balinterdi-activewarehouse-etl 0.9.2

data/test/performance/delimited.ctl

@@ -0,0 +1,30 @@
+# puts "executing delimited.ctl"
+
+source :in, {
+  :file => 'delimited.txt',
+  :parser => :delimited
+}, 
+[ 
+  :first_name,
+  :last_name,
+  :ssn,
+  {
+    :name => :age,
+    :type => :integer
+  },
+  :sex
+]
+
+transform :ssn, :sha1
+transform(:ssn){ |v| v[0,24] }
+transform :sex, :decode, {:decode_table_path => 'delimited_decode.txt'}
+
+destination :out, {
+  :file => 'delimited.out.txt'
+},
+{
+  :order => [:first_name, :last_name, :name, :ssn, :age, :sex],
+  :virtual => {
+    :name => Proc.new { |row| "#{row[:first_name]} #{row[:last_name]}" }
+  }
+}

data/test/processor_test.rb

@@ -0,0 +1,38 @@
+require File.dirname(__FILE__) + '/test_helper'
+
+class Person < ActiveRecord::Base
+end
+
+# Test pre- and post-processors
+class ProcessorTest < Test::Unit::TestCase
+  # Test bulk import functionality
+  
+  context "the bulk import processor" do
+    should "should import successfully" do
+      assert_nothing_raised { do_bulk_import }
+      assert_equal 3, Person.count
+    end
+  end
+  
+  def test_bulk_import_with_empties
+    assert_raise(ActiveRecord::StatementInvalid) { do_bulk_import('bulk_import_with_empties.txt') }
+  end
+
+  def test_truncate
+    # TODO: implement test
+  end
+  
+  private
+  
+  def do_bulk_import(file = 'bulk_import.txt')
+    control = ETL::Control::Control.new(File.join(File.dirname(__FILE__), 'delimited.ctl'))
+    configuration = {
+      :file => "data/#{file}",
+      :truncate => true,
+      :target => :data_warehouse,
+      :table => 'people'
+    }
+    processor = ETL::Processor::BulkImportProcessor.new(control, configuration)
+    processor.process
+  end
+end

data/test/row_processor_test.rb

@@ -0,0 +1,17 @@
+require File.dirname(__FILE__) + '/test_helper'
+
+# Test row processors
+class RowProcessorTest < Test::Unit::TestCase
+  def test_copy_field_processor
+    
+  end
+  def test_hierarchy_exploder_processor
+    
+  end
+  def test_rename_processor
+    
+  end
+  def test_sequence_processor
+    
+  end
+end

data/test/sax.ctl

@@ -0,0 +1,26 @@
+# puts "executing fixed_width.ctl"
+
+source :in, {
+  :file => 'data/sax.xml',
+  :parser => :sax
+}, 
+{
+  :write_trigger => 'people/person',
+  :fields => {
+    :first_name => 'people/person/first_name',
+    :last_name => 'people/person/last_name',
+    :ssn => 'people/person/social_security_number',
+    :age => 'people/person[age]'
+  }
+}
+
+transform :ssn, :sha1
+transform(:ssn){ |v| v[0,24] }
+transform :age, :type, {:type => :number}
+
+destination :out, {
+  :file => 'output/sax.out.txt'
+}, 
+{
+  :order => [:first_name, :last_name, :ssn, :age]
+}

