 4. + const reqUrl = new URL(innerRequest.url()); + if (reqUrl.protocol !== "http:" && reqUrl.protocol !== "https:") { + throw new TypeError( + "Request url protocol must be 'http:' or 'https:'", + ); + } + if (innerRequest.method !== "GET") { + throw new TypeError("Request method must be GET"); + } + // Step 5. + const innerResponse = toInnerResponse(response); + // Step 6. + if (innerResponse.status === 206) { + throw new TypeError("Response status must not be 206"); + } + // Step 7. + const varyHeader = getHeader(innerResponse.headerList, "vary"); + if (varyHeader) { + const fieldValues = varyHeader.split(",").map((field) => field.trim()); + for (const fieldValue of fieldValues) { + if ( + fieldValue === "*" + ) { + throw new TypeError("Vary header must not contain '*'"); + } + } + } + + // Step 8. + if (innerResponse.body !== null && innerResponse.body.unusable()) { + throw new TypeError("Response body must not already used"); + } + + // Remove fragment from request URL before put. + reqUrl.hash = ""; + + // Step 9-11. + const rid = await core.opAsync( + "op_cache_put", + { + cacheId: this[_id], + requestUrl: reqUrl.toString(), + responseHeaders: innerResponse.headerList, + requestHeaders: innerRequest.headerList, + responseHasBody: innerResponse.body !== null, + responseStatus: innerResponse.status, + responseStatusText: innerResponse.statusMessage, + }, + ); + if (innerResponse.body) { + const reader = innerResponse.body.stream.getReader(); + while (true) { + const { value, done } = await reader.read(); + if (done) { + await core.shutdown(rid); + core.close(rid); + break; + } else { + await core.write(rid, value); + } + } + } + // Step 12-19: TODO(@satyarohith): do the insertion in background. + } + + /** See https://w3c.github.io/ServiceWorker/#cache-match */ + async match(request, options) { + const prefix = "Failed to execute 'match' on 'Cache'"; + webidl.requiredArguments(arguments.length, 1, { prefix }); + request = webidl.converters["RequestInfo_DOMString"](request, { + prefix, + context: "Argument 1", + }); + const p = await this.#matchAll(request, options); + if (p.length > 0) { + return p[0]; + } else { + return undefined; + } + } + + /** See https://w3c.github.io/ServiceWorker/#cache-delete */ + async delete(request, _options) { + const prefix = "Failed to execute 'delete' on 'Cache'"; + webidl.requiredArguments(arguments.length, 1, { prefix }); + request = webidl.converters["RequestInfo_DOMString"](request, { + prefix, + context: "Argument 1", + }); + // Step 1. + let r = null; + // Step 2. + if (ObjectPrototypeIsPrototypeOf(RequestPrototype, request)) { + r = request; + if (request.method !== "GET") { + return false; + } + } else if ( + typeof request === "string" || + ObjectPrototypeIsPrototypeOf(URLPrototype, request) + ) { + r = new Request(request); + } + return await core.opAsync("op_cache_delete", { + cacheId: this[_id], + requestUrl: r.url, + }); + } + + /** See https://w3c.github.io/ServiceWorker/#cache-matchall + * + * Note: the function is private as we don't want to expose + * this API to the public yet. + * + * The function will return an array of responses. + */ + async #matchAll(request, _options) { + // Step 1. + let r = null; + // Step 2. + if (ObjectPrototypeIsPrototypeOf(RequestPrototype, request)) { + r = request; + if (request.method !