Cloudant Fundamentals: The Document

General tips on data modeling (part 1 of 10)


Cloudant is a JSON document store, based on Apache CouchDB, running as-a-service in the IBM Cloud. The form of JSON you store in the database is up to you. You don’t need to tell the database about the schema you’re using ahead of time.

Here’s a typical document:

JSON allows you to represent your data through key/value pairs that are:

  • strings
  • numbers
  • booleans
  • objects
  • arrays of any of the above

Cloudant doesn’t restrict you on how detailed or “nested” your objects are (other database services restrict you to top-level key/values, for instance), so you can have arrays of objects that contain arrays of objects, if that is what you need.

Documents with different schemas within the same database? You’ll see that, in Cloudant, we can and should do that. Photo by Kristina Tripkovic on Unsplash.

Rules of thumb

Avoid thinking in joins

Cloudant doesn’t do joins like a relational database, so the JSON objects tend to be denormalised (i.e., contain repetitions of data held elsewhere) . Don’t try to model your data in a relational way and join it yourself in your app.

Size matters

Cloudant will refuse to store documents over 1MB in size, but you shouldn’t be going anywhere near that hard limit. The example document above is less than 1KB.

Schemaless doesn’t mean schema-free

Although Cloudant won’t object if you use a different schema for each document in your database, that’s probably not what you want. One of the first things I do when creating a new project is design the schema, in other words sketch the JSON that represents the objects I’m going to save and the data types of each key. I don’t have to tell Cloudant what the schema is, but that doesn’t mean I shouldn’t give careful consideration to my data model before I jump into the code.

Unlike other databases, Cloudant doesn’t enforce data types or mandatory fields or permitted ranges of values, but it’s likely that you still want to do that in your app.

Multiple object types in the same database

One widely used pattern that is odd to folks coming from a relational background is storing different object types in the same database. You could have blog posts and authors in the same Cloudant database using their own separate schema. One convention is to use the type field to indicate which flavour of object it is.

Write-only document patterns

Cloudant allows documents to be updated, but not on a field-by-field basis. For example, your e-commerce site can’t do:

Your app would have to fetch the whole document and write the whole, modified document back to the database.

Think about whether you can adopt a “write only” approach. For the e-commerce example, a database could contain:

  • a document every time inventory is received into the warehouse
  • a document every time an item is sold

The sum of the stock levels could then be calculated for each product by aggregating the documents. Using this technique you are only ever writing documents, not updating an existing document over and over.

If you’re interested in data modelling, then be sure to read this guide from the Cloudant documentation and watch Joan Touzet’s excellent talk 10 Common Misconceptions about CouchDB:

Next time

Next time we’ll look at a vital component of a Cloudant document: the _id field.