My solution to extract email and link from pdf data

link_extractor.rb

# Public: Class that parses the word extraction output from the `pdftoolkit-extract`
#         tool and returns web links and emails to be added to a publication at a
#         later stage.
#
#         The format in which the tool outputs is as follows:
#
#         <WORD> <X1> <Y1> <X2> <Y2>
#
#         An example of what the original PDF looks like can be find here:
#         
#         http://i.imgur.com/LqFxvzE.png
class LinkExtractor
  # Public: Initializes a link extractor with the raw output from the extraction
  #         tool.
  def initialize(words)
    @words = words
  end

  # Public: Runs link detection and yields each result as links are picked up.
  def each_link
    detect_links.each do |link|
      yield link
    end
  end

private  
  # Private: The detection algorithm.
  def detect_links
    # TODO: Implement method
    terms = get_term(@words)
    links = []
    terms.each do |term|
      parts = term.split("\n")
      readable = []
      x1s = []
      x2s = []
      y1s = []
      y2s = []

      #TODO: thinking of solution which use single Block, trigger "split" once, 
      # and assigned all: readable, x1s, x2s, y1s, y2s
      parts.select { |x| readable << x.split(" ")[0] } 
      parts.select { |x| x1s << x.split(" ")[1] } 
      parts.select { |x| y1s << x.split(" ")[2] } 
      parts.select { |x| x2s << x.split(" ")[3] } 
      parts.select { |x| y2s << x.split(" ")[4] } 

      #TODO: create a helper action to validate valid email
      # create a helper action to validate valid link
      readable.each_with_index do |text,i|
        if text == "@" && (i+3) < readable.size
          links << {
            :type=>:email, 
            :content=> "#{readable[i-1]}#{readable[i]}#{readable[i+1]}#{readable[i+2]}#{readable[i+3]}", 
            :rect=>
            {
              :x1=> x1s[i-1].to_f, 
              :x2=> x2s[i+3].to_f, 
              :y1=> y1s[i-1].to_f, 
              :y2=> y2s[i+3].to_f
            }
          }
        elsif (text == "www" && readable[i+1]=="." && (i+4 < readable.size) )
          links << {
            :type=>:link, 
            :content=>"#{text}#{readable[i+1]}#{readable[i+2]}#{readable[i+3]}#{readable[i+4]}", 
            :rect=>
            {
              :x1=>x1s[i].to_f, 
              :x2=>x2s[i+4].to_f, 
              :y1=>y1s[i].to_f, 
              :y2=>y2s[i+4].to_f
            }
          }
        end
      end
    end
    links
  end

  def get_term(words)
    words.split("\n\n")
  end

end

require 'minitest/autorun'

class LinkExtractorTest < Minitest::Unit::TestCase
  def test_emails
    fixture = <<-STR
CALL 250.934100 182.967400 282.310500 173.787400
: 283.425300 182.967400 285.117300 173.787400
01327 289.278900 182.967400 323.618100 173.787400
305232 326.447700 182.967400 372.389700 173.787400

West 138.282900 165.399400 168.477300 153.951400
- 171.998100 165.399400 176.162100 153.951400
wdistric 179.273700 165.399400 226.673700 153.951400
@ 228.076500 165.399400 239.068500 153.951400
ford 240.339300 165.399400 264.020100 153.951400
. 265.349700 165.399400 267.041700 153.951400
com 268.059300 165.399400 293.304900 153.951400
Trust 320.753700 165.399400 351.342900 153.951400
Ford 355.294500 165.399400 381.124500 153.951400
/ 381.951300 165.399400 388.359300 153.951400
Retail 389.330100 165.399400 423.173700 153.951400
- 426.693300 165.399400 430.857300 153.951400
retail12 434.664900 165.399400 476.472900 153.951400
@ 478.115700 165.399400 489.107700 153.951400
ford 490.378500 165.399400 514.060500 153.951400
. 515.390100 165.399400 517.082100 153.951400
com 518.100900 165.399400 543.347700 153.951400

North 94.534500 147.399400 127.539300 135.951400
/ 128.326500 147.399400 134.734500 135.951400
SNIC 134.877300 147.399400 164.742900 135.951400
- 168.198900 147.399400 172.362900 135.951400
nsnic123 176.166900 147.399400 226.365300 135.951400
@ 228.016500 147.399400 239.008500 135.951400
ford 240.275700 147.399400 263.957700 135.951400
. 265.287300 147.399400 266.979300 135.951400
com 267.996900 147.399400 293.238900 135.951400
East 321.407700 147.399400 346.836900 135.951400
- 350.357700 147.399400 354.521700 135.951400
edistric 358.001700 147.399400 401.745300 135.951400