== "GET") { + return []; + } + } else if ( + typeof request === "string" || + ObjectPrototypeIsPrototypeOf(URLPrototype, request) + ) { + r = new Request(request); + } + + // Step 5. + const responses = []; + // Step 5.2 + if (r === null) { + // Step 5.3 + // Note: we have to return all responses in the cache when + // the request is null. + // We deviate from the spec here and return an empty array + // as we don't expose matchAll() API. + return responses; + } else { + // Remove the fragment from the request URL. + const url = new URL(r.url); + url.hash = ""; + const innerRequest = toInnerRequest(r); + const matchResult = await core.opAsync( + "op_cache_match", + { + cacheId: this[_id], + requestUrl: url.toString(), + requestHeaders: innerRequest.headerList, + }, + ); + if (matchResult) { + const [meta, responseBodyRid] = matchResult; + let body = null; + if (responseBodyRid !== null) { + body = readableStreamForRid(responseBodyRid); + } + const response = new Response( + body, + { + headers: meta.responseHeaders, + status: meta.responseStatus, + statusText: meta.responseStatusText, + }, + ); + responses.push(response); + } + } + // Step 5.4-5.5: don't apply in this context. + + return responses; + } + } + + webidl.configurePrototype(CacheStorage); + webidl.configurePrototype(Cache); + const CacheStoragePrototype = CacheStorage.prototype; + + let cacheStorage; + window.__bootstrap.caches = { + CacheStorage, + Cache, + cacheStorage() { + if (!cacheStorage) { + cacheStorage = webidl.createBranded(CacheStorage); + } + return cacheStorage; + }, + }; +})(this); diff --git a/ext/cache/Cargo.toml b/ext/cache/Cargo.toml new file mode 100644 index 000000000..6f6808fe7 --- /dev/null +++ b/ext/cache/Cargo.toml @@ -0,0 +1,22 @@ +# Copyright 2018-2022 the Deno authors. All rights reserved. MIT license. + +[package] +name = "deno_cache" +version = "0.1.0" +authors = ["the Deno authors"] +edition = "2021" +license = "MIT" +readme = "README.md" +repository = "https://github.com/denoland/deno" +description = "Implementation of Cache API for Deno" + +[lib] +path = "lib.rs" + +[dependencies] +async-trait = "0.1" +deno_core = { version = "0.152.0", path = "../../core" } +rusqlite = { version = "0.28.0", features = ["unlock_notify", "bundled"] } +serde = { version = "1.0.129", features = ["derive"] } +sha2 = "0.10.2" +tokio = { version = "1.19", features = ["full"] } diff --git a/ext/cache/README.md b/ext/cache/README.md new file mode 100644 index 000000000..7e58f6e4e --- /dev/null +++ b/ext/cache/README.md @@ -0,0 +1,24 @@ +# deno_cache + +This crate implements the Cache API for Deno. + +The following APIs are implemented: + +- [`CacheStorage::open()`][cache_storage_open] +- [`CacheStorage::has()`][cache_storage_has] +- [`CacheStorage::delete()`][cache_storage_delete] +- [`Cache::match()`][cache_match] +- [`Cache::put()`][cache_put] +- [`Cache::delete()`][cache_delete] + +Cache APIs don't support the [query options][query_options] yet. + +Spec: https://w3c.github.io/ServiceWorker/#cache-interface + +[query_options]: https://w3c.github.io/ServiceWorker/#dictdef-cachequeryoptions +[cache_storage_open]: https://developer.mozilla.org/en-US/docs/Web/API/CacheStorage/open +[cache_storage_has]: https://developer.mozilla.org/en-US/docs/Web/API/CacheStorage/has +[cache_storage_delete]: https://developer.mozilla.org/en-US/docs/Web/API/CacheStorage/delete +[cache_match]: https://developer.mozilla.org/en-US/docs/Web/API/Cache/match +[cache_put]: https://developer.mozilla.org/en-US/docs/Web/API/Cache/put +[cache_delete]: https://developer.mozilla.org/en-US/docs/Web/API/Cache/delete diff --git a/ext/cache/lib.deno_cache.d.ts b/ext/cache/lib.deno_cache.d.ts new file mode 100644 index 000000000..3b03512fc --- /dev/null +++ b/ext/cache/lib.deno_cache.d.ts @@ -0,0 +1,72 @@ +// Copyright 2018-2022 the Deno authors. All rights reserved. MIT