scrape

README

scrape

Fast web scraping

Table of Contents

• Description
• Output Format
• Usage as a CLI Application
• Usage as a Server
  • Web Interface
  • API
  • Healthchecks
  • Authorization
• Database Options
• Building and Developing
  • Building
  • Using the Docker
• Roadmap
• Acknowledgements

Description

scrape provides a self-contained low-to-no-setup tool to grab metadata and text content from web pages. The server provides a REST API to scrape web metadata, with support for batches, using either a direct client, with headless browser option that's useful for pages that need javascript to load.

Results are stored, so subsequent fetches of a particular URL are fast. Install the binary, and operate it as a shell command or as a server with a REST API. The default SQLite storage backend is performance-optimized and can store to disk or in memory. MySQL is also supported. Resources are stored with a configurable TTL.

The scrape cli tool provides shell access to scraped content via command-line entry or CSV files, and also provides database management functionality. scrape-server provides web and API access to content metadata in one-offs or batches.

RSS and Atom feeds are supported via an endpoint in scrape-server. Loading a feed returns the parsed results for all item links in the feed.

Authorization via JWT keys is supported with a configuration option. The companion scrape-jwt-encode tool can be used to generate tokens, and to make the secret you need to securely sign and verify JWT tokens.

The scrape and scrape-server binaries should be buildable and runnable in any environment where go and SQLite3 (or a MySQL server) are present. A docker build is also included. See make help for build instructions.

Output Format

JSON output is a superset of Trafilatura fields. Empty fields may be omitted in responses.

Field	Type	Description
url	String (URL)	The (canonical) URL for the page, as reported by the page itself. If the page doesn't supply that, this field will contain the same value as RequestedURL
requested_url	String (URL)	The URL that was actually requested. (Some URL params (e.g. utm_*) may be stripped before the outbound request)
original_url	String (URL)	Exactly the url that was in the inbound request
fetch_time	ISO8601	The time that URL was retrieved
fetch_method	String	The type of client used to fetch this resource (DefaultClient or HeadlessBrowser)
status_code	Int	The status code returned by the target server when fetching this page
error	String	Error message(s), if there were any, while processing this page
hostname	Domain name	The domain serving this resource
date	ISO8601	The publish date of the page, in UTC time
sitename	String	Identifies the publisher. Can be domain, company name, or other text, IRL usage not consistent
title	String	The page's title
authors	[]String	Authors
description	String	Page summary or excerpt
categories	[]String	Content categories, if supplied
tags	[]String	Tags, if supplied
language	String	2-letter language code
page_type	String	If it's there it's usually "article" following the `og`` usage
image	String (URL)	Hero image link
license	String	Generally empty
content_text	String	The text of the page, with all HTML removed

Parsed field content is largely dependent on metadata included in the page. GIGO/YMMV.

Here's an example, with long fields truncated:

{
  "url": "https://www.nasa.gov/missions/webb/nasas-webb-stuns-with-new-high-definition-look-at-exploded-star/",
  "requested_url": "https://www.nasa.gov/missions/webb/nasas-webb-stuns-with-new-high-definition-look-at-exploded-star/",
  "original_url": "https://www.nasa.gov/missions/webb/nasas-webb-stuns-with-new-high-definition-look-at-exploded-star/",
  "fetch_time": "2024-01-09T03:57:44Z",
  "fetch_method": "DefaultClient",
  "status_code": 200,
  "hostname": "www.nasa.gov",
  "date": "2023-12-10T00:00:00Z",
  "sitename": "NASA",
  "title": "NASA’s Webb Stuns With New High-Definition Look at Exploded Star - NASA",
  "authors": [
      "Steve Sabia"
  ],
  "description": "Like a shiny, round ornament ready to be placed in the perfect spot on a holiday tree, supernova remnant Cassiopeia A (Cas A) gleams in a new image from",
  "categories": [
    "Astrophysics",
    "Goddard Space Flight Center",
    "James Webb Space Telescope (JWST)",
    "Missions",
    "Nebulae",
    "Science \u0026 Research",
    "Stars",
    "Supernovae",
    "The Universe"
  ],
  "tags": [
    "Astrophysics",
    "Goddard Space Flight Center",
    "James Webb Space Telescope (JWST)",
    "Missions",
    "Nebulae",
    "Science \u0026 Research",
    "Stars",
    "Supernovae",
    "The Universe"
  ],
  "language": "en",
  "image": "https://www.nasa.gov/wp-content/uploads/2023/12/webb-stsci-01hggzdyh8ghhssnwzd71mf0xh-2k.png",
  "page_type": "article",
  "content_text": "Mysterious features hide in near-infrared light Like a shiny, round ornament ready to be placed in the perfect spot on a holiday tree, supernova remnant Cassiopeia A (Cas A) gleams in a new image from NASA’s James Webb Space Telescope. As part of the 2023...(there's about 10x more content in this example, truncated in the docs for readability)",
}

