Just a quick reminder of the difference between authentication and authorization:

If you’ve used Couchbase before, you’re familiar with the login to what we sometimes call the "Admin Web Console".

However, the Web Console is really not just for admins, it’s for developers too. But until now, you didn’t really have a lot of control built-in to Couchbase about who can log in and (more importantly) what they’re allowed to do.

So, I’d like to introduce you to Couchbase’s new first-class user feature.

There’s still a full administrator user. This is the login that you create when you first install Couchbase. This is the user who is unrestricted, and can do anything, including creating new users. So, for instance, a full administrator can see the "Security" link in the navigation, while other users can’t.

Now, once on this security page, you can add, edit, and delete users.

A user can identify a person, but it can also identify some service or process. For instance, if you’re writing an ASP.NET application, you may want to create a user with a limited set of permissions called "web-service". Therefore, the credentials for that "user" would not be for a person, but for an ASP.NET application.

Next, try adding a new Couchbase user by clicking "+ Add User". I’m going to create a user called "fts_admin", with a name of "Full Text Search Admin", a password, and a single role: FTS Admin of the travel-sample bucket (FTS = Full Text Search).

Here’s an animation of adding that user:

Some notes about the above animation:

More on the difference between FTS Admin and FTS Searcher later.

Now that this user is created, I can login as fts_admin. This user’s authentication is handled within Couchbase.

First, in the above animation, note that the fts_admin user has a much more limited set of options compared to the full admin user.

Next, it’s worth pointing out that users can reset their password:

Since I’ve already created an fts_admin user with the FTS Admin role, I’ll create another user called fts_searcher that only has the FTS Searcher role for the travel-sample bucket.

{
  "type": "fulltext-index",
  "name": "hotels",
  "sourceType": "couchbase",
  "sourceName": "travel-sample",

// ... snip ...

}

{
  "message": "Forbidden. User needs one of the following permissions",
  "permissions": [
    "cluster.bucket[travel-sample].fts!write"
  ]
}

{
  "status": "ok"
}

Next, let’s use the REST API to search the index for the word 'breakfast'.

First, make a POST to the /api/index/hotels/query endpoint, again with the proper credentials and port number.

http://fts_admin:password@192.168.1.10:8094/api/index/hotels/query

or

http://fts_searcher:password@192.168.1.10:8094/api/index/hotels/query

Both users should be able to execute a search using that index.

Next, in the body of the POST should be a simple JSON object. Again, you don’t normally have to create this by hand — your SDK of choice or the Web Console UI can do this for you.

{
  "explain": true,
  "fields": [
    "*"
  ],
  "highlight": {},
  "query": {
    "query": "breakfast"
  }
}

Finally, the result of this search request will be a large JSON response. Look within the "hits" sub-document for "fragments" to verify that the search worked. Here’s a snippet of my search for "breakfast". Again, the full result is on Github.

// ... snip ...

        "reviews.content": [
          "… to watch TV. <mark>Breakfast</mark> was served every morning along with a copy of the Times-Picayune. I took my <mark>breakfast</mark> downstairs in the patio, the coffee was very good. The continental <mark>breakfast</mark> is nothing to…"
        ]
      },

// ... snip ...

There are some bugs and some incomplete features.

If you have questions, the best way to contact me is either Twitter @mgroves or email me matthew.groves@couchbase.com.

When I’m writing N1QL queries, I need to be able to understand how well (or how badly) my query (and my cluster) is performing in order to make improvements and diagnose issues.

With this latest developer version of Couchbase Server 5.0, some new tools have been added to your N1QL-writing toolbox.

First, some review.

There are multiple ways for a developer to execute N1QL queries.

In this post, I’ll be mainly using Query Workbench.

There are two system catalogs that are already available to you in Couchbase Server 4.5 that I’ll be talking about today.

