tabula-extractor 0.6.6-java → 0.7.0-java

data/lib/tabula/whitespace.rb

@@ -1,50 +0,0 @@
-require 'algorithms'
-module Tabula
-  module Whitespace
-
-    # Detect whitespace in a document (not yet used in Tabula)
-    # Described in "Two Geometric Algorithms for layout analysis" (Thomas Breuer)
-    # http://pdf.aminer.org/000/140/219/two_geometric_algorithms_for_layout_analysis.pdf
-
-    def self.find_closest(text_elements, x, y)
-      text_elements.sort_by { |te|
-        Math.sqrt((x - te.midpoint[0]) ** 2 + (y - te.midpoint[1]) ** 2)
-      }.first
-    end
-
-
-    def self.find_whitespace(text_elements, bounds)
-      queue = Containers::PriorityQueue.new
-      queue.push([bounds, text_elements], bounds.width * bounds.height)
-      rv = []
-
-
-      while !queue.empty?
-        r, obstacles = queue.pop
-        if obstacles.empty?
-          return r
-        end
-
-        pivot = self.find_closest(obstacles, *r.midpoint)
-
-        subrectangles = [
-                         ZoneEntity.new(r.top, pivot.right, r.right - pivot.right, pivot.top - r.top),
-                         ZoneEntity.new(r.top, r.left, pivot.left - r.left, pivot.top - r.top),
-                         ZoneEntity.new(pivot.bottom, r.left, pivot.left - r.left, r.bottom - pivot.bottom),
-                         ZoneEntity.new(pivot.bottom, pivot.right, r.right - pivot.right, r.bottom - pivot.bottom)
-                        ]
-        subrectangles.each do |sub_r|
-          obs = obstacles.select { |s|
-            s.overlaps?(sub_r)
-          }
-          if obs.empty?
-            rv << sub_r
-          else
-            queue.push([sub_r, obs], sub_r.width * sub_r.height)
-          end
-        end
-      end
-      return rv
-    end
-  end
-end

data/vertical_rulings_bug.rb

@@ -1,29 +0,0 @@
-require './lib/tabula'
-
-input_filename = "vertical_rulings_bug.pdf"
-out = File.new("output.xls", 'w')
-
-extractor = Tabula::Extraction::CharacterExtractor.new(input_filename, :all) #:all ) # 1..2643
-extractor.extract.each_with_index do |pdf_page, page_index|
-
-  lines = Tabula::Ruling::clean_rulings(Tabula::LSD::detect_lines_in_pdf_page(input_filename, page_index))
-  page_areas = [[0, 0, 1000, 1700]]
-
-  scale_factor = pdf_page.width / 1700
-  puts scale_factor
-
-  vertical_rulings = [0, 360, 506, 617, 906, 1034, 1160, 1290, 1418, 1548].map{|n| Geometry::Segment.new_by_arrays([n * scale_factor, 0], [n * scale_factor, 1000])}
-
-  page_areas.each do |page_area|
-    text = pdf_page.get_text( page_area ) #all the characters within the given area.
-
-    Tabula::Writers.send(:TSV,
-                         Tabula.make_table_with_vertical_rulings(text, {:vertical_rulings => vertical_rulings, :merge_words => true, :dontmerge => true}), 
-                         out)
-  end
-end
-out.close
-
-
-#with dontmerge false (i.e. if we merge) we get crap. STCITY and no spaces in any cities.
-#with dontmerge true (or commented out), MORGANTOWNWV, and some spaces (e.g. BRYN MAWR, but not FRESHMEADOWS)