Usage as a CLI Application

Installing for shell usage

go install github.com/efixler/scrape/cmd/scrape@latest

The scrape command provides single and batch retrieval, using or bypassing the connected storage db. It also provides commands to manage the backing store.

Quickstart

> scrape http://www.foo.com/some/url/path

That's actually it. The database will be created if it doesn't exist already.

scrape % ./scrape -h
Usage: 
        scrape [flags] :url [...urls]

In addition to http[s] URLs, file:/// urls are supported, using the current working directory as the base path.

Flags:
 
  -h
        Show this help message
  -clear
        Clear the database and exit
  -csv value
        CSV file path
  -csv-column value
        The index of the column in the CSV that contains the URLs
        Environment: SCRAPE_CSV_COLUMN (default 1)
  -database value
        Database type:path
        Environment: SCRAPE_DB (default sqlite:scrape_data/scrape.db)
  -db-password value
        Database password
        Environment: SCRAPE_DB_PASSWORD
  -db-user value
        Database user
        Environment: SCRAPE_DB_USER
  -headless
        Use headless browser for extraction
  -log-level value
        Set the log level [debug|error|info|warn]
        Environment: SCRAPE_LOG_LEVEL (default WARN)
  -maintain
        Execute database maintenance and exit
  -migrate
        Migrate the database to the latest version (creating if necessary)
  -notext
        Skip text content
        Environment: SCRAPE_NOTEXT
  -ping
        Ping the database and exit
  -user-agent value
        User agent to use for fetching
        Environment: SCRAPE_USER_AGENT (default Mozilla/5.0 (X11; Linux x86_64; rv:88.0) Gecko/20100101 Firefox/88.0)

Usage as a Server

The server provides a REST API to get resource data one-at-a-time or in bulk. The root URL serves up a page that can be used to spot check results for any url.

Running scrape-server with no arguments will bring up a server on port 8080 with a SQLite database, storing scraped pages for 30 days and without any authorization controls. See below for guidance on how to change these settings.

Installation

go install github.com/efixler/scrape/cmd/scrape-server@latest

scrape % ./build/scrape-server -h

Usage:
-----
scrape-server [-port nnnn] [-h]

Some options have environment variable equivalents. Invalid environment settings
are ignored. Command line options override environment variables.

If environment variables are set, they'll override the defaults displayed in this 
help message.
 
Command line options:
--------------------

  -h
        Show this help message
  -database value
        Database type:path
        Environment: SCRAPE_DB (default sqlite:scrape_data/scrape.db)
  -db-password value
        Database password
        Environment: SCRAPE_DB_PASSWORD
  -db-user value
        Database user
        Environment: SCRAPE_DB_USER
  -enable-headless
        Enable headless browser extraction functionality
        Environment: SCRAPE_ENABLE_HEADLESS
  -log-level value
        Set the log level [debug|error|info|warn]
        Environment: SCRAPE_LOG_LEVEL (default info)
  -port value
        Port to run the server on
        Environment: SCRAPE_PORT (default 8080)
  -profile
        Enable profiling at /debug/pprof
        Environment: SCRAPE_PROFILE
  -signing-key value
        Base64 encoded HS256 key to verify JWT tokens. Required for JWT auth, and enables JWT auth if set.
        Environment: SCRAPE_SIGNING_KEY
  -ttl value
        TTL for fetched resources
        Environment: SCRAPE_TTL (default 720h0m0s)
  -user-agent value
        User agent for fetching
        Environment: SCRAPE_USER_AGENT (default Mozilla/5.0 (X11; Linux x86_64; rv:88.0) Gecko/20100101 Firefox/88.0)

Web Interface

The root path of the server (/) is browsable and provides a simple url to test URLs and results.