Both active_requests and completed_requests return not only the original N1QL query text, but also related information: request time, request id, execution time, scan consistency, and so on. This can be useful information. Here’s an example that looks at a simple query (select * from `travel-sample`) while it’s running by executing select * from system:active_requests;

{
	"active_requests": {
	  "clientContextID": "805f519d-0ffb-4adf-bd19-15238c95900a",
	  "elapsedTime": "645.4333ms",
	  "executionTime": "645.4333ms",
	  "node": "10.0.75.1",
	  "phaseCounts": {
		"fetch": 6672,
		"primaryScan": 7171
	  },
	  "phaseOperators": {
		"fetch": 1,
		"primaryScan": 1
	  },
	  "phaseTimes": {
		"authorize": "500.3µs",
		"fetch": "365.7758ms",
		"parse": "500µs",
		"primaryScan": "107.3891ms"
	  },
	  "requestId": "80787238-f4cb-4d2d-999f-7faff9b081e4",
	  "requestTime": "2017-02-10 09:06:18.3526802 -0500 EST",
	  "scanConsistency": "unbounded",
	  "state": "running",
	  "statement": "select * from `travel-sample`;"
	}
}

First, I want to point out that phaseTimes is a new addition to the results. It’s a quick and dirty way to get a sense of the query cost without looking at the whole profile. It gives you the overall cost of each request phase without going into detail of each operator. In the above example, for instance, you can see that parse took 500µs and primaryScan took 107.3891ms. This might be enough information for you to go on without diving into META().plan.

However, with the new META().plan, you can get very detailed information about the query plan. This time, I’ll execute SELECT *, META().plan FROM system:active_requests;

[
  {
    "active_requests": {
      "clientContextID": "75f0f401-6e87-48ae-bca8-d7f39a6d029f",
      "elapsedTime": "1.4232754s",
      "executionTime": "1.4232754s",
      "node": "10.0.75.1",
      "phaseCounts": {
        "fetch": 12816,
        "primaryScan": 13231
      },
      "phaseOperators": {
        "fetch": 1,
        "primaryScan": 1
      },
      "phaseTimes": {
        "authorize": "998.7µs",
        "fetch": "620.704ms",
        "primaryScan": "48.0042ms"
      },
      "requestId": "42f50724-6893-479a-bac0-98ebb1595380",
      "requestTime": "2017-02-15 14:44:23.8560282 -0500 EST",
      "scanConsistency": "unbounded",
      "state": "running",
      "statement": "select * from `travel-sample`;"
    },
    "plan": {
      "#operator": "Sequence",
      "#stats": {
        "#phaseSwitches": 1,
        "kernTime": "1.4232754s",
        "state": "kernel"
      },
      "~children": [
        {
          "#operator": "Authorize",
          "#stats": {
            "#phaseSwitches": 3,
            "kernTime": "1.4222767s",
            "servTime": "998.7µs",
            "state": "kernel"
          },
          "privileges": {
            "default:travel-sample": 1
          },
          "~child": {
            "#operator": "Sequence",
            "#stats": {
              "#phaseSwitches": 1,
              "kernTime": "1.4222767s",
              "state": "kernel"
            },
            "~children": [
              {
                "#operator": "PrimaryScan",
                "#stats": {
                  "#itemsOut": 13329,
                  "#phaseSwitches": 53319,
                  "execTime": "26.0024ms",
                  "kernTime": "1.3742725s",
                  "servTime": "22.0018ms",
                  "state": "kernel"
                },
                "index": "def_primary",
                "keyspace": "travel-sample",
                "namespace": "default",
                "using": "gsi"
              },
              {
                "#operator": "Fetch",
                "#stats": {
                  "#itemsIn": 12817,
                  "#itemsOut": 12304,
                  "#phaseSwitches": 50293,
                  "execTime": "18.5117ms",
                  "kernTime": "787.9722ms",
                  "servTime": "615.7928ms",
                  "state": "services"
                },
                "keyspace": "travel-sample",
                "namespace": "default"
              },
              {
                "#operator": "Sequence",
                "#stats": {
                  "#phaseSwitches": 1,
                  "kernTime": "1.4222767s",
                  "state": "kernel"
                },
                "~children": [
                  {
                    "#operator": "InitialProject",
                    "#stats": {
                      "#itemsIn": 11849,
                      "#itemsOut": 11848,
                      "#phaseSwitches": 47395,
                      "execTime": "5.4964ms",
                      "kernTime": "1.4167803s",
                      "state": "kernel"
                    },
                    "result_terms": [
                      {
                        "expr": "self",
                        "star": true
                      }
                    ]
                  },
                  {
                    "#operator": "FinalProject",
                    "#stats": {
                      "#itemsIn": 11336,
                      "#itemsOut": 11335,
                      "#phaseSwitches": 45343,
                      "execTime": "6.5002ms",
                      "kernTime": "1.4157765s",
                      "state": "kernel"
                    }
                  }
                ]
              }
            ]
          }
        },
        {
          "#operator": "Stream",
          "#stats": {
            "#itemsIn": 10824,
            "#itemsOut": 10823,
            "#phaseSwitches": 21649,
            "kernTime": "1.4232754s",
            "state": "kernel"
          }
        }
      ]
    }
  }, ...
]