license. + +// deno-lint-ignore-file no-var + +/// <reference no-default-lib="true" /> +/// <reference lib="esnext" /> + +/** @category Cache API */ +declare var caches: CacheStorage; + +/** @category Cache API */ +declare interface CacheStorage { + /** Open a cache storage for the provided name. */ + open(cacheName: string): Promise<Cache>; + /** Check if cache already exists for the provided name. */ + has(cacheName: string): Promise<boolean>; + /** Delete cache storage for the provided name. */ + delete(cacheName: string): Promise<boolean>; +} + +/** @category Cache API */ +declare interface Cache { + /** + * Put the provided request/response into the cache. + * + * How is the API different from browsers? + * 1. You cannot match cache objects using by relative paths. + * 2. You cannot pass options like `ignoreVary`, `ignoreMethod`, `ignoreSearch`. + */ + put(request: RequestInfo | URL, response: Response): Promise<void>; + /** + * Return cache object matching the provided request. + * + * How is the API different from browsers? + * 1. You cannot match cache objects using by relative paths. + * 2. You cannot pass options like `ignoreVary`, `ignoreMethod`, `ignoreSearch`. + */ + match( + request: RequestInfo | URL, + options?: CacheQueryOptions, + ): Promise<Response | undefined>; + /** + * Delete cache object matching the provided request. + * + * How is the API different from browsers? + * 1. You cannot delete cache objects using by relative paths. + * 2. You cannot pass options like `ignoreVary`, `ignoreMethod`, `ignoreSearch`. + */ + delete( + request: RequestInfo | URL, + options?: CacheQueryOptions, + ): Promise<boolean>; +} + +/** @category Cache API */ +declare var Cache: { + prototype: Cache; + new (name: string): Cache; +}; + +/** @category Cache API */ +declare var CacheStorage: { + prototype: CacheStorage; + new (): CacheStorage; +}; + +/** @category Cache API */ +interface CacheQueryOptions { + ignoreMethod?: boolean; + ignoreSearch?: boolean; + ignoreVary?: boolean; +} diff --git a/ext/cache/lib.rs b/ext/cache/lib.rs new file mode 100644 index 000000000..350efbc38 --- /dev/null +++ b/ext/cache/lib.rs @@ -0,0 +1,214 @@ +// Copyright 2018-2022 the Deno authors. All rights reserved. MIT license. + +mod sqlite; +use deno_core::ByteString; +pub use sqlite::SqliteBackedCache; + +use async_trait::async_trait; +use deno_core::error::AnyError; +use deno_core::include_js_files; +use deno_core::op; +use deno_core::serde::Deserialize; +use deno_core::serde::Serialize; +use deno_core::Extension; +use deno_core::OpState; +use deno_core::Resource; +use deno_core::ResourceId; + +use std::cell::RefCell; +use std::path::PathBuf; +use std::rc::Rc; +use std::sync::Arc; + +#[derive(Deserialize, Serialize, Debug, Clone)] +#[serde(rename_all = "camelCase")] +pub struct CachePutRequest { + pub cache_id: i64, + pub request_url: String, + pub request_headers: Vec<(ByteString, ByteString)>, + pub response_headers: Vec<(ByteString, ByteString)>, + pub response_has_body: bool, + pub response_status: u16, + pub response_status_text: String, +} + +#[derive(Deserialize, Serialize, Debug)] +#[serde(rename_all = "camelCase")] +pub struct CacheMatchRequest { + pub cache_id: i64, + pub request_url: String, + pub request_headers: Vec<(ByteString, ByteString)>, +} + +#[derive(Debug, Deserialize, Serialize)] +#[serde(rename_all = "camelCase")] +pub struct CacheMatchResponse(CacheMatchResponseMeta, Option<ResourceId>); + +#[derive(Debug, Deserialize, Serialize)] +#[serde(rename_all = "camelCase")] +pub struct CacheMatchResponseMeta { + pub response_status: u16, + pub