The pulldown on the right lets you select between loading results for a page url or for a feed, or scraping a page using a headless browser instead of a direct http client.

API

batch [POST]

Returns the metadata for the supplied list of URLs. Returned metadatas are not guaranteed to be in the same order as the request.

The batch endpoint behaves indentically to the extract endpoint in all ways except two:

1. The endpoint returns an array of the JSON payload described above
2. When individual items have errors, the request will still return with a 200 status code. Inspect the payload for individual items to determine the status of an individual item request.

Param	Description	Required
urls	A JSON array of the urls to fetch	Y

extract [GET, POST]

Fetch the metadata and text content for the specified URL. Returns JSON payload as decribed above.

If the server encounters an error fetching a requested URL, the status code for the request will be set to 422 (Unprocessable Entity). This may change.

The returned JSON payload will include a StatusCode field in all cases, along with an Error field when there's an error fetching or parsing the requested content.

Params

Param	Description	Required
url	The url to fetch. Should be url encoded.	Y

Errors

StatusCode	Description
415	The requested resource was for a content type not supported by this service
422	The request could not be completed
504	The request for the target url timed out

In all other cases, requests should return a 200 status code, and any errors received when fetching a resource will be included in the returned JSON payload.

extract/headless [GET, POST]

Identical to the extract endpoint, but uses a headless browser to scrape content instead of a direct http client. This endpoint is likely to change, with its functionality folded into the extract endpoint using a param.

feed [GET, POST]

Feed parses an RSS or Atom feed and returns the parsed results for each of the item links in the feed.

Params

Param	Description	Required
url	The feed url to fetch. Should be url encoded.	Y

Errors

StatusCode	Description
422	The url was not a valid feed
504	Request for the feed timed out

Global Params

These params work for any endpoint

Param	Value	Description
pp	1	Pretty print JSON output

Healthchecks

scrape has two healthchecks:

/.well-known/health

This is a JSON endpoint that returns data on the application's state, including memory and database runtime info.

/.well-known/heartbeat

This just returns a status 200 with the content OK

Authorization

By default, scrape runs without any authorization, all endpoints are open. JWT based authentication is supported, via the scrape-jwt-encode tool and a scrape-server configuration option. Here's how to enable it:

Generate a Secret

(./build is the path for executables built locally with make update this path if your binaries are elsewhere)

Running scrape-jwt-encode with the make-key flag will generate a cryptographically random HS256 secret, and encode it to Base64.

scrape % ./build/scrape-jwt-encode -make-key
Be sure to save this key, as it can't be re-generated:
b4RThFbyMKfQE3+jAjJcR5rjVgVOeA2Ub9eethtX83M=

You will need this secret to generate tokens and to configure the server for authentication; you'll need to save it, but don't share it via non-secure means, etc.

If the key is in your environment as SCRAPE_SIGNING_KEY it'll be picked up by both the JWT encoding tool and the server.

Generate Tokens

Tokens are also generated using scrape-jwt-encode. Here's the output of scrape-jwt-encode -h:

scrape % ./build/scrape-jwt-encode -h       

Generates JWT tokens for the scrape service. Also makes the signing key to use for the tokens.

Usage: 
-----
scrape-jwt-encode -sub subject [-signing-key key] [-exp expiration] [-aud audience]
scrape-jwt-encode -make-key

  -aud string
        Audience (recipient) for the key (default "moz")
  -exp value
        Expiration date for the key, in RFC3339 format. Default is 1 year from now. (default 2025-05-16T11:45:45.842362-04:00)
  -make-key
        Generate a new signing key
  -signing-key value
        HS256 key to sign the JWT token
        Environment: SCRAPE_SIGNING_KEY (default &[])
  -sub string
        Subject (holder name) for the key (required)

When generating a key, only a subject is required. Authorization doesn't actually check this value, but it's a good idea to use unique identifying values here, as this may get used in the future and may also get logged.