The above output comes from the Query Workbench.

Note the new "plan" part. It contains a tree of operators that combine to execute the N1QL query. The root operator is a Sequence, which itself has a collection of child operators like Authorize, PrimaryScan, Fetch, and possibly even more Sequences.

To get this information when using cbq or the REST API, you’ll need to turn on the "profile" feature.

You can do this in cbq by entering set -profile timings; and then running your query.

You can also do this with the REST API on a per request basis (using the /query/service endpoint and passing a querystring parameter of profile=timings, for instance).

You can turn on the setting for the entire node by making a POST request to http://localhost:8093/admin/settings, using Basic authentication, and a JSON body like:

{
  "completed-limit": 4000,
  "completed-threshold": 1000,
  "controls": false,
  "cpuprofile": "",
  "debug": false,
  "keep-alive-length": 16384,
  "loglevel": "INFO",
  "max-parallelism": 1,
  "memprofile": "",
  "pipeline-batch": 16,
  "pipeline-cap": 512,
  "pretty": true,
  "profile": "timings",
  "request-size-cap": 67108864,
  "scan-cap": 0,
  "servicers": 32,
  "timeout": 0
}

Notice the profile setting. It was previously set to off, but I set it to "timings".

You may not want to do that, especially on nodes being used by other people and programs, because it will affect other queries running on the node. It’s better to do this on a per-request basis.

It’s also what Query Workbench does by default.

There’s a lot of information in META().plan about how the plan is executed. Personally, I prefer to look at a simplified graphical version of it in Query Workbench by clicking the "Plan" icon (which I briefly mentioned in a previous post about the new Couchbase Web Console UI).

Let’s look at a slightly more complex example. For this exercise, I’m using the travel-sample bucket, but I have removed one of the indexes (DROP INDEX `travel-sample.def_sourceairport;`).

I then execute a N1QL query to find flights between San Francisco and Miami:

SELECT r.id, a.name, s.flight, s.utc, r.sourceairport, r.destinationairport, r.equipment
FROM `travel-sample` r
UNNEST r.schedule s
JOIN `travel-sample` a ON KEYS r.airlineid
WHERE r.sourceairport = 'SFO'
AND r.destinationairport = 'MIA'
AND s.day = 0
ORDER BY a.name;

Executing this query (on my single-node local machine) takes about 10 seconds. That’s definitely not an acceptible amount of time, so let’s look at the plan to see what the problem might be (I broke it into two lines so the screenshots will fit in the blog post).

Looking at that plan, it seems like the costliest parts of the query are the Filter and the Join. JOIN operations work on keys, so they should normally be very quick. But it looks like there are a lot of documents being joined.

The Filter (the WHERE part of the query) is also taking a lot of time. It’s looking at the sourceairport and destinationairport fields. Looking elsewhere in the plan, I see that there is a PrimaryScan. This should be a red flag when you are trying to write performant queries. PrimaryScan means that the query couldn’t find an index other than the primary index. This is roughly the equivalent of a "table scan" in relational database terms. (You may want to drop the primary index so that these issues get bubbled-up faster, but that’s a topic for another time).

