bring cookie handling in-house. Fixes issue #324 (#325)

* bring cookie handling in-house. Fixes issue #324

Co-authored-by: Glynn Bird <glynnbird@apache.org>

Author: glynnbird

lib/cookie.js

@@ -0,0 +1,129 @@
+const { URL } = require('url')
+
+// a simple cookie jar
+class CookieJar {
+  // create new empty cookie jar
+  constructor () {
+    this.jar = []
+  }
+
+  // remove expired cookies
+  clean () {
+    const now = new Date().getTime()
+    for (let i = 0; i < this.jar.length; i++) {
+      const c = this.jar[i]
+      if (c.ts < now) {
+        this.jar.splice(i, 1)
+        i--
+      }
+    }
+  }
+
+  // add a cookie to the jar
+  add (cookie, url) {
+    // see if we have this cookie already
+    const oldCookieIndex = this.findByName(url, cookie.name)
+
+    // if we do, update it
+    if (oldCookieIndex >= 0) {
+      // update existing cookie
+      this.jar[oldCookieIndex].value = cookie.value
+      this.jar[oldCookieIndex].expires = cookie.expires
+      this.jar[oldCookieIndex].ts = new Date(cookie.expires).getTime()
+    } else {
+      // otherwise, just add it
+      this.jar.push(cookie)
+    }
+  }
+
+  // locate a cookie by name & url
+  findByName (url, name) {
+    this.clean()
+    const now = new Date().getTime()
+    const parsedURL = new URL(url)
+    for (let i = 0; i < this.jar.length; i++) {
+      const c = this.jar[i]
+      if (c.origin === parsedURL.origin &&
+          c.name === name &&
+          c.ts >= now) {
+        return i
+      }
+    }
+    return -1
+  }
+
+  // get a list of cookies to send for a supplied URL
+  getCookieString (url) {
+    let i
+    // clean up deceased cookies
+    this.clean()
+
+    // find cookies that match the url
+    const now = new Date().getTime()
+    const parsedURL = new URL(url)
+    const retval = []
+    for (i = 0; i < this.jar.length; i++) {
+      const c = this.jar[i]
+      // if match domain name and timestamp
+      if ((c.origin === parsedURL.origin ||
+          (c.domain && parsedURL.hostname.endsWith(c.domain))) &&
+          c.ts >= now) {
+        // if cookie has httponly flag and this is not http(s), ignore
+        if (c.httponly && !['http:', 'https:'].includes(parsedURL.protocol)) {
+          continue
+        }
+
+        // if cookie has a path and it doesn't match incoming url, ignore
+        if (c.path && !parsedURL.pathname.startsWith(c.path)) {
+          continue
+        }
+
+        // if cookie has a secure flag and the transport isn't secure, ignore
+        if (c.secure && parsedURL.protocol !== 'https:') {
+          continue
+        }
+
+        // add to list of returned cookies
+        retval.push(c.value)
+      }
+    }
+    // if we've got cookies to return
+    if (retval.length > 0) {
+      // join them with semi-colons
+      return retval.join('; ')
+    } else {
+      // otherwise a blank string
+      return ''
+    }
+  }
+
+  // parse a 'set-cookie' header of the form:
+  //   AuthSession=YWRtaW46NjM5ODgzQ0Y6TuB66MczvkZ7axEJq6Fz0gOdhKY; Version=1; Expires=Tue, 13-Dec-2022 13:54:19 GMT; Max-Age=60; Path=/; HttpOnly
+  parse (h, url) {
+    const parsedURL = new URL(url)
+
+    // split components by ; and remove whitespace
+    const bits = h.split(';').map(s => s.trim())
+
+    // extract the cookie's value from the start of the string
+    const cookieValue = bits.shift()
+
+    // start a cookie object
+    const cookie = {
+      name: cookieValue.split('=')[0], // the first part of the value
+      origin: parsedURL.origin,
+      pathname: parsedURL.pathname,
+      protocol: parsedURL.protocol
+    }
+    bits.forEach((e) => {
+      const lr = e.split('=')
+      cookie[lr[0].toLowerCase()] = lr[1] || true
+    })
+    // calculate expiry timestamp
+    cookie.ts = new Date(cookie.expires).getTime()
+    cookie.value = cookieValue
+    this.add(cookie, url)
+  }
+}
+
+module.exports = CookieJar

lib/nano.js

@@ -10,20 +10,20 @@
 // License for the specific language governing permissions and limitations under
 // the License.
 