response_status_text: String, + pub request_headers: Vec<(ByteString, ByteString)>, + pub response_headers: Vec<(ByteString, ByteString)>, +} + +#[derive(Deserialize, Serialize, Debug)] +#[serde(rename_all = "camelCase")] +pub struct CacheDeleteRequest { + pub cache_id: i64, + pub request_url: String, +} + +#[async_trait] +pub trait Cache: Clone { + async fn storage_open(&self, cache_name: String) -> Result<i64, AnyError>; + async fn storage_has(&self, cache_name: String) -> Result<bool, AnyError>; + async fn storage_delete(&self, cache_name: String) -> Result<bool, AnyError>; + + async fn put( + &self, + request_response: CachePutRequest, + ) -> Result<Option<Rc<dyn Resource>>, AnyError>; + async fn r#match( + &self, + request: CacheMatchRequest, + ) -> Result< + Option<(CacheMatchResponseMeta, Option<Rc<dyn Resource>>)>, + AnyError, + >; + async fn delete(&self, request: CacheDeleteRequest) + -> Result<bool, AnyError>; +} + +#[op] +pub async fn op_cache_storage_open<CA>( + state: Rc<RefCell<OpState>>, + cache_name: String, +) -> Result<i64, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + cache.storage_open(cache_name).await +} + +#[op] +pub async fn op_cache_storage_has<CA>( + state: Rc<RefCell<OpState>>, + cache_name: String, +) -> Result<bool, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + cache.storage_has(cache_name).await +} + +#[op] +pub async fn op_cache_storage_delete<CA>( + state: Rc<RefCell<OpState>>, + cache_name: String, +) -> Result<bool, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + cache.storage_delete(cache_name).await +} + +#[op] +pub async fn op_cache_put<CA>( + state: Rc<RefCell<OpState>>, + request_response: CachePutRequest, +) -> Result<Option<ResourceId>, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + match cache.put(request_response).await? { + Some(resource) => { + let rid = state.borrow_mut().resource_table.add_rc_dyn(resource); + Ok(Some(rid)) + } + None => Ok(None), + } +} + +#[op] +pub async fn op_cache_match<CA>( + state: Rc<RefCell<OpState>>, + request: CacheMatchRequest, +) -> Result<Option<CacheMatchResponse>, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + match cache.r#match(request).await? { + Some((meta, None)) => Ok(Some(CacheMatchResponse(meta, None))), + Some((meta, Some(resource))) => { + let rid = state.borrow_mut().resource_table.add_rc_dyn(resource); + Ok(Some(CacheMatchResponse(meta, Some(rid)))) + } + None => Ok(None), + } +} + +#[op] +pub async fn op_cache_delete<CA>( + state: Rc<RefCell<OpState>>, + request: CacheDeleteRequest, +) -> Result<bool, AnyError> +where + CA: Cache + 'static, +{ + let cache = get_cache::<CA>(&state)?; + cache.delete(request).await +} + +pub fn get_cache<CA>(state: &Rc<RefCell<OpState>>) -> Result<CA, AnyError> +where + CA: Cache + 'static, +{ + let mut state = state.borrow_mut(); + if let Some(cache) = state.try_borrow::<CA>() { + Ok(cache.clone()) + } else { + let create_cache = state.borrow::<CreateCache<CA>>().clone(); + let cache = create_cache.0(); + state.put(cache); + Ok(state.borrow::<CA>().clone()) + } +} + +#[derive(Clone)] +pub struct CreateCache<C: Cache + 'static>(pub Arc<dyn Fn() -> C>); + +pub fn init<CA: Cache + 'static>( + maybe_create_cache: Option<CreateCache<CA>>, +) -> Extension { + Extension::builder() + .js(include_js_files!