Let’s add an index on the sourceairport field and see if that helps.

CREATE INDEX `def_sourceairport` ON `travel-sample`(`sourceairport`);

Now, running the same query as above, I get the following plan:

This query took ~100ms (on my single-node local machine) which is much more acceptible. The Filter and the Join still take up a large percentage of the time, but thanks to the IndexScan replacing the PrimaryScan, there are many fewer documents that those operators have to deal with. Perhaps the query could be improved even more with an additional index on the destinationairport field.

The answer to performance problems is not always in tweaking queries. Sometimes you might need to add more nodes to your cluster to address the underlying problem.

Look at the PrimaryScan information in META().plan. Here’s a snippet:

"~children": [
  {
    "#operator": "PrimaryScan",
    "#stats": {
      "#itemsOut": 13329,
      "#phaseSwitches": 53319,
      "execTime": "26.0024ms",
      "kernTime": "1.3742725s",
      "servTime": "22.0018ms",
      "state": "kernel"
    },
    "index": "def_primary",
    "keyspace": "travel-sample",
    "namespace": "default",
    "using": "gsi"
  }, ... ]

The servTime value indicates how much time is spent by the Query service to wait on the Key/Value data storage. If the servTime is very high, but there is a small number of documents being processed, that indicates that the indexer (or the key/value service) can’t keep up. Perhaps they have too much load coming from somewhere else. So this means that something weird is running someplace else or that your cluster is trying to handle too much load. Might be time to add some more nodes.

Similarly, the kernTime is how much time is spent waiting on other N1QL routines. This might mean that something else downstream in the query plan has a problem, or that the query node is overrun with requests and are having to wait a lot.

The new META().plan functionality and the new Plan UI combine in Couchbase Server 5.0 to improve the N1QL writing and profiling process.

Stay tuned to the Couchbase Blog for information about what’s coming in the next developer build.

Interested in trying out some of these new features? Download Couchbase Server 5.0 today!

We want feedback! Developer releases are coming every month, so you have a chance to make a difference in what we are building.

Bugs: If you find a bug (something that is broken or doesn’t work how you’d expect), please file an issue in our JIRA system at issues.couchbase.com or submit a question on the Couchbase Forums. Or, contact me with a description of the issue. I would be happy to help you or submit the bug for you (my Couchbase handlers high-five me every time I submit a good bug).

Feedback: Let me know what you think. Something you don’t like? Something you really like? Something missing? Now you can give feedback directly from within the Couchbase Web Console. Look for the icon at the bottom right of the screen.

In some cases, it may be tricky to decide if your feedback is a bug or a suggestion. Use your best judgement, or again, feel free to contact me for help. I want to hear from you. The best way to contact me is either Twitter @mgroves or email me matthew.groves@couchbase.com.

Couchbase (and many other document databases) use JSON objects for data. JSON is a powerful, human readable format to store data. When comparing to data types in relational tables, there are some similarities, and there are some important differences.

All JSON data is made up of 6 types: string, number, boolean, array, object, and null. There are a lot of data types available in SQL Server. Let’s start with a table that is a kind of "literal" translation, and work from there.

It’s important to understand how JSON works. I’ve listed some high-level differences between JSON data types and SQL Server data types. Assuming you already understand SQL data types, you might want to spend some time learning more about JSON and JSON data types.

A string in SQL Server is often defined by a length. nvarchar(50) or nvarchar(MAX) for instance. In JSON, you don’t need to define a length. Just use a string.

A number in SQL Server varies widely based on what you are using it for. The number type in JSON is flexible, in that it can store integers, decimal, or floating point. In specialized circumstances, like if you need a specific precision or you need to store very large numbers, you may want to store a number as a string instead.

A boolean in JSON is true/false. In SQL Server, it’s roughly equivalent: a bit that represents true/false.