    STR
    
    assert_equal links_for(fixture), [
      {:type=>:email, :content=>"[email protected]", :rect=>{:x1=>179.2737, :x2=>293.3049, :y1=>165.3994, :y2=>153.9514}},
      {:type=>:email, :content=>"[email protected]", :rect=>{:x1=>434.6649, :x2=>543.3477, :y1=>165.3994, :y2=>153.9514}},
      {:type=>:email, :content=>"[email protected]", :rect=>{:x1=>176.1669, :x2=>293.2389, :y1=>147.3994, :y2=>135.9514}}
    ]
  end
  
  def test_urls
    fixture = <<-STR
0181 339.203831 221.163031 356.562792 211.117225
- 358.920122 221.163031 361.306201 211.117225
458845 361.808957 221.163031 390.528143 211.117225
| 394.763276 221.163031 395.515513 211.117225
www 399.430156 221.163031 422.122463 211.117225
. 421.572410 221.163031 423.575183 211.117225
pols 424.251228 221.163031 443.201512 211.117225
. 443.651914 221.163031 445.654686 211.117225
nl 446.661196 221.163031 455.280747 211.117225

5 306.554777 188.411494 310.363879 178.371012
. 311.461091 188.411494 312.630174 178.371012
03 313.425103 188.411494 322.188432 178.371012
CASE 339.382200 188.411494 361.757467 178.371012
IH 365.514557 188.411494 373.635849 178.371012
& 377.382365 188.411494 383.184656 178.371012
Steyr 386.360906 188.411494 406.465606 178.371012

De 339.650514 177.497927 350.637016 168.676570
Kolk 354.436100 177.497927 372.006923 168.676570
2 374.862116 177.497927 379.097645 168.676570
, 379.806761 177.497927 381.062087 168.676570
8255 383.645969 177.497927 402.346499 168.676570
PE 406.366023 177.497927 415.307022 168.676570
SWIFTERBANT 418.718047 177.497927 482.538383 168.676570

0321 339.203831 166.605654 356.562792 156.549201
- 358.920122 166.605654 361.306201 156.549201
335515 362.326420 166.605654 390.528143 156.549201
| 394.763276 166.605654 395.515513 156.549201
www 399.430156 166.605654 422.122463 156.549201
. 421.572410 166.605654 423.575183 156.549201
caseih 424.653699 166.605654 453.066849 156.549201
. 453.751970 166.605654 455.754743 156.549201
com 456.833259 166.605654 475.524594 156.549201
    STR
    
    assert_equal links_for(fixture), [
      {:type=>:link, :content=>"www.pols.nl", :rect=>{:x1=>399.430156, :x2=>455.280747, :y1=>221.163031, :y2=>211.117225}},
      {:type=>:link, :content=>"www.caseih.com", :rect=>{:x1=>399.430156, :x2=>475.524594, :y1=>166.605654, :y2=>156.549201}}
    ]
  end
  
  def test_false_positives
    fixture = <<-STR
kand 88.149000 210.623500 100.209000 205.103500
. 101.151000 210.623500 101.655000 205.103500
rechten 102.597000 210.623500 121.227000 205.103500
UFSIA 123.639000 210.623500 139.515000 205.103500
91 141.345000 210.623500 146.547000 205.103500
, 148.371000 210.623500 148.899000 205.103500
bijz 151.509000 210.623500 159.489000 205.103500
. 160.173000 210.623500 160.677000 205.103500
lic 161.625000 210.623500 166.425000 205.103500
. 167.175000 210.623500 167.679000 205.103500
int 168.627000 210.623500 174.183000 205.103500
. 174.951000 210.623500 175.455000 205.103500
pol 176.397000 210.623500 183.561000 205.103500
UIA 185.985000 210.623500 194.859000 205.103500
93 196.689000 210.623500 202.899000 205.103500

Plataandreef 88.221000 203.621500 119.775000 198.425500
9 121.509000 203.621500 124.401000 198.425500
, 125.169000 203.621500 125.697000 198.425500
2900 128.103000 203.621500 140.937000 198.425500
SCHOTEN 142.917000 203.621500 169.935000 198.425500
, 170.889000 203.621500 171.417000 198.425500

✆ 87.999000 196.581437 92.319000 191.423500
03 94.335000 196.581437 100.515000 191.423500
- 101.103000 196.581437 102.603000 191.423500
658 103.173000 196.581437 112.659000 191.423500
38 114.627000 196.581437 120.909000 191.423500
39 122.877000 196.581437 129.141000 191.423500
, 129.909000 196.581437 130.437000 191.423500
7 133.133836 196.581437 137.469773 191.423500
03 139.791000 196.581437 145.971000 191.423500
- 146.559000 196.581437 148.059000 191.423500
658 148.629000 196.581437 158.115000 191.423500
30 160.083000 196.581437 166.383000 191.423500
92 168.399000 196.581437 174.483000 191.423500

ADRIAENSEN 247.988800 465.641300 286.946800 460.259300
, 287.726800 465.641300 288.668800 460.259300
OSWALD 290.924800 465.641300 316.688200 460.259300
, 317.486200 465.641300 318.014200 460.259300
b 320.624200 465.641300 323.480200 460.259300
. 324.290200 465.641300 324.794200 460.259300
w 325.424200 465.641300 329.774200 460.259300
. 330.074200 465.641300 330.578200 460.259300
ir 331.526200 465.641300 334.160200 460.259300
. 334.190200 465.641300 334.694200 460.259300
UGent 337.334200 465.641300 353.426200 460.259300
53 355.496200 465.641300 361.694200 460.259300    
    STR
    
    assert_empty links_for(fixture)
  end
  
  def links_for(fixture)
    ([]).tap do |links|
      LinkExtractor.new(fixture).each_link { |l| links << l }
    end
  end
end