( + prefix "deno:ext/cache", + "01_cache.js", + )) + .ops(vec![ + op_cache_storage_open::decl::<CA>(), + op_cache_storage_has::decl::<CA>(), + op_cache_storage_delete::decl::<CA>(), + op_cache_put::decl::<CA>(), + op_cache_match::decl::<CA>(), + op_cache_delete::decl::<CA>(), + ]) + .state(move |state| { + if let Some(create_cache) = maybe_create_cache.clone() { + state.put(create_cache); + } + Ok(()) + }) + .build() +} + +pub fn get_declaration() -> PathBuf { + PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("lib.deno_cache.d.ts") +} diff --git a/ext/cache/sqlite.rs b/ext/cache/sqlite.rs new file mode 100644 index 000000000..1e5591839 --- /dev/null +++ b/ext/cache/sqlite.rs @@ -0,0 +1,503 @@ +// Copyright 2018-2022 the Deno authors. All rights reserved. MIT license. + +use async_trait::async_trait; +use deno_core::error::AnyError; +use deno_core::parking_lot::Mutex; +use deno_core::AsyncRefCell; +use deno_core::AsyncResult; +use deno_core::ByteString; +use deno_core::Resource; +use deno_core::ZeroCopyBuf; +use rusqlite::params; +use rusqlite::Connection; +use rusqlite::OptionalExtension; +use tokio::io::AsyncReadExt; +use tokio::io::AsyncWriteExt; + +use std::borrow::Cow; +use std::path::PathBuf; +use std::rc::Rc; +use std::sync::Arc; +use std::time::SystemTime; +use std::time::UNIX_EPOCH; + +use crate::Cache; +use crate::CacheDeleteRequest; +use crate::CacheMatchRequest; +use crate::CacheMatchResponseMeta; +use crate::CachePutRequest; + +#[derive(Clone)] +pub struct SqliteBackedCache { + pub connection: Arc<Mutex<Connection>>, + pub cache_storage_dir: PathBuf, +} + +impl SqliteBackedCache { + pub fn new(cache_storage_dir: PathBuf) -> Self { + { + std::fs::create_dir_all(&cache_storage_dir) + .expect("failed to create cache dir"); + let path = cache_storage_dir.join("cache_metadata.db"); + let connection = rusqlite::Connection::open(&path).unwrap_or_else(|_| { + panic!("failed to open cache db at {}", path.display()) + }); + connection + .execute( + "CREATE TABLE IF NOT EXISTS cache_storage ( + id INTEGER PRIMARY KEY, + cache_name TEXT NOT NULL UNIQUE + )", + (), + ) + .expect("failed to create cache_storage table"); + connection + .execute( + "CREATE TABLE IF NOT EXISTS request_response_list ( + id INTEGER PRIMARY KEY, + cache_id INTEGER NOT NULL, + request_url TEXT NOT NULL, + request_headers BLOB NOT NULL, + response_headers BLOB NOT NULL, + response_status INTEGER NOT NULL, + response_status_text TEXT, + response_body_key TEXT, + last_inserted_at INTEGER UNSIGNED NOT NULL, + FOREIGN KEY (cache_id) REFERENCES cache_storage(id) ON DELETE CASCADE, + + UNIQUE (cache_id, request_url) + )", + (), + ) + .expect("failed to create request_response_list table"); + SqliteBackedCache { + connection: Arc::new(Mutex::new(connection)), + cache_storage_dir, + } + } + } +} + +#[async_trait] +impl Cache for SqliteBackedCache { + /// Open a cache storage. Internally, this creates a row in the + /// sqlite db if the cache doesn't exist and returns the internal id + /// of the cache. + async fn storage_open(&self, cache_name: String) -> Result<i64, AnyError> { + let db = self.connection.clone(); + let cache_storage_dir = self.cache_storage_dir.clone(); + tokio::task::spawn_blocking(move || { + let db = db.lock(); + db.execute( + "INSERT OR IGNORE INTO cache_storage (cache_name) VALUES (?1)", + params![cache_name], + )?; + let cache_id = db.query_row( + "SELECT id FROM cache_storage WHERE cache_name = ?1", + params![cache_name], + |row| { + let id: i64 = row.get(0)?; + Ok(id) + }, + )?; + let responses_dir = get_responses_dir(cache_storage_dir, cache_id); + std::fs::create_dir_all(&responses_dir)?; + Ok::<i64, AnyError>(cache_id) + }) + .await? + } + + /// Check if a cache with the provided name exists. + /// Note: this doesn't check the disk, it only checks the sqlite db. + async fn storage_has(&self, cache_name: String) -> Result<bool, AnyError> { + let