BT
x Your opinion matters! Please fill in the InfoQ Survey about your reading habits!

Data Modeling: Sample E-Commerce System with MongoDB

Posted by Christian Kvalheim on Jun 06, 2012 |

The goal of this article is to demonstrate a data model for a simple e-commerce application that focuses on a couple of key structures contained in typical e-commerce systems. These include product categories, shopping carts, and orders. All example code is in JavaScript using the MongoDB shell but you may adapt all parts of this prototypical code to any MongoDB driver is available.

Overview

Simple e-commerce systems are a good starting point for data modeling with document databases like MongoDB. These examples easily demonstrate core concepts of application development with MongoDB and contain several patterns that you can reuse in other problem domains. MongoDB lets you organize you data in "BSON documents," which you can think of as a "typed JSON" documents. A simple prototypical document might resemble the following:


In a relational database you might express this using two tables. The first table would contain the person's details like first and last name, while a second table would contain details associated with people, with a foreign key back to the persons table.

MongoDB has a rich set of query operators and update operators that let you access documents easily and also perform atomic updates on single fields, arrays or subdocuments of a document.

Requirements

A running MongoDB instance and the mongo shell.

All examples use the shell to avoid distracting from the model with language specific implementation details. You can implement all of the examples and patterns in this article with driver for MongoDB and your language of choice.

These examples purposely avoid more complex error handling to focus on basic operations like accessing and modifying data.

Components

Simsong has just launched their brand new OneDroid phone range to the eager reception of the consumer market. Your task is to build a e-commerce system to take advantage of this huge opportunity and the stock we got allocated.

The Product Catalog

The first step is to design the schema for the website. Consider an initial product schema:

{
  sku: "111445GB3",
  title: "Simsong One mobile phone",
  description: "The greatest Onedroid phone on the market .....",

  manufacture_details: {
    model_number: "A123X",
    release_date: new ISODate("2012-05-17T08:14:15.656Z")
  },

  shipping_details: {
    weight: 350,
    width: 10,
    height: 10,
    depth: 1
  },

  quantity: 99,

  pricing: {
    price: 1000
  }
}

This data model stores physical details like manufacturing and shipping information as embedded documents in the larger product document, which makes sense because these physical details are unique features of the product. This gives the document "strong data locality," which allows easy mapping in an object oriented environment.

To insert the document in the products collection, use the following commands.

mongo    
use ecommerce    
db.products.insert({
  sku: "111445GB3",
  title: "Simsong One mobile phone",
  description: "The greatest Onedroid phone on the market .....",

  manufacture_details: {
    model_number: "A123X",
    release_date: new ISODate("2012-05-17T08:14:15.656Z")
  },

  shipping_details: {
    weight: 350,
    width: 10,
    height: 10,
    depth: 1
  },

  quantity: 99,

  pricing: {
    price: 1000
  }
})

The first command (mongo) starts the mongodb console and connects to the local Mongo DB console on localhost and port 27017. The next chooses the ecommerce database (use ecommerce) and the third inserts the product document in the products collection. Going forward all commands are assuming you are in the Mongo DB shell using the ecommerce database.

The products data model has a unique sku that identifies the product, title, description, a stock quantity, and pricing information about the item.

All products have categories. In the case of the Simsong One it's a 15G phone and also has a FM receiver. As a result, This product falls into both the mobile/15G and the radio/fm categories. Add the categories to the existing document, with the following update() operation:

 db.products.update({sku: "111445GB3"}, {$set: { categories: ['mobile/15G', 'mobile/fm'] }});

To support efficient queries using the categories field, add an index on the categories field for the products collection:

 db.products.ensureIndex({categories:1 })

This returns all the products for a specific category using the index and an anchored regular expression. As long as the regular expression is case sensitive and anchored, MongoDB will use the index to return the query. For example, fetch all the products in the category that begins with mobile/fm:

 db.products.find({categories: /^mobile\/fm/})

To be able to provide a list of all the products in a category, amend the data model with a collection of documents for each category. In this collection, each document represents a category and contains the path for that category in category tree. These documents would resemble the following.

 {
   title: "Mobiles containing a FM radio",
   parent: "mobile",
   path: "mobile/fm"
 }

Insert the document into the categories collection and add indexes to this collection:

 db.categories.insert({title: "Mobiles containing a FM radio", parent: "mobile",  path: "mobile/fm"})
 db.categories.insert({title: "Mobiles with 15G support", parent: "mobile",  path: "mobile/15G"})
 db.categories.ensureIndex({parent: 1, path: 1})
 db.categories.ensureIndex({path: 1})

There are two paths in each category: this allows the application to use the same method to find all categories for a specific category root as used for finding products by category. For example, to return all sub-categories of the category "mobile", use the following query:

db.categories.find({parent: /^mobile/}, {_id: 0, path: 1})

This will return the following documents:

 {"path": "mobile/fm"}
 {"path": "mobile/15G"}

Using these path values, the application can use this method to access the category tree and extract more sub-categories with a single index supported query. Furthermore, the application can pull all the documents for a specific category using this path value.

