fetch-politely

Fetch Politely

A library for ensuring polite outgoing HTTP requests that respect robots.txt and aren't made too close to each other

Installation

npm install fetch-politely --save

Usage

Simple:

var fetchInstance = new FetchPolitely(function (err, url, message) {
  if (err) { return; }

  // The URL has been cleared for fetching – the hostname isn't throttled and robots.txt doesn't ban it
}, {
  // Robots.txt checking requires specification of a User Agent as Robots.txt can contain User Agent specific rules
  // See http://en.wikipedia.org/wiki/User_agent for more info in format
  userAgent: 'Your-Application-Name/your.app.version (http://example.com/optional/full/app/url)',
});

// When a slot has been reserved the callback sent in the constructor will be called
fetchInstance.requestSlot('http://foo.example.org/interesting/content/');

FetchPolitely()

var fetchInstance = new FetchPolitely(callback, [options]);

Parameters

• callback – function (err, url, message, [content]) {};, called for each succesful request slot

Options

• throttleDuration – for how long in milliseconds to throttle requests to each hostname. Defaults to 10 seconds.
• returnContent – whether to fetch and return the content with the callback when a URL has received a request slot. Defaults to false.
• logger – a Bunyan compatible logger library. Defaults to bunyan-duckling which uses console.log()/.error().
• lookup – an object or class that keeps track of throttled hosts and queued URL:s. Defaults to PoliteLookup.
• lookupOptions – an object that defines extra lookup options.
• allowed – a function that checks whether a URL is allowed to be fetched. Defaults to PoliteRobot.allowed().
• robotCache – a cache method used by PoliteRobot to cache fetched robots.txt. Defaults to wrapped lru-cache.
• robotCacheLimit – a limit of the number of items to keep in the default lru-cache of PoliteRobot.
• robotPool – an HTTP agent to use for the request-library of PoliteRobot.
• userAgent – required by PoliteRobot and options.returnContent. The User Agent to use for HTTP requests.

Methods

• requestSlot – tries to reserve a request slot for a URL.

Static

• FetchPolitely.PoliteError – a very polite error object used for eg. informing about denied URL:s
• FetchPolitely.PoliteLookup – defines the interface for keeping track of throttled hosts and queued URL:s
• FetchPolitely.PolitePGLookup – alternative lookup that uses PostgreSQL as the backend
• FetchPolitely.PoliteRobot – checks whether URL:s are allowed to be fetched according to Robots.txt.

fetchInstance.requestSlot()

fetchInstance.requestSlot(url, [message], [options]);

Parameters

• url – the URL to reserve a request slot for
• message – a JSON-encodeable optional message containing eg. instructions for the FetchPolitely callback.

Options

• allow – if set to true the URL will always be allowd and not be sent to the allowed function.
• allowDuplicates – if set to false no more than one item of every url + message combination will be queued.

PoliteLookup

The simplest of simple implementations for keeping track of throttled hosts and queued URL:s. Handles it all in-memory. Same interface can be used to build a database backend for this though.

PolitePGLookup

A PostgreSQL + Knex-driven lookup that throttles hosts and queues URL using database tables.

Use by setting up the tables in pglookup.sql and include by setting the FetchPolitely options to:

{
  lookup: FetchPolitely.PolitePGLookup,
  lookupOptions: {
    knex: knexInstance
  }
}

Pull Requests are welcome if someone wants to pull out the Knex-dependency. Most projects where this has been used with Postgres has been using Knex so it got used here as well.

lookupOptions

• knex – required – the database connection to use, provided through a Knex object.
• purgeWindow – the minimum interval in milliseconds between two host purges. Defaults to 500 ms.
• concurrentReleases – how many parallell database lookups to perform to check for released URL:s. Defaults to 2.
• releasesPerBatch – how many URL:s to fetch in each database lookup. Defaults to 5.
• onlyDeduplicateMessages – bool that if set will only deduplicate URL:s with the same message when deduplicating. Defaults to false.

Lint / Test

npm test or to watch, install grunt-cli then do grunt watch