data/test/scd/1.txt

@@ -0,0 +1 @@
+Bob,Smith,200 South Drive,Boston,MA,32123

data/test/scd/2.txt

@@ -0,0 +1 @@
+Bob,Smith,1010 SW 23rd St,Los Angeles,CA,90392

data/test/scd/3.txt

@@ -0,0 +1 @@
+Bob,Smith,280 Pine Street,Los Angeles,CA,90392

data/test/scd_test.rb

@@ -0,0 +1,263 @@
+require File.dirname(__FILE__) + '/test_helper'
+
+class ScdTest < Test::Unit::TestCase
+  context "when working with a slowly changing dimension" do
+    setup do
+      @connection = ETL::Engine.connection(:data_warehouse)
+      @connection.delete("DELETE FROM person_dimension")
+      @end_of_time = DateTime.parse('9999-12-31 00:00:00')
+    end
+    context "of type 1" do
+      context "on run 1" do
+        setup do
+          do_type_1_run(1)
+        end
+        should "insert record" do
+          assert_equal 1, count_bobs
+        end
+        should "set the original address" do
+          assert_boston_address(find_bobs.first)
+        end
+        should "set the original id" do
+          assert_equal 1, find_bobs.first.id
+        end
+        should "skip the load if there is no change" do
+          do_type_1_run(1)
+          lines = lines_for('scd_test_type_1.txt')
+          assert lines.empty?, "scheduled load expected to be empty, was #{lines.size} records"
+        end
+      end
+      context "on run 2" do
+        setup do
+          do_type_1_run(1)
+          do_type_1_run(2)
+        end
+        should "delete the old record" do
+          assert_equal 1, count_bobs, "new record created, but old not deleted: #{find_bobs.inspect}"
+        end
+        should "update the address" do
+          assert_los_angeles_address(find_bobs.last)
+        end
+        should "keep id" do
+          assert_equal 1, find_bobs.first.id
+        end
+        should "only change once even if run again" do
+          do_type_1_run(2)
+          assert_equal 1, count_bobs
+          lines = lines_for('scd_test_type_1.txt')
+          assert lines.empty?, "scheduled load expected to be empty, was #{lines.size} records"
+        end
+        should "revert address on new record" do
+          do_type_1_run(1)
+          assert_boston_address(find_bobs.first)
+        end
+        should "keep record on revert" do
+          do_type_1_run(1)
+          assert_equal 1, count_bobs
+        end
+      end
+    end
+    context "of type 2" do
+      context "on run 1" do
+        setup do
+          do_type_2_run(1)
+        end
+        should "insert record" do
+          assert_equal 1, count_bobs
+        end
+        should "set the original record" do
+          assert_boston_address(find_bobs.first)
+        end
+        should "set the original id" do
+          assert_equal 1, find_bobs.first.id
+        end
+        should "set the effective date" do
+          # TODO: This is a test bug - if the tests don't run at the correct
+          # time, this timestamp may not match the timestamp used during the
+          # creation of the SCD row
+          assert_equal current_datetime, find_bobs.first.effective_date, 
+            "failure might be a test bug - see test notes"
+        end
+        should "set the end date" do
+          assert_equal @end_of_time, find_bobs.first.end_date
+        end
+        should "set the latest version flag" do
+          assert find_bobs.first.latest_version?
+        end
+        should "skip the load if there is no change" do
+          do_type_2_run(1)
+          lines = lines_for('scd_test_type_2.txt')
+          assert lines.empty?, "scheduled load expected to be empty, was #{lines.size} records"
+        end
+        
+      end
+      context "on run 2" do
+        setup do
+          do_type_2_run(1)
+          do_type_2_run(2)
+        end
+        should "insert new record" do
+          assert_equal 2, count_bobs
+        end
+        should "keep the primary key of the original version" do
+          assert_not_nil find_bobs.detect { |bob| 1 == bob.id }
+        end
+        should "increment the primary key for the new version" do
+          assert_not_nil find_bobs.detect { |bob| 2 == bob.id }
+        end
+        should "expire the old record" do
+          original_bob = find_bobs.detect { |bob| 1 == bob.id }
+          new_bob = find_bobs.detect { |bob| 2 == bob.id }
+          assert_equal new_bob.effective_date, original_bob.end_date
+        end
+        should "keep the address for the expired record" do
+          assert_boston_address(find_bobs.detect { |bob| 1 == bob.id })
+        end
+        should "update the address on the new record" do
+          assert_los_angeles_address(find_bobs.detect { |bob| 2 == bob.id })
+        end
+        should "activate the new record" do
+          # TODO: This is a test bug - if the tests don't run at the correct
+          # time, this timestamp may not match the timestamp used during the
+          # creation of the SCD row
+          assert_equal current_datetime, find_bobs.detect { |bob| 2 == bob.id }.effective_date, 
+            "failure might be a test bug - see test notes"
