Crazyton - Better than module_function
I created a new gem that gives you the nice syntax of module_functions but with the advantages of classes.
An example:
class ApiStub
include Crazyton
def sign_up(email, password, status = 200, response = default_sign_up_request)
stub_request(:post, "https://some-service.com")
.with(body: { email: email, password: password })
.to_return(status: status, body: response.to_json)
end
private
def default_sign_up_request
{
event: :signed_up,
user_id: 123
}
end
end
# Invoke
ApiStub.sign_up("demo@example.com", "password123")
You can get it here: GitHub - abuisman/crazyton
Why?
Modules with module_function and are a pretty good way to define helpers and make them available in a nice way:
module AppSettings
module_function
def google_analytics_enabled?
!!ENV.fetch('GOOGLE_ANALYTICS', false)
end
end
## Usage:
AppSettings.google_analytics_enabled?
But they aren’t great once things get more complex.
Let’s say that I am creating a few interfaces to some websites. Let’s call them crawlers.
This is fine:
require 'uri'
require 'net/http'
require 'openssl'
require 'nokogiri'
module AbuismanCom
module_function
def latest_post
get_selector_text("https://abuisman.com/", ".posts ul li:first-of-type")
end
def get_selector_text(url, selector)
url = URI(url)
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
http.verify_mode = OpenSSL::SSL::VERIFY_NONE
request = Net::HTTP::Get.new(url)
response = http.request(request)
response.read_body
document = Nokogiri::HTML.parse(response.read_body)
document.css(selector).text.split("\n").map(&:strip).join(" ")
end
end
puts AbuismanCom.latest_post
=> Diffing Rails credentials in Rails 6.0 10/05
But when I add another site, I recognise that I want to reuse some of the code and things are going in the wrong direction:
require 'uri'
require 'net/http'
require 'openssl'
require 'nokogiri'
module AbuismanCom
module_function
def latest_post
HttpHelper.get_selector_text("https://abuisman.com/", ".posts ul li:first-of-type")
end
end
module HackerNews
module_function
def top_post
HttpHelper.get_selector_text("https://news.ycombinator.com/", "#hnmain .itemlist tr:first-of-type")
end
end
module HttpHelper
module_function
def get_selector_text(url, selector)
url = URI(url)
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
http.verify_mode = OpenSSL::SSL::VERIFY_NONE
request = Net::HTTP::Get.new(url)
response = http.request(request)
response.read_body
document = Nokogiri::HTML.parse(response.read_body)
document.css(selector).text.split("\n").map(&:strip).join(" ")
end
end
puts AbuismanCom.latest_post
=> Diffing Rails credentials in Rails 6.0 10/05
puts HackerNews.top_post
=> 1. Tokyo police lose 2 floppy disks containing personal info on citizens (mainichi.jp)
I don’t like the call to HttpHelper.get_selector_text. You can fix that by using extend(HttpHelper) on the other two modules, so that is also ok. But I don’t want to do that for every website module that I create.
I will want to add more helpers to the crawler and a caching feature and I don’t want to keep repeating those things in a list of extend calls. I could get a long way with something like extend CrawlerStuff which in turn extends those things, but then I’d just be avoiding inheritance for the sake of avoiding it or for that sweet sweet HackerNews.top_post syntax.
Insert include Crazyton
Instead, I can use inheritance from a base class and define whatever I need there:
require 'uri'
require 'net/http'
require 'openssl'
require 'nokogiri'
require 'crazyton'
class BaseCrawler
include Crazyton
def get_selector_text(url, selector)
url = URI(url)
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
http.verify_mode = OpenSSL::SSL::VERIFY_NONE
request = Net::HTTP::Get.new(url)
response = http.request(request)
response.read_body
document = Nokogiri::HTML.parse(response.read_body)
document.css(selector).text.split("\n").map(&:strip).join(" ")
end
end
class AbuismanCom < BaseCrawler
def latest_post
get_selector_text("https://abuisman.com/", ".posts ul li:first-of-type")
end
end
class HackerNews < BaseCrawler
def top_post
get_selector_text("https://news.ycombinator.com/", "#hnmain .itemlist tr:first-of-type")
end
end
Without all the module_function and extend stuff the crawlers look a lot cleaner. The < BaseCrawler is very basic ruby stuff so I don’t need to explain to others in the team why you’d have to do some unfamiliar module setup. Even the include Crazyton happens only once inside of the BaseCrawler.
What IS strange then is that you would have to explain to the others why top_post is suddenly callable like a class method instead of it being an instance method. In the end though I like how little bootstrapping code there is in the classes.
Because we are using classes and inheritance I could also give the BaseCrawler and interface:
require 'uri'
require 'net/http'
require 'openssl'
require 'nokogiri'
require 'crazyton'
class BaseCrawler
include Crazyton
def get_selector_text(selector)
url = URI(base_url)
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
http.verify_mode = OpenSSL::SSL::VERIFY_NONE
request = Net::HTTP::Get.new(url)
response = http.request(request)
response.read_body
document = Nokogiri::HTML.parse(response.read_body)
document.css(selector).text.split("\n").map(&:strip).join(" ")
end
private
def base_url
raise "You have to implement base_url"
end
end
class AbuismanCom < BaseCrawler
def latest_post
get_selector_text(".posts ul li:first-of-type")
end
private
def base_url
"https://abuisman.com/"
end
end
class HackerNews < BaseCrawler
def top_post
get_selector_text("#hnmain .itemlist tr:first-of-type")
end
private
def base_url
"https://news.ycombinator.com/"
end
end
Ah yes, did I mention private methods? Because modules don’t have those, but classes do.
In summary, with Crazyton you get:
- smooth syntax
- inheritance
- standard looking ruby with a twist
- private methods
- a free Singleton instance
Have fun!