Here's the output from token generation, after putting the key from above into the environment. The claims are printed out for reference, but it's the token at the bottom that you want to share with API consumers.

scrape % export SCRAPE_SIGNING_KEY=b4RThFbyMKfQE3+jAjJcR5rjVgVOeA2Ub9eethtX83M=
scrape % ./build/scrape-jwt-encode -sub some_user

Claims:
------
{
  "iss": "scrape",
  "sub": "some_user",
  "aud": [
    "moz"
  ],
  "exp": 1747410570,
  "iat": 1715874570
}

Token:
-----
eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzY3JhcGUiLCJzdWIiOiJzb21lX3VzZXIiLCJhdWQiOlsibW96Il0sImV4cCI6MTc0NzQxMDU3MCwiaWF0IjoxNzE1ODc0NTcwfQ.4AapsWfYAK78JhP9AyhupmZAHLeRDyIPlE8ODwwsVRg

Using Tokens When Making API Requests

When authorization is enabled on the server, API requests must have an Authorization header that begins with the string Bearer (with a space) followed by the token.

Enabling Authorization In scrape-server

To enable authorization on the server either:

1. Start the server with the SCRAPE_SIGNING_KEY environment variable
2. Pass a -signing-key [key] argument to the server when starting up

When enabled, scrape will check the following qualities of the token, and reject API requests with a 401 unless there is a token passed in the Authorization header fulfilling the following criteria:

1. It's a valid JWT token
2. The token has been signed with the same signing key that the server is using
3. The token's issuer is scrape
4. The token is not expired

Healthcheck paths don't require authorization, and neither does the web interface at the root URL. (The test console uses a short-lived key to authorize requests -- it is possible to lift a key from here and use it for a little while; trying to strike a balance here between securing access and making exercising the server easy. You will also need to reload the test console periodically or calls from here will 401 as well)

Database Options

scrape supports SQLite or MySQL for data storage. Your choice depends on your requirements and environment.

SQLite

SQLite is the default storage engine for scrape. There's no need for setup and you can get running right away. The database will autocreate if it doesn't exist.

SQLite is ideal when there's a 1:1 relationship between the service and its backing store, or if you're running the service on a workstation or a 'real' computer.

If you deploy the Docker container to a cloud provider like GCP or AWS there will usually be a few hundred MB of disk storage associated with the container. Storage requirements typically translate to about 3000 url metadatas per 100MB.

When your container shuts down, previously stored data will be lost. This may or may not be a problem, depending on your application. Disk space can be monitored with the healthcheck.

It is also possible to mount a block drive to a container for persistent storage independent of the container lifecycle.

To specify a non-default path to a SQLite database using the command line -database switch or the equivalent SCRAPE_DB environment variable, use the form sqlite:/path/to.db. The special form sqlite::memory: is supported for a transient, in-memory database.

MySQL

MySQL will be a good choice for applications that run under higher volumes or where multiple service instances want to share a storage backend.

Here are the configuration options for MySQL:

Flag	Environment	Description	Example
-database	SCRAPE_DB	mysql: + addr:port	mysql:mysql.domain.co:3306
-db-password	SCRAPE_DB_PASSWORD	Password	lkajd901e109i^jhj%
-db-user	SCRAPE_DB_USER	Username for mysql connections	scrape_user

Create the MySQL database by running scrape -migrate with the applicable values above. For database creation a privileged user is required. The database will be provisioned with two roles; scrape_app for app operations and a scrape_admin role with full privileges to the schema. Assign these roles to users as appropriate.

Building and Developing

Building

Best to build with make. The Makefile has a help target, here is its output:

scrape % make help

Usage:
  make 
  build            build the binaries, to the build/ folder (default target)
  clean            clean the build directory
  docker-build     build a docker image on the current platform, for local use
  docker-push      push an amd64/arm64 docker to Docker Hub or to a registry specified by CONTAINER_REGISTRY
  docker-run       run the local docker image, binding to port 8080, or the env value of SCRAPE_PORT
  release-tag      create a release tag at the next patch version. Customize with TAG_MESSAGE and/or TAG_VERSION
  test             run the tests
  test-mysql       run the MySQL integration tests
  vet              fmt, vet, and staticcheck
  cognitive        run the cognitive complexity checker
  help             show this help message

Using the Docker

The docker-build target will build a docker on the current architecture. Run this image with docker-run (or with the normal Docker cli or desktop tools) to bring up a local instance of the service for testing. You don't need any Go tooling installed to build and run with the Docker.

