require File.expand_path(File.dirname(__FILE__) + '/../../spec_helper')

include Fathom

describe DataNode do

  before do
    @values = [1,2,3,4,5]
    @opts = {:values => @values}
    @dn = DataNode.new(@opts)
    @vector = GSL::Vector.ary_to_gv(@values)
  end
  
  it "should initialize requiring values in the options" do
    lambda{DataNode.new}.should raise_error(/values/)
    lambda{DataNode.new(:values => @values)}.should_not raise_error
  end
  
  it "should make the values readable" do
    @dn.values.should eql(@values)
  end
  
  it "should allow an optional name for the node" do
    @dn = DataNode.new(:values => @values, :name => "Demo Name")
    @dn.name.should eql("Demo Name")
  end
  
  it "should take an optional distribiution" do
    @dn = DataNode.new(@opts.merge(:distribution => :gaussian))
    @dn.distribution.should eql(Fathom::Distributions::Gaussian)
  end

  it "should create a vector from the values" do
    @dn.vector.should ==(@vector)
  end
  
  it "should provide the standard deviation" do
    @dn.standard_deviation.should ==(@vector.sd)
  end
  
  it "should alias sd and std for standard_deviation" do
    @dn.sd.should eql(@dn.standard_deviation)
    @dn.std.should eql(@dn.standard_deviation)
  end
  
  it "should be able to produce the mean" do
    @dn.mean.should eql(@vector.mean)
  end

  it "should generate a random variable that fits the data's distribution" do
    @dn.rand.should be_a(Float)
  end
  
  it "should have a name_sym method" do
    dn = DataNode.new(:name => "Demo Node", :values => [1,2,3])
    dn.name_sym.should eql(:demo_node)
  end

  it "should offer the lower bounds at a default confidence level of 0.05" do
    GSL::Cdf.should_receive(:gaussian_Pinv).with(0.05, @dn.vector.sd).and_return(0.0)
    @dn.inverse_cdf
  end
  
  it "should offer the lower bounds at an arbitrary confidence level" do
    GSL::Cdf.should_receive(:gaussian_Pinv).with(0.95, @dn.vector.sd).and_return(0.0)
    @dn.inverse_cdf(:confidence_interval => 0.95)
  end
  
  it "should be able to calculate the upper bound by passing upper as a parameter" do
    GSL::Cdf.should_receive(:gaussian_Qinv).with(0.05, @dn.vector.sd).and_return(0.0)
    @dn.inverse_cdf(:confidence_interval => 0.05, :upper => true)
  end
  
  it "should be able to calculate a lower_bound, an alias for inverse_cdf" do
    GSL::Cdf.should_receive(:gaussian_Pinv).with(0.05, @dn.vector.sd).and_return(0.0)
    @dn.lower_bound
  end
  
  it "should be able to pass arguments to lower_bound" do
    GSL::Cdf.should_receive(:gaussian_Pinv).with(0.1, @dn.vector.sd).and_return(0.0)
    @dn.lower_bound(:confidence_interval => 0.1)
  end
  
  it "should be able to calculate an upper_bound, a shortcut for inverse_cdf(confidence_interval, false)" do
    GSL::Cdf.should_receive(:gaussian_Qinv).with(0.05, @dn.vector.sd).and_return(0.0)
    @dn.upper_bound
  end
  
  it "should be able to pass arguments to upper_bound" do
    GSL::Cdf.should_receive(:gaussian_Qinv).with(0.1, @dn.vector.sd).and_return(0.0)
    @dn.upper_bound(:confidence_interval => 0.1)
  end
  
  it "should offset the cdf results by the mean" do
    @dn.lower_bound.should eql(GSL::Cdf.gaussian_Pinv(0.05, @dn.vector.sd) + @dn.vector.mean)
  end
  
  it "should return the lower and upper bounds of a confidence interval" do
    @dn.interval_values.should eql([@dn.lower_bound, @dn.upper_bound])
  end
  
end