crawl jobs

app/models/industries_job.rb

-class IndustriesJob < ApplicationRecord
-  belongs_to :job
-  belongs_to :industry
-end

app/models/job.rb

@@ -3,6 +3,7 @@ class Job < ApplicationRecord
   has_many :applied_jobs
   has_many :histories
   has_many :favorites
+  has_many :locations_jobs
   has_many :locations, through: :locations_jobs
   has_and_belongs_to_many :industries
 end

app/models/location.rb

 class Location < ApplicationRecord
   CITY_VIETNAM_NUMBER = 70
+  has_many :locations_jobs
   has_many :jobs, through: :locations_jobs
 end

lib/tasks/crawler.rake

 require "open-uri"
-task crawl_companies_jobs: :environment do
-  crawl_companies_and_jobs(10)
+task crawl_jobs: :environment do
+  job_links = get_job_links(1)
+  crawl_jobs(job_links)
 end
 
 task crawl_industries_locations: :environment do
   crawl_industries_and_locations
 end
 
-def crawl_companies_and_jobs(page)
+def get_job_links(page)
+  job_links = []
   page.times do |i|
-    company_links, job_links = get_company_and_job_links("https://careerbuilder.vn/viec-lam/tat-ca-viec-lam-trang-#{i}-vi.html")
-    crawl_companies(company_links)
-    crawl_jobs(job_links)
-  end
-end
-
-def get_company_and_job_links(base_link)
-  document = Nokogiri::HTML(open(base_link))
-  companies_xml = document.xpath('//div/a[@class="company-name"]/@href')
-  company_links = companies_xml.map(&:value)
+    document = Nokogiri::HTML(open("https://careerbuilder.vn/viec-lam/tat-ca-viec-lam-trang-#{i}-vi.html"))
     jobs_xml = document.xpath('//div/a[@class="job_link"]/@href')
-  job_links = jobs_xml.map(&:value)
-
-  [company_links, job_links]
-end
-
-def crawl_companies(company_links)
-  company_links.each do |link|
-    crawl_company(link)
+    jobs_xml.each { |i| job_links << i.value}
   end
+  job_links
 end
 
 def crawl_company(company_link)
   begin
-    document = Nokogiri::HTML(open(company_link))
+    uri = URI.parse(URI.escape(company_link)) #fix error: uri must be ascii only
+    document = Nokogiri::HTML(open(uri))
     company_name = document.css(".content .name").text
-    return if company_name.empty?
+    return nil if company_name.empty?
 
-    exist = Company.find_by(name: company_name).present?
-    return if exist
+    company = Company.find_by(name: company_name)
+    return company if company.present?
     
     puts company_name
     company_address = document.css(".content p")[1].text
@@ -46,6 +34,7 @@ def crawl_company(company_link)
     company = Company.create!(name: company_name, address: company_address, description: company_description)
   rescue => exception
     puts exception
+    return nil
   end
 end
 
@@ -57,16 +46,39 @@ end
 
 def crawl_job(job_link)
   begin
-    document = Nokogiri::HTML(open(job_link))
-    job_company = document.css('')
-    job_title = document.css('.job-desc p.title').text
-    job_salary = document.css('')
-    job_experience = document.css('')
-    job_level = document.css('')
-    job_expiration_date = document.css('')
-    job_description = document.css('')
+    uri = URI.parse(URI.escape(job_link)) #fix error: uri must be ascii only
+    document = Nokogiri::HTML(open(uri))
+    job_title = document.at_css('.job-desc p.title').text
+    return if job_title.empty?
+
+    job_company_link = document.at_css('.job-desc a.job-company-name')[:href]
+    job_company = crawl_company(job_company_link)
+    return if job_company.nil?
+    
+    job_detail = document.css('.job-detail-content div.detail-box')
+    job_location_name = job_detail[0].css('p a').map{ |val| val.text.strip }
+    job_locations = Location.where(city: job_location_name)
+    job_industry_names = job_detail[1].css('ul li')[1].css('p a').map{ |val| val.text.strip }
+    job_industries = Industry.where(name: job_industry_names)
+    job_salary = job_detail[2].css('ul li')[0].css('p').text.strip
+    job_experience = job_detail[2].css('ul li')[1].css('p').text.strip
+    job_level = job_detail[2].css('ul li')[2].css('p').text.strip
+    job_expiration_date = job_detail[2].css('ul li')[3].css('p').text.strip
+    job_description = document.css('.job-detail-content .detail-row').to_s
+
+    Job.find_or_create_by(title: job_title, company_id: job_company.id) do |job|
+      job.salary = job_salary
+      job.experience = job_experience
+      job.level = job_level
+      job.expiration_date = job_expiration_date
+      job.description = job_description
+      job.industries << job_industries
+      job.locations << job_locations
+    end
+    puts job_title
   rescue => exception
-    
+    puts exception
+    return exception
   end
 end
 

test/fixtures/industries_jobs.yml

-# Read about fixtures at http://api.rubyonrails.org/classes/ActiveRecord/FixtureSet.html
-
-one:
-  job: one
-  industry: one
-
-two:
-  job: two
-  industry: two

test/models/industries_job_test.rb

-require 'test_helper'
-
-class IndustriesJobTest < ActiveSupport::TestCase
-  # test "the truth" do
-  #   assert true
-  # end
-end