To push a image to a registry, use docker-push. This will build a multiplatform amd64/arm64 image and deploy it upstream. This image is appropriate for cloud platform deployment. The registry username or organization should match the username or organization of the working repo, and you need the appropriate permissions.

The docker builds with local sources so use caution when pushing to registries.

By default, the Docker will run using a sqlite database at /scrape_data/scrape.db on the container itself. This can be changed via the SCRAPE_DB environment variable. You can also use a mount to mount this locally.

The docker-run make target will mount a local folder called docker/data and bind that to the container for file storage. If you want to use a file-based db you can use this directory, or update the Makefile to mount the desired local directory.

Roadmap

• Outbound request pacing
• Expose outbound request options (headers, etc)
• Client preference settings (e.g. use headless for specific domains)
• Authentication hooks

Feature request or bug? Post issues here.

Acknowledgements

scrape is powered by:

• go-trafilatura HTML parsing
• gofeed Atom and RSS parsing
• sqlite data storage
• go-sqlite3 SQLite client
• go-mysql MySQL client

Documentation

Overview

Package scrape provides a simple interface for fetching and storing web pages' metadata and text content. The `scrape` and `scrape-server` commands provide a command-line interface and a REST API, respectively.

Index

• type StorageBackedFetcher
  • func NewStorageBackedFetcher(fetcherFactory fetch.Factory, storageFactory store.Factory) (*StorageBackedFetcher, error)
  • func (f StorageBackedFetcher) Batch(urls []string, options fetch.BatchOptions) <-chan *resource.WebPage
  • func (f *StorageBackedFetcher) Close() error
  • func (f StorageBackedFetcher) Delete(url *nurl.URL) (bool, error)
  • func (f *StorageBackedFetcher) Fetch(url *nurl.URL) (*resource.WebPage, error)
  • func (f StorageBackedFetcher) Open(ctx context.Context) error
  • func (f *StorageBackedFetcher) WithAlternateURLFetcher(ctx context.Context, uf fetch.URLFetcher) (*StorageBackedFetcher, error)

Types

type StorageBackedFetcher

type StorageBackedFetcher struct {
	Fetcher fetch.URLFetcher
	Storage store.URLDataStore
	// contains filtered or unexported fields
}

func NewStorageBackedFetcher

func NewStorageBackedFetcher(
	fetcherFactory fetch.Factory,
	storageFactory store.Factory,
) (*StorageBackedFetcher, error)

func (StorageBackedFetcher) Batch

func (f StorageBackedFetcher) Batch(urls []string, options fetch.BatchOptions) <-chan *resource.WebPage

func (*StorageBackedFetcher) Close

func (f *StorageBackedFetcher) Close() error

Close() will be invoked when the context sent to Open() is done If that context doesn't get cancelled, Close() must be called to release resources.

func (StorageBackedFetcher) Delete

func (f StorageBackedFetcher) Delete(url *nurl.URL) (bool, error)

func (*StorageBackedFetcher) Fetch

func (f *StorageBackedFetcher) Fetch(url *nurl.URL) (*resource.WebPage, error)

func (StorageBackedFetcher) Open

func (f StorageBackedFetcher) Open(ctx context.Context) error

The context passed to Open() will be passed on to child components so that they can hook into the context directly, specifically to close and release resources on cancellation.

func (*StorageBackedFetcher) WithAlternateURLFetcher

func (f *StorageBackedFetcher) WithAlternateURLFetcher(ctx context.Context, uf fetch.URLFetcher) (*StorageBackedFetcher, error)

WithAlternateURLFetcher returns new SBF using the same storage but a different url fetcher. This is to support headless fetching, where we want to use a different underlying http client but the same storage. Call this _after_ Open() has been called. On the source fetcher. This function will Open() the passed URLFetcher with passed context, which should be the same context that was passed to Open() on the source fetcher. Do not call Open() on the returned fetcher.