The Cart

A cart in an e-commerce system, allows users to reserve items from the inventory and keep them until they check out and pay for the items. The application must ensure that at any point in time there are not more items in carts than there are in stock and that if the users abandons the cart, the application must return the items from the cart to the inventory without loosing track of any objects. Take the following document, which models the cart:

 {
   _id: "the_users_session_id",
   status:'active'

   quantity: 2,
   total: 2000,

   products: []
 }

The products array contains the list of products the customer intends to purchase. Use the following insert() operation to create the cart:

  db.carts.insert({
   _id: "the_users_session_id",
   status:'active',
   quantity: 2,
   total: 2000,
   products: []});

If the inventory had 99 items, after this operation, the inventory should have 97 items. To prevent "overselling," the application must move items from the inventory to the cart. To support these operations applications must perform a set of updates, and be able "rollback" changes if something goes awry. Begin by adding a product to the customer's cart with the following operation:

  db.carts.update({
     _id: "the_users_session_id", status:'active'
   }, {
     $set: { modified_on: ISODate() },
     $push: {
       products: {
         sku: "111445GB3", quantity: 1, title: "Simsong One mobile phone", price:1000
       }
     }
   });

Then, check to ensure that the inventory can support adding the product to the customers cart:

 db.products.update({
     sku: "111445GB3", quantity: {$gte: 1}
   }, {
     $inc: {quantity: -1},
     $push: {
       in_carts: {
         quantity:1, id: "the_users_session_id", timestamp: new ISODate()
       }
     }
   })

This operation only succeeds if there is sufficient inventory, and the application must detect the operation's success or failure. Call getLastError to fetch the result of the attempted update:

  if(!db.runCommand({getLastError:1}).updatedExisting) {
   db.carts.update({
       _id: "the_users_session_id"
     }, {
       $pull: {products: {sku:"111445GB3"}}
   })

}

If updatedExisting is false in the resulting document, the operation failed and the application must "roll back" the attempt to add the product to the users cart. This pattern ensures that the application cannot have more products in carts than the available inventory.

In addition to simply adding objects to carts, there are a number of cart related operations that the application must be able to support:

  • users may add or remove objects from the cart.

  • users may abandon a cart and the application must return items in the cart to inventory.

The next sequence of operations allow the application to ensure that carts are up to date and that the application has enough inventory to cover it. Update the cart with the new quantity, using the following update() operation:

var new_quantity = 2;
var old_quantity = 1;
var quantity_delta = new_quantity - old_quantity;

db.carts.update({
    _id: "the_users_session_id", "products.sku": "111445GB3", status: "active"
  }, {
    $set: {
      modified_on: new ISODate(),
      "products.$.qty": new_quantity
  }
})

Notice that there are some defined variables called newquantity, oldquantity and quantity_delta to contain the new and previous quantity in the cart as well as the delta that needs to be requested from the inventory.

Now, remove the additional item from the inventory update the number of items in the shopping cart:

  db.products.update({
     sku: "111445GB3",
     "in_carts.id": "the_users_session_id",
     quantity: {
       $gte: 1
     }
   }, {
     $inc: { quantity: (-1)*quantity_delta },
     $set: {
       "in_carts.$.quantity": new_quantity, timestamp: new ISODate()
     }
   })

Ensure the application has enough inventory for the operation. If there is not sufficient inventory, the application must rollback the last operation. The following operation checks for errors using getLastError and rolls back the operation if it returns an error:

 if(!db.runCommand({getLastError:1}).updatedExisting) {
   db.carts.update({
       _id: "the_users_session_id", "products.sku": "111445GB3"
     }, {
       $set : { "in_carts.$.quantity": old_quantity}
     })
 }

If a user abandons the purchase process or the shopping cart grows stale and times out, the application must return cart content to the inventory. This operation requires a loop that finds all expired or canceled carts and then returns the content of each cart to the inventory. Begin by finding all sufficiently "stale" carts, and use an operation that resembles the following:

 var carts = db.carts.find({status:"expiring"})
 for(var i = 0; i < carts.length; i++) {
   var cart = carts[i]

   for(var j = 0; j < cart.products.length; j++) {
     var product = cart.products[i]

     db.products.update({
         sku: product.sku,
         "in_carts.id": cart._id,
         "in_carts.quantity": product.quantity
       }, {
         $inc: {quantity: item.quantity},
         $pull: {in_carts: {id: cart._id}}
       })
   }

   db.carts.update({
       _id: cart._id,
       $set: {status: 'expired'}
     })
 }

This operation walks all products in each cart and returns them to the inventory and removes cart identifiers from the in_carts array in the product documents. Once the application has returned all of the items to the inventory, the application sets the cart's status to expired.