db = self.connection.clone(); + tokio::task::spawn_blocking(move || { + let db = db.lock(); + let cache_exists = db.query_row( + "SELECT count(cache_name) FROM cache_storage WHERE cache_name = ?1", + params![cache_name], + |row| { + let count: i64 = row.get(0)?; + Ok(count > 0) + }, + )?; + Ok::<bool, AnyError>(cache_exists) + }) + .await? + } + + /// Delete a cache storage. Internally, this deletes the row in the sqlite db. + async fn storage_delete(&self, cache_name: String) -> Result<bool, AnyError> { + let db = self.connection.clone(); + let cache_storage_dir = self.cache_storage_dir.clone(); + tokio::task::spawn_blocking(move || { + let db = db.lock(); + let maybe_cache_id = db + .query_row( + "DELETE FROM cache_storage WHERE cache_name = ?1 RETURNING id", + params![cache_name], + |row| { + let id: i64 = row.get(0)?; + Ok(id) + }, + ) + .optional()?; + if let Some(cache_id) = maybe_cache_id { + let cache_dir = cache_storage_dir.join(cache_id.to_string()); + if cache_dir.exists() { + std::fs::remove_dir_all(cache_dir)?; + } + } + Ok::<bool, AnyError>(maybe_cache_id.is_some()) + }) + .await? + } + + async fn put( + &self, + request_response: CachePutRequest, + ) -> Result<Option<Rc<dyn Resource>>, AnyError> { + let db = self.connection.clone(); + let cache_storage_dir = self.cache_storage_dir.clone(); + let now = SystemTime::now().duration_since(UNIX_EPOCH)?; + let response_body_key = if request_response.response_has_body { + Some(hash(&format!( + "{}_{}", + &request_response.request_url, + now.as_nanos() + ))) + } else { + None + }; + + if let Some(body_key) = response_body_key { + let responses_dir = + get_responses_dir(cache_storage_dir, request_response.cache_id); + let response_path = responses_dir.join(&body_key); + let file = tokio::fs::File::create(response_path).await?; + Ok(Some(Rc::new(CachePutResource { + file: AsyncRefCell::new(file), + db, + put_request: request_response, + response_body_key: body_key, + start_time: now.as_secs(), + }))) + } else { + insert_cache_asset(db, request_response, None).await?; + Ok(None) + } + } + + async fn r#match( + &self, + request: CacheMatchRequest, + ) -> Result< + Option<(CacheMatchResponseMeta, Option<Rc<dyn Resource>>)>, + AnyError, + > { + let db = self.connection.clone(); + let cache_storage_dir = self.cache_storage_dir.clone(); + let query_result = tokio::task::spawn_blocking(move || { + let db = db.lock(); + let result = db.query_row( + "SELECT response_body_key, response_headers, response_status, response_status_text, request_headers + FROM request_response_list + WHERE cache_id = ?1 AND request_url = ?2", + (request.cache_id, &request.request_url), + |row| { + let response_body_key: Option<String> = row.get(0)?; + let response_headers: Vec<u8> = row.get(1)?; + let response_status: u16 = row.get(2)?; + let response_status_text: String = row.get(3)?; + let request_headers: Vec<u8> = row.get(4)?; + let response_headers: Vec<(ByteString, ByteString)> = deserialize_headers(&response_headers); + let request_headers: Vec<(ByteString, ByteString)> = deserialize_headers(&request_headers); + Ok((CacheMatchResponseMeta {request_headers, response_headers,response_status,response_status_text}, response_body_key)) + }, + ); + result.optional() + }) + .await??; + + match query_result { + Some((cache_meta, Some(response_body_key))) => { + // From https://w3c.github.io/ServiceWorker/#request-matches-cached-item-algorithm + // If there's Vary header in the response, ensure all the + // headers of the cached request match the query request. + if let Some(vary_header) = + get_header("vary", &cache_meta.response_headers) + { + if !vary_header_matches( + &vary_header, + &request.request_headers, + &cache_meta.request_headers, + ) { + return Ok(None); + } + } + let response_path = + get_responses_dir(cache_storage_dir, request.cache_id) + .join(response_body_key); + let file = tokio::fs::File::open(response_path).await?; + return Ok(Some(( + cache_meta, + Some(Rc::new(CacheResponseResource::new(file))), + ))); + } + Some((cache_meta, None)) => { + return Ok(Some((cache_meta, None))); + } + None => return Ok(None), + } + } + + async fn delete( + &self, + request: CacheDeleteRequest, + ) -> Result<bool, AnyError> { + let db = self.connection.clone(); + tokio::task::spawn_blocking(move || { + // TODO(@satyarohith): remove the response body from disk if one exists + let db = db.lock(); + let rows_effected = db.execute( + "DELETE FROM request_response_list WHERE cache_id = ?1 AND request_url = ?2", + (request.cache_id, &request.request_url), + )?; + Ok::<bool, AnyError>(rows_effected > 0) + }) + .await? + } +} + +async fn insert_cache_asset( + db: Arc<Mutex<rusqlite::Connection>>, + put: CachePutRequest, + body_key_start_time: Option<(String, u64)>, +) -> Result<Option<String>, deno_core::anyhow::Error> { + tokio::task::spawn_blocking(move || { + let maybe_response_body = { + let db = db.lock(); + let mut response_body_key = None; + if let Some((body_key, start_time)) = body_key_start_time { + response_body_key = Some(body_key); + let last_inserted_at = db.query_row(" + SELECT last_inserted_at FROM request_response_list + WHERE cache_id = ?1 AND request_url = ?2", + (put.cache_id, &put.request_url), |row| { + let last_inserted_at: i64 = row.get(0)?; + Ok(last_inserted_at) + }).optional()?; + if let Some(last_inserted) = last_inserted_at { + // Some other worker has already inserted this resource into the cache. + // Note: okay to unwrap() as it is always present when response_body_key is present. + if start_time > (last_inserted as u64) { + return Ok(None); + } + } + } + db.query_row( + "INSERT OR REPLACE INTO request_response_list + (cache_id, request_url, request_headers, response_headers, + response_body_key, response_status, response_status_text, last_inserted_at) + VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8) + RETURNING response_body_key", + ( + put.cache_id, + put.request_url, + serialize_headers(&put.request_headers), + serialize_headers(&put.response_headers), + response_body_key, + put.response_status, + put.response_status_text, + SystemTime::now().duration_since(UNIX_EPOCH)?.as_secs(), + ), + |row| { + let response_body_key: Option<String> = row.get(0)?; + Ok(response_body_key) + }, + )? + }; + Ok::<Option<String>, AnyError>(maybe_response_body) + }).await? +} + +#[inline] +fn get_responses_dir(cache_storage_dir: PathBuf, cache_id: i64) -> PathBuf { + cache_storage_dir + .join(cache_id.to_string()) + .join("responses") +} + +/// Check if the headers provided in the vary_header match +/// the query request headers and the cached request headers. +fn vary_header_matches( + vary_header: &ByteString, + query_request_headers: &[(ByteString, ByteString)], + cached_request_headers: &[(ByteString, ByteString)], +) -> bool { + let vary_header = match std::str::from_utf8(vary_header) { + Ok(vary_header) => vary_header, + Err(_) => return false, + }; + let headers = get_headers_from_vary_header(vary_header); + for header in headers { + let query_header = get_header(&header, query_request_headers); + let cached_header = get_header(&header, cached_request_headers); + if query_header != cached_header { + return false; + } + } + true +} + +fn get_headers_from_vary_header(vary_header: &str) -> Vec<String> { + vary_header + .split(',') + .map(|s| s.trim().to_lowercase()) + .collect() +} + +fn get_header( + name: &str, + headers: &[(ByteString, ByteString)], +) -> Option<ByteString> { + headers + .iter() + .find(|(k, _)| { + if let Ok(k) = std::str::from_utf8(k) { + k.eq_ignore_ascii_case(name) + } else { + false + } + }) + .map(|(_, v)| v.to_owned()) +} + +impl deno_core::Resource for SqliteBackedCache { + fn name(&self) -> std::borrow::Cow<str> { + "SqliteBackedCache".into() + } +} + +pub struct CachePutResource { + pub db: Arc<Mutex<rusqlite::Connection>>, + pub put_request: CachePutRequest, + pub response_body_key: String, + pub file: AsyncRefCell<tokio::fs::File>, + pub start_time: u64, +} + +impl CachePutResource { + async fn write(self: Rc<Self>, data: ZeroCopyBuf) -> Result<usize, AnyError> { + let resource = deno_core::RcRef::map(&self, |r| &r.file); + let mut file = resource.borrow_mut().await; + file.write_all(&data).await?; + Ok(data.len()) + } + + async fn shutdown(self: Rc<Self>) -> Result<(), AnyError> { + let resource = deno_core::RcRef::map(&self, |r| &r.file); + let mut file = resource.borrow_mut().await; + file.flush().await?; + file.sync_all().await?; + insert_cache_asset( + self.db.clone(), + self.put_request.clone(), + Some((self.response_body_key.clone(), self.start_time)), + ) + .await?; + Ok(()) + } +} + +impl Resource for CachePutResource { + fn name(&self) -> Cow<str> { + "CachePutResource".into() + } + + fn write(self: Rc<Self>, buf: ZeroCopyBuf) -> AsyncResult<usize> { + Box::pin(self.write(buf)) + } + + fn shutdown(self: Rc<Self>) -> AsyncResult<()> { + Box::pin(self.shutdown()) + } +} + +pub struct CacheResponseResource { + file: AsyncRefCell<tokio::fs::File>, +} + +impl CacheResponseResource { + fn new(file: tokio::fs::File) -> Self { + Self { + file: AsyncRefCell::new(file), + } + } + + async fn read( + self: Rc<Self>, + mut buf: ZeroCopyBuf, + ) -> Result<(usize, ZeroCopyBuf), AnyError> { + let resource = deno_core::RcRef::map(&self, |r| &r.file); + let mut file = resource.borrow_mut().await; + let nread = file.read(&mut buf).await?; + Ok((nread, buf)) + } +} + +impl Resource for CacheResponseResource { + fn name(&self) -> Cow<str> { + "CacheResponseResource".into() + } + + fn read_return( + self: Rc<Self>, + buf: ZeroCopyBuf, + ) -> AsyncResult<(usize, ZeroCopyBuf)> { + Box::pin(self.read(buf)) + } +} + +pub fn hash(token: &str) -> String { + use sha2::Digest; + format!("{:x}", sha2::Sha256::digest(token.as_bytes())) +} + +fn serialize_headers(headers: &[(ByteString, ByteString)]) -> Vec<u8> { + let mut serialized_headers = Vec::new(); + for (name, value) in headers { + serialized_headers.extend_from_slice(name); + serialized_headers.extend_from_slice(b"\r\n"); + serialized_headers.extend_from_slice(value); + serialized_headers.extend_from_slice(b"\r\n"); + } + serialized_headers +} + +fn deserialize_headers( + serialized_headers: &[u8], +) -> Vec<(ByteString, ByteString)> { + let mut headers = Vec::new(); + let mut piece = None; + let mut start = 0; + for (i, byte) in serialized_headers.iter().enumerate() { + if byte == &b'\r' && serialized_headers.get(i + 1) == Some(&b'\n') { + if piece.is_none() { + piece = Some(start..i); + } else { + let name = piece.unwrap(); + let value = start..i; + headers.push(( + ByteString::from(&serialized_headers[name]), + ByteString::from(&serialized_headers[value]), + )); + piece = None; + } + start = i + 2; + } + } + assert!(piece.is_none()); + assert_eq!(start, serialized_headers.len()); + headers +} |