Boolean method not returning in different situations [RUBY] - ruby

Question

I'm building a simple web-scraper (scraping jobs from indeed.com) for practice and I'm trying to implement the following method (low_salary?(salary)). The aim is for the method to compare a minimum (i.e. desired) salary, compare it with the offered salary contained in the job object (#salary):
class Job
attr_reader :title, :company, :location, :salary, :url
def initialize(title, company, location, salary, url)
#title = title
#company = company
#location = location
#salary = salary
#url = url
end
def low_salary?(minimum_salary)
return if !#salary
minimum_salary < #salary.split(/[^\d]/)[1..2].join.to_i
end
end
The method works fine when comapring #salary and the min_salary variable given to it, the delete_if appropriately deletes the elements that return true for low_salary? and returns correctly when #salary is nil (indeed listings don't always include the salary so my assumption is that there will be some nil values) in the following test program (Also: I am unsure as to why minimum_salary < #salary works but #salary < minimum_salary doesn't, but this does exactly what I want it to do):
require_relative('job_class.rb')
job = Job.new("designer", "company", "location", "£23,000 a year", "url")
job_results = []
job_results.push(job)
min_salary = 50000
print job.low_salary?(min_salary)
job_results.delete_if { |job| job.low_salary?(min_salary) }
print job_results
However in my scraper program, I get a no method error when calling the method: job_class.rb:16:in "low_salary?": undefined method `join' for nil:NilClass (NoMethodError)
require 'nokogiri'
require 'open-uri'
require_relative 'job_class.rb'
class JobSearchTool
def initialize(job_title, location, salary)
#document = Nokogiri::HTML(open("https://uk.indeed.com/jobs?q=#{job_title.gsub('-', '+')}&l=#{location}"))
#job_listings = #document.css('div.mosaic-provider-jobcards > a')
#salary = salary.to_i
#job_results = []
end
def scrape_jobs
#job_listings.each do |job_card|
#job_results.push(Job.new(
job_card.css('h2 > span').text, #title
job_card.css('span.companyName').text, #company
job_card.css('div.companyLocation').text, #location
job_card.css('span.salary-snippet').text, #salary
job_card['href']) #url
)
end
end
def format_jobs
#job_results.each do |job|
puts <<~JOB
#{job.title} - #{job.company} in #{job.location} :#{job.salary}
Apply at: #{job.url}
---------------------------------------------------------------------------------
JOB
end
end
def check_salary
#job_results.delete_if { |job| job.low_salary?(#salary) }
end
def run
scrape_jobs
check_salary
format_jobs
end
if __FILE__ == $0
job_search_tool = JobSearchTool.new(ARGV[0], ARGV[1], ARGV[2])
job_search_tool.run
end
Obviously something from the scraper programme is influencing the method somehow, but I can't understand what it could be. I'm using the method in the exact same way as the test program, so what difference is causing the method not to return when #salary is nil?

Answer 1

A quick search on the URL you're scraping shows there are job posts that don't have a salary, so, when you get the data from that HTML element and initialize a new Job object, the salary is an empty string, and knowing that "".split(/[^\d]/)[1..2] returns nil, that's the error you get.
You must add a way to handle job posts without a salary:
class Job
attr_reader :title, :company, :location, :salary, :url
def initialize(title, company, location, salary, url)
#title = title
#company = company
#location = location
#salary = salary.to_s # Explicit conversion of nil to string
#url = url
end
def low_salary?(minimum_salary)
return if parsed_salary.zero? # parsed_salary returns always an integer,
# so you can check when is zero,
# and not just when is falsy
minimum_salary < parsed_salary
end
private
def parsed_salary
salary[/(?<=£)(\d|,)*(?=\s)/]
.to_s # converts nil to "" if the regex doesn't capture anything
.tr(",", "") # removes the commas to parse the string as an integer
.to_i # parses the string to its corresponding integer representation
end
end
Notice the regex isn't meant to capture everything, but it works with the salary as rendered in the website.