-const { HttpsCookieAgent, HttpCookieAgent } = require('http-cookie-agent/http')
 const { URL } = require('url')
+const http = require('http')
+const https = require('https')
 const assert = require('assert')
 const querystring = require('qs')
 const axios = require('axios')
-const { CookieJar } = require('tough-cookie')
-const cookieJar = new CookieJar()
 const stream = require('stream')
 const pkg = require('../package.json')
-const AGENT_DEFAULTS = { cookies: { jar: cookieJar }, keepAlive: true, maxSockets: 50, keepAliveMsecs: 30000 }
+const AGENT_DEFAULTS = { keepAlive: true, maxSockets: 50, keepAliveMsecs: 30000 }
+const defaultHttpAgent = new http.Agent(AGENT_DEFAULTS)
+const defaultHttpsAgent = new https.Agent(AGENT_DEFAULTS)
 const SCRUBBED_STR = 'XXXXXX'
-const defaultHttpAgent = new HttpCookieAgent(AGENT_DEFAULTS)
-const defaultHttpsAgent = new HttpsCookieAgent(AGENT_DEFAULTS)
 const ChangesReader = require('./changesreader.js')
+const CookieJar = require('./cookie.js')
 const MultiPartFactory = require('./multipart.js')
 
 function isEmpty (val) {
@@ -77,6 +77,9 @@ module.exports = exports = function dbScope (cfg) {
   const log = typeof cfg.log === 'function' ? cfg.log : dummyLogger
   const parseUrl = 'parseUrl' in cfg ? cfg.parseUrl : true
 
+  // create cookieJar for this Nano
+  cfg.cookieJar = new CookieJar()
+
   function maybeExtractDatabaseComponent () {
     if (!parseUrl) {
       return
@@ -123,6 +126,16 @@ module.exports = exports = function dbScope (cfg) {
     let body = response.data
     response.statusCode = statusCode
 
+    // cookie parsing
+    if (response.headers) {
+      const h = response.headers['set-cookie']
+      if (h && h.length) {
+        h.forEach((header) => {
+          cfg.cookieJar.parse(header, req.url)
+        })
+      }
+    }
+
     // let parsed
     const responseHeaders = Object.assign({
       uri: scrubURL(req.url),
@@ -282,7 +295,6 @@ module.exports = exports = function dbScope (cfg) {
     }
 
     if (isJar) {
-      req.jar = cookieJar
       req.withCredentials = true
     }
 
@@ -350,6 +362,12 @@ module.exports = exports = function dbScope (cfg) {
       req.qs = qs
     }
 
+    // add any cookies for this domain
+    const cookie = cfg.cookieJar.getCookieString(req.uri)
+    if (cookie) {
+      req.headers.cookie = cookie
+    }
+
     if (opts.body) {
       if (Buffer.isBuffer(opts.body) || opts.dontStringify) {
         req.body = opts.body
@@ -375,8 +393,6 @@ module.exports = exports = function dbScope (cfg) {
     // ?drilldown=["author","Dickens"]&drilldown=["publisher","Penguin"]
     req.qsStringifyOptions = { arrayFormat: 'repeat' }
 
-    cfg.cookies = cookieJar.getCookiesSync(cfg.url)
-
     // This where the HTTP request is made.
     // Nano used to use the now-deprecated "request" library but now we're going to
     // use axios, so let's modify the "req" object to suit axios
@@ -409,8 +425,6 @@ module.exports = exports = function dbScope (cfg) {
     // add http agents
     req.httpAgent = cfg.requestDefaults.agent || defaultHttpAgent
     req.httpsAgent = cfg.requestDefaults.agent || defaultHttpsAgent
-    req.httpAgent.jar = req.httpAgent.jar ? req.httpAgent.jar : cookieJar
-    req.httpsAgent.jar = req.httpsAgent.jar ? req.httpsAgent.jar : cookieJar
     const ax = axios.create({
       httpAgent: req.httpAgent,
       httpsAgent: req.httpsAgent