+        end
+        should "set the end date for the new record" do
+          assert_equal @end_of_time, find_bobs.detect { |bob| 2 == bob.id }.end_date
+        end
+        should "shift the latest version" do
+          original_bob = find_bobs.detect { |bob| 1 == bob.id }
+          new_bob = find_bobs.detect { |bob| 2 == bob.id }
+          assert !original_bob.latest_version?
+          assert new_bob.latest_version?
+        end
+        should "only execute a change once" do
+          do_type_2_run(2)
+          assert_equal 2, count_bobs
+          lines = lines_for('scd_test_type_2.txt')
+          assert lines.empty?, "scheduled load expected to be empty, was #{lines.size} records"
+        end
+        should "insert new records on revert" do
+          do_type_2_run(1)
+          assert_equal 3, count_bobs
+        end
+        should "update address on new record on revert" do
+          do_type_2_run(1)
+          assert_boston_address(find_bobs.detect { |bob| 3 == bob.id })
+        end
+        should "only delete one row on an scd change" do
+          # Two records right now
+          assert_equal 2, count_bobs
+          do_type_2_run(1) # put third version in (same as first version, but that's irrelevant)
+          # was failing because first and second versions were being deleted.
+          assert_equal 3, count_bobs
+        end
+      end
+      context "on non sdc fields that change" do
+        setup do
+          do_type_2_run_with_only_city_state_zip_scd(1)
+          do_type_2_run_with_only_city_state_zip_scd(2)
+        end
+        should "not create an extra record" do
+          do_type_2_run_with_only_city_state_zip_scd(3)
+          assert_equal 2, count_bobs
+        end
+        should "keep id" do
+          do_type_2_run_with_only_city_state_zip_scd(3)
+          assert_not_nil find_bobs.detect { |bob| 2 == bob.id }
+        end
+        should "keep dates" do
+          old_bob = find_bobs.detect { |bob| 2 == bob.id }
+          do_type_2_run_with_only_city_state_zip_scd(3)
+          new_bob = find_bobs.detect { |bob| 2 == bob.id }
+          assert_equal old_bob.end_date, new_bob.end_date
+          assert_equal old_bob.effective_date, new_bob.effective_date
+        end
+        should "keep the latest version flag" do
+          do_type_2_run_with_only_city_state_zip_scd(3)
+          assert find_bobs.detect { |bob| 2 == bob.id }.latest_version?
+        end
+        should "treat non scd fields like type 1 fields" do
+          do_type_2_run_with_only_city_state_zip_scd(3)
+          assert_los_angeles_address(find_bobs.detect { |bob| 2 == bob.id }, "280 Pine Street")
+        end
+        should "skip load when there is no change" do
+          do_type_2_run_with_only_city_state_zip_scd(2)
+          lines = lines_for('scd_test_type_2.txt')
+          assert lines.empty?, "scheduled load expected to be empty, was #{lines.size} records"
+        end
+      end
+    end
+  end
+  
+  def do_type_2_run(run_num)
+    ENV['run_number'] = run_num.to_s
+    assert_nothing_raised do
+      run_ctl_file("scd_test_type_2.ctl")
+    end
+  end
+  
+  def do_type_2_run_with_only_city_state_zip_scd(run_num)
+    ENV['type_2_scd_fields'] = Marshal.dump([:city, :state, :zip_code])
+    do_type_2_run(run_num)
+  end
+  
+  def do_type_1_run(run_num)
+    ENV['run_number'] = run_num.to_s
+    assert_nothing_raised do
+      run_ctl_file("scd_test_type_1.ctl")
+    end
+  end
+  
+  def lines_for(file)
+    File.readlines(File.dirname(__FILE__) + "/output/#{file}")
+  end
+  
+  def run_ctl_file(file)
+    ETL::Engine.process(File.dirname(__FILE__) + "/#{file}")
+  end
+  
+  def count_bobs
+    @connection.select_value(
+      "SELECT count(*) FROM person_dimension WHERE first_name = 'Bob' and last_name = 'Smith'").to_i
+  end
+  
+  def find_bobs
+    bobs = @connection.select_all(
+      "SELECT * FROM person_dimension WHERE first_name = 'Bob' and last_name = 'Smith'")
+    bobs.each do |bob|
+      def bob.id
+        self["id"].to_i
+      end
+      def bob.effective_date
+        DateTime.parse(self["effective_date"])
+      end
+      def bob.end_date
+        DateTime.parse(self["end_date"])
+      end
+      def bob.latest_version?
+        ActiveRecord::ConnectionAdapters::Column.value_to_boolean(self["latest_version"])
+      end
+    end
+    bobs
+  end
+  
+  def current_datetime
+    DateTime.parse(Time.now.to_s(:db))
+  end
+  
+  def assert_boston_address(bob, street = "200 South Drive")
+    assert_equal street, bob['address'], bob.inspect
+    assert_equal "Boston", bob['city'], bob.inspect
+    assert_equal "MA", bob['state'], bob.inspect
+    assert_equal "32123", bob['zip_code'], bob.inspect
+  end
+  
+  def assert_los_angeles_address(bob, street = "1010 SW 23rd St")
+    assert_equal street, bob['address'], bob.inspect
+    assert_equal "Los Angeles", bob['city'], bob.inspect
+    assert_equal "CA", bob['state'], bob.inspect
+    assert_equal "90392", bob['zip_code'], bob.inspect
+  end
+end

