Scraping and analyzing recent articles on bitcoin

Question

This script scrapes recent articles on bitcoin, does sentiment analysis, and does some mock trading based on the sentiment of the articles. I'm looking for advice on code style, and I would love to learn how to write beautiful Ruby.

require 'coinbase'
require 'sanitize'
require 'cgi'
require 'htmlentities'
require 'sentimental'
require 'simple-rss'
require 'open-uri'
require 'sqlite3'
require 'date'

@db = SQLite3::Database.open('instructions.db')
@db.execute("CREATE TABLE IF NOT EXISTS instructions2(Id INTEGER PRIMARY KEY AUTOINCREMENT, Date TEXT, Do_now TEXT, Do_later TEXT, Buy_price TEXT, Sell_price TEXT)")
@rss = SimpleRSS.parse(open('http://fulltextrssfeed.com/news.google.com/news?pz=1&cf=all&ned=us&hl=en&q=Bitcoin&output=rss'))
@coinbase = Coinbase::Client.new(ENV['COINBASEKEY'])
@coder = HTMLEntities.new
Sentimental.load_defaults
@analyzer = Sentimental.new
@ran_today = false
@scores_today = []
@log = File.open('log.txt', 'w')

def write_log(me)
  @log << DateTime.now.to_s + " : " + me.to_s
end

def check_sentiment
  @ran_today = true
  @rss.items.each do |item|
    @scores_today.push(@analyzer.get_score(clean(item.description)))
  end
  react_sentiment
end

def react_sentiment
  avg = @scores_today.reduce(:+).to_f / @scores_today.size
  do_now = :nothing
  do_later = :nothing
  #if negative, sell now, buy when low
  if avg > 3.0
    do_now = :buy
    do_later = :sell
  elsif avg < -2.0
    do_now = :sell
    do_later = :buy
  end
  write_log("day's sentiment score : " + avg.to_s)
  method(do_now).call
  write_log(@db.execute("INSERT INTO instructions2(Date, Do_now, Do_later, Buy_price, Sell_price) VALUES (?, ?, ?, ?, ?)", (Date.today+2).to_s, do_now.to_s, do_later.to_s, check_buy.to_s, check_sell.to_s))
end

def read_instructions
  @db.execute("SELECT Id, Date, Do_now, Do_later, Buy_price, Sell_price FROM instructions2;").each do |id, date, do_now, do_later, buy_price, sell_price|
    if Date.today.to_s == date
      if do_now == "buy"
        if check_buy < sell_price.to_f
          buy
        else
          puts "want to buy, buy " + buy_price + "  > sell " + sell_price
        end
      elsif do_now == "sell" 
        if check_sell > buy_price.to_f
          sell
        else
         puts "want to sell, sell " + sell_price + " > buy " + buy_price
        end
      end
      @db.execute("DELETE * FROM instructions2 WHERE Date=?", Date.today.to_s)
    end
  end
end


def buy
  write_log("would buy at " + check_buy.to_s + " at " + Date.today.to_s)
end

def sell
  write_log("would sell at " + check_sell.to_s + " at " + Date.today.to_s)
end

def nothing
  write_log("would do nothing at " + Date.today.to_s)
end

def check_sell
  @coinbase.sell_price(1)
end

def check_buy
  @coinbase.buy_price(1)
end

def clean(stuff)
  Sanitize.clean(CGI.unescapeHTML(stuff).to_s.force_encoding('UTF-8'))
end

check_sentiment
read_instructions
@ran_today = true
while true
  if Time.now.hour == 23 and not @ran_today
    check_sentiment
  elsif Time.now.hour == 1
    read_instructions
  elsif Time.now.hour == 24
    @ran_today = false
    @scores_today = []
  end
  sleep(60000)
end

Answer 1

The @ variable prefix should only be used inside a class to denote instance variables. That said, for a script like this, you might like to wrap all the logic in a class (or rather a module since you don't need multiple instances).

Use string interpolation. So instead of

DateTime.now.to_s + " : " + me.to_s

you can write

"#{DateTime.now} : #{me}"

Answer 2

A few observations:

1. It seems that you fill @rss only when the program starts, and never again, which means that each time you run check_sentiment you read the same (stale) data.
2. When you react_sentiment you call method(do_now).call, and when you read_instructions you also call do_now. You never use do_later though...
3. In your main loop you sleep(60000), between iterations, which means you sleep 60,000 seconds - about 17 hours. This means you will miss some of the check points you intended to run. Generally speaking, maybe running this indefinitely for a couple of check points a day is not recommended, and better way would be to use cron jobs to run at 0100 and at 2300.

4. Encapsulation - you seem to have a lot of instance members in your code, some of which need to be maintained manually (like @scores_today), some maintenance is neglected (like @rss), and some are forgotten (like @coder, which is never used). To prevent this, you need to think which members are actually part of your state, and which are not. If there is no good reason to keep them - don't:

   def check_sentiment
     @ran_today = true
     rss = SimpleRSS.parse(open('http://fulltextrssfeed.com/news.google.com/news?pz=1&cf=all&ned=us&hl=en&q=Bitcoin&output=rss'))
     analyzer = Sentimental.new
     scores_today = rss.items.map do |item|
       analyzer.get_score(clean(item.description)))
     end
     react_sentiment(scores_today)
   end
   
   def react_sentiment
     avg = scores_today.reduce(:+).to_f / scores_today.size
   
     # ... etc
   end