Data Modeling using Pacer

Brought to you by

What Is Data Modeling?

Mike buys a book online

Data Modeling is about going back and forth between the real-world (on the left) and the graph (on the right).

VS.

Pacer Extensions

A fundamental building block of Pacer

Without extensions

def comments_to_posts_by(person)
   person
      .out_e(:POSTED)
      .in_v(type: 'post')
      .in_e(:IS_ABOUT)
      .in_v(type: 'comment')
end
   

With extensions

def comments_to_posts_by(person)
   person.posts.comments
end
   

Pacer Extensions

What do they look like?

module Person

   # ...

   module Vertex
 
      def posts
         self.out_e(:POSTED).in_v(type: 'post')
      end

   end

   # ...

end

Case Study - NorthWind

• Dataset of a fictional company, NorthWind Traders.
• Created by Microsoft
• Imported into a property graph by Neo4j

Setup

$ cd pacer-northwind
$ irb

jruby-1.7.18 :001 > require 'pacer'
 => true

jruby-1.7.18 :002 > g = Pacer.tg
 => #<PacerGraph tinkergraph[vertices:0 edges:0]

jruby-1.7.18 :003 > Pacer::GraphML.import(g, 'northwind.graphml')
 => true

From the root folder of our repo

And then

NorthWind

# Find out the different types of vertices ...

jruby-1.7.18 :004 > g.v[:type].uniq
"Order"  "Category" "Employee" 
"Customer" "Product"  "Supplier"
Total: 6

Let's explore the dataset ...

First Extension - Customer

In a file called customer.rb

module NorthWind
    module Customer

        

            
                
            

   
    end	
end

• Extensions are plain old Ruby modules
• Create a Vertex module
• Define vertex extension methods
• Implement them (self is a vertex object)

Using The Extension

Try it in the IRB ...

jruby-1.7.18 :100 > load 'customer.rb'
 => true

jruby-1.7.18 :101 > c = g.v(NorthWind::Customer, 
                            type: 'Customer').first
 => #<V[n10]>

jruby-1.7.18 :102 > c.orders
#<V[n255]> #<V[n913]> #<V[n450]> #<V[n505]> #<V[n989]> 
#<V[n504]> #<V[n909]> #<V[n437]> #<V[n544]> #<V[n565]>
Total: 10

What just happened?

• All vertices
• Whose type is 'Customer'
• Extended by the Customer extension

Filter By Extension

Only vertices whose type is 'Customer' should be extended by the Customer extension.

Filter By Extension

jruby-1.7.18 :103 > load 'customer.rb'
 => true

jruby-1.7.18 :104 > include NorthWind
 => Object

jruby-1.7.18 :105 > g.v(Customer).first.orders
#<V[n255]> #<V[n913]> #<V[n450]> #<V[n505]> #<V[n989]> 
#<V[n504]> #<V[n909]> #<V[n437]> #<V[n544]> #<V[n565]>
Total: 10

Filter By Extension

Try to get the orders of multiple customers ...

jruby-1.7.18 :106 > g.v(Customer).orders
NoMethodError
...
Total: 91
undefined method `orders' for 
#<GraphV -> V-Property(NorthWind::Customer)>

Vertex vs. Route

module NorthWind
    module Customer

        def self.route_conditions(graph)
            {type: "Customer"}
        end

        module Vertex  # Contains vertex extension methods
            # ... 
        end

        module Route   # Contains route extension methods

            def orders
                self.out_e(:PURCHASED).in_v(type: "Order")
            end

        end

    end
end

jruby-1.7.18 :107 > load 'customer.rb'
 => true

jruby-1.7.18 :108 > g.v(Customer).limit(3).orders
#<V[n255]>  #<V[n913]>  #<V[n450]>  #<V[n505]>  #<V[n989]>  
#<V[n504]>  #<V[n909]>  #<V[n437]>  #<V[n544]>  #<V[n565]>  
#<V[n225]>  #<V[n487]>  #<V[n903]>  #<V[n1020]> #<V[n748]>  
#<V[n847]>  #<V[n785]>
Total: 17
 => #<GraphV -> V-Property(NorthWind::Customer) -> V-Range(-1...3) 
    -> outE(:PURCHASED) -> inV -> V-Property(type=="Order")>

Vertex vs. Route

Reload and try it in the IRB

jruby-1.7.18 :109 > g.v(Customer).first.orders
#<V[n255]> #<V[n913]> #<V[n450]> #<V[n505]> #<V[n989]> 
#<V[n504]> #<V[n909]> #<V[n437]> #<V[n544]> #<V[n565]>
Total: 10
 => #<outE(:PURCHASED) -> inV -> V-Property(type=="Order")>

Route-extension methods can be called on a single element:

* Because an element can be converted to a route with one element.

Vertex vs. Route

Exercise

• Create extensions for all 6 types:
  • Category 
  • Customer
  • Employee
  • Order
  • Product
  • Supplier
• Conventions:
  • Each extension in its own file
  • Under the NorthWind namespace.
  • Filenames are lower-case

Exercise

• Define route conditions.
  • Based on the type property of a vertex.
• Explore the dataset to discover all types of relations.
  • Use the IRB and/or write a script.
  • Remember: A vertex can have edges of multiple labels.
• Write route-extension methods for all relations.

Display Name

Customers have a few properties.

jruby-1.7.18 :111 > g.v(Customer).first
 => #<V[n10]>

Yet, in the IRB, we only display their internal id.

jruby-1.7.18 :110 > g.v(Customer).first.properties
 => {"phone"=>"(171) 555-1212", 
     "companyName"=>"B's Beverages", 
     "customerID"=>"BSBEV", 
     "type"=>"Customer"}

Can we define some sort of a  toString method? 

Display Name

Add the following vertex extension method:

def display_name
    "#{self[:companyName]} (#{self[:customerID]})"
end

And ...

jruby-1.7.18 :112 > load 'customer.rb'
 => true
jruby-1.7.18 :113 > g.v(Customer).first
 => #<V[n10] B's Beverages (BSBEV)>

Exercise

Add a display_name to every extension you created.

After doing so, you should be able to do the following:

# Get a customer
jruby-1.7.18 :114 > c = g.v(Customer).first
 => #<V[n10] B's Beverages (BSBEV)>

# Get their orders
jruby-1.7.18 :115 > c.orders
#<V[n255] Order 10289> #<V[n913] Order 10947> #<V[n450] Order 10484> ... 
Total: 10

# Get the employees who sold something to this customer
jruby-1.7.18 :116 > c.orders.employees.uniq
#<V[n203] Robert King, Sales Representative> ... 
Total: 7

# Which other customers did these employees serve?
jruby-1.7.18 :117 > c.orders
jruby-1.7.18 :118 >  .employees.uniq
jruby-1.7.18 :119 >  .orders.customers
jruby-1.7.18 :120 >  .uniq.is_not(c)
#<V[n59] Princesa Isabel Vinhos (PRINI)>  #<V[n75] Suprêmes délices (SUPRD)> ...
Total: 89

What's next?

• Common patterns
• Practical tips and tricks
• Considerations when modeling a realistic dataset

We have seen the basic concepts for Data Modeling using Pacer.

Next, we will go through