data/test/scd_test_type_1.ctl

@@ -0,0 +1,43 @@
+source :in, {
+  :file => "scd/#{ENV['run_number']}.txt",
+  :parser => :delimited
+},
+[
+  :first_name,
+  :last_name,
+  :address,
+  :city,
+  :state,
+  :zip_code
+]
+
+# NOTE: These are not usually required for a type 1 SCD dimension, but since
+# we're sharing this table with the type 2 tests, they're necessary.
+transform :effective_date, :default, :default_value => Time.now.to_s(:db)
+transform :end_date, :default, :default_value => '9999-12-31 00:00:00'
+transform :latest_version, :default, :default_value => true
+
+destination :out, {
+  :file => 'output/scd_test_type_1.txt',
+  :natural_key => [:first_name, :last_name],
+  :scd => {
+    :type => 1,
+    :dimension_target => :data_warehouse,
+    :dimension_table => 'person_dimension'
+  },
+  :scd_fields => [:address, :city, :state, :zip_code]
+}, 
+{
+  :order => [
+    :id, :first_name, :last_name, :address, :city, :state, :zip_code, :effective_date, :end_date, :latest_version
+  ],
+  :virtual => {
+    :id => ETL::Generator::SurrogateKeyGenerator.new(:target => :data_warehouse, :table => 'person_dimension')
+  }
+}
+
+post_process :bulk_import, {
+  :file => 'output/scd_test_type_1.txt',
+  :target => :data_warehouse,
+  :table => 'person_dimension'
+}

data/test/scd_test_type_2.ctl

@@ -0,0 +1,42 @@
+source :in, {
+  :file => "scd/#{ENV['run_number']}.txt",
+  :parser => :delimited
+},
+[
+  :first_name,
+  :last_name,
+  :address,
+  :city,
+  :state,
+  :zip_code
+]
+
+
+
+
+
+
+destination :out, {
+  :file => 'output/scd_test_type_2.txt',
+  :natural_key => [:first_name, :last_name],
+  :scd => {
+    :type => 2,
+    :dimension_target => :data_warehouse,
+    :dimension_table => 'person_dimension'
+  },
+  :scd_fields => ENV['type_2_scd_fields'] ? Marshal.load(ENV['type_2_scd_fields']) : [:address, :city, :state, :zip_code]
+}, 
+{
+  :order => [
+    :id, :first_name, :last_name, :address, :city, :state, :zip_code, :effective_date, :end_date, :latest_version
+  ],
+  :virtual => {
+    :id => ETL::Generator::SurrogateKeyGenerator.new(:target => :data_warehouse, :table => 'person_dimension')
+  }
+}
+
+post_process :bulk_import, {
+  :file => 'output/scd_test_type_2.txt',
+  :target => :data_warehouse,
+  :table => 'person_dimension'
+}