Checkout

When the user clicks the "confirm" button in the checkout portion of the application, the application creates an "order" document that reflects the entire order. Consider the following operation:

 db.orders.insert({
   created_on: new ISODate("2012-05-17T08:14:15.656Z"),

   shipping: {
     customer: "Peter P Peterson",
     address: "Longroad 1343",
     city: "Peterburg",
     region: "",
     state: "PE",
     country: "Peteonia",
     delivery_notes: "Leave at the gate",

     tracking: {
       company: "ups",
       tracking_number: "22122X211SD",
       status: "ontruck",
       estimated_delivery: new ISODate("2012-05-17T08:14:15.656Z")
     },
   },

   payment: {
     method: "visa",
     transaction_id: "2312213312XXXTD"
   }

   products: {
     {quantity: 2, sku:"111445GB3", title: "Simsong mobile phone", unit_cost:1000, currency:"USDA"}
   }
 })

For a relational databases you might need to model this as a set of tables: for orders, shipping, tracking, and payment. Using MongoDB one can create a single document that is self-contained, easy to understand, and simply maps into an object oriented application. After inserting the this document the application must ensure inventory is up to date before completing the checkout. Begin by setting the cart as finished, with the following operation:

 db.carts.update({
     _id: "the_users_session_id"
   }, {
     $set: {status:"complete"}
   });

Use the following operation to remove the cart identifer from all product records:

 db.products.update({
     "in_carts.id": "the_users_session_id"
   }, {
     $pull: {in_carts: {id: "the_users_session_id"}}
   }, false, true);

By using "multi-update," which is the last argument in the update() method, this operation will update all matching documents in one set of operations.

Conclusion

The rich document capabilities atomic operation guarantees in MongoDB makes it possible to model many different applications in MongoDB. Even the rigorous requirements of conventional applications like e-commerce system are possible in a document database. This data model (i.e. "schema design,") is useful for developing applications around any restricted resource system, not just e-commerce systems.

The close relationship match between object oriented application code and documents leads to more simple data models and less glue code between the data storage system and the application-level code.

About the Author

Christian Kvalheim, Node.js driver developer and evangalist for Mongo DB has worked as a developer and Director of Engineering for the last 12 years. He's worked for Thoughtworks, IMSI, Xing and now 10gen on a variety of different technology platforms. Besides the Node.js driver he spends his time in Erlang, Java and C.

Hello stranger!

You need to Register an InfoQ account or or login to post comments. But there's so much more behind being registered.

Get the most out of the InfoQ experience.

Tell us what you think

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

MongoDB with Spring by Ilias Tsagklis

Nice article. MongoDB has a lot of promise. You can also check how to use it on a higher level combined with Spring's power: MongoDB with Spring Data project

Re: MongoDB with Spring by Francois MAROT

There's also Jongo, young but promising: blog.xebia.com/2012/06/29/jongo/

cart expiration by gabriel duncan

i know this is two years after the fact, but i'm in a situation where i need to handle cart expiration. could you recommend how to do this in the most efficient way?

more specifically, you change the cart status from 'active' to 'expiring'. what is the best way to do this? how to keep track of time a cart has had an item in it? do you just run a polling process on the server like a cron job that checks all carts after an interval?

any help would be greatly appreciated!

Re: cart expiration by Anup Marwadi

Hello, I have implemented a similar MongoDB shopping cart.

Here's how I resolved the Cart Expiration:
1. We've built a ticketing system which means that users can keep tickets in cart only for 'x' minutes. If they do not complete the checkout within 'x' minutes, we release the tickets back to the pool.
2. When the first set of data (we call it a line item) is added to a cart, we provide an expiration date for the cart. The cart is also marked as In-Process.
3. We have a Quartz Job (Quartz.Net if you're using .NET, which we are), that runs every second and checks up on which carts have exceeded the Expiration duration and then releases the tickets back to the pool and expires the cart by marking the status as Expired. Once the cart is marked as Expired, the user gets a notification on the front-end (through a polling Javascript (or Websockets whichever is easier) and the user will be directed back to the landing page after disposing the cart.

Not only does this help the Cart expiration issue, but it also gives our customers an idea on Cart Analytics.

Inventory Management using MongoDB by Anup Marwadi

Hello, this is a great article and has been the basis of a Mongo based shopping cart that I design. Thanks!

One of the things that I'm concerned about is the ability to manage Inventory/Stock accurately due to the inability to "lock" records in MongoDB like you would do on a relational database so that exclusive read/writes can happen to manage and update inventory without others intervening. I would highly appreciate it if someone else who has had experience on this can provide us with some guidance.

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

5 Discuss

Educational Content

General Feedback
Bugs
Advertising
Editorial
InfoQ.com and all content copyright © 2006-2014 C4Media Inc. InfoQ.com hosted at Contegix, the best ISP we've ever worked with.
Privacy policy
BT