In JSON, any value can be null. In SQL Server, you set this on a field-by-field basis. If a field in SQL Server is not set to "nullable", then it will be enforced. In a JSON document, there is no such enforcement.

JSON has no date data type. Often dates are stored as UNIX timestamps, but you could also use string representations or other formats for dates. The N1QL query language has a variety of date functions available, so if you want to use N1QL on dates, you can use those functions to plan your date storage accordingly.

In SQL Server, there is a geography data type. In Couchbase, the GeoJSON format is supported.

There are some other specialized data types in SQL Server, including hierarchyid, and xml. Typically, these would be unrolled in JSON objects and/or referenced by key (as explored in part 1 of this blog series on data modeling). You can still store XML/JSON within a string if you want, but if you do, then you can’t use the full power of N1QL on those fields.

Depending on your organization and your team, you may have to bring in people from multiple roles to ensure a successful migration. If you have a DBA, that DBA will have to know how to run and manage Couchbase just as well as SQL Server. If you are DevOps, or have a DevOps team, it’s important to involve them early on, so that they are aware of what you’re doing and can help you coordinate your efforts. Moving to a document database does not mean that you no longer need DBAs or Ops or DevOps to be involved. These roles should also be involved when doing data modeling, if possible, so that they can provide input and understand what is going on.

After you’ve designed your model with part 1 on data modeling, you can start moving data over to Couchbase.

For a naive migration (1 row to 1 document), you can write a very simple program to loop through the tables, columns, and values of a relational database and spit out corresponding documents. A tool like Dapper would handle all the data type translations within C# and feed them into the Couchbase .NET SDK.

Completely flat data is relatively uncommon, however, so for more complex models, you will probably need to write code to migrate from the old relational model to the new document model.

Here are some things you want to keep in mind when writing migration code (of any kind, but especially relational-to-nonrelational):

public class ShoppingCartMigrator
{
    readonly IBucket _bucket;
    readonly SqlToCbContext _context;

    public ShoppingCartMigrator(IBucket bucket, SqlToCbContext context)
    {
        _bucket = bucket;
        _context = context;
    }
}

public bool Go()
{
    var carts = _context.ShoppingCarts
        .Include(x => x.Items)
        .ToList();
    foreach (var cart in carts)
    {
        var cartDocument = new Document<dynamic>
        {
            Id = cart.Id.ToString(),
            Content = MapCart(cart)
        };
        var result = _bucket.Insert(cartDocument);
        if (!result.Success)
        {
            Console.WriteLine($"There was an error migrating Shopping Cart {cart.Id}");
            return false;
        }
        Console.WriteLine($"Successfully migrated Shopping Cart {cart.Id}");
    }
    return true;
}

public void Cleanup()
{
    Console.WriteLine("Delete all shopping carts...");
    var result = _bucket.Query<dynamic>("DELETE FROM `sqltocb` WHERE type='ShoppingCart';");
    if (!result.Success)
    {
        Console.WriteLine($"{result.Exception?.Message}");
        Console.WriteLine($"{result.Message}");
    }
}

Not everything in SQL Server has a direct counterpart in Couchbase. In some cases, it won’t ever have a counterpart. In some cases, there will be a rough equivalent. Some features will arrive in the future, as Couchbase is under fast-paced, active, open-source development, and new features are being added when appropriate.

Also keep in mind that document databases and NoSQL databases often force business logic out of the database to a larger extent than relational databases. As nice as it would be if Couchbase Server had every feature under the sun, there are always tradeoffs. Some are technical in nature, some are product design decisions. Tradeoffs could be made to add relational-style features, but at some point in that journey, Couchbase stops being a fast, scalable database and starts being "just another" relational database. There is certainly a lot of convergence in both relational and non-relational databases, and a lot of change happening every year.

With that in mind, stay tuned for the final blog post in the series. This will cover the changes to application coding that come with using Couchbase, including:

This blog post compared and contrasted the data features available in Couchbase Server with SQL Server. If you are currently using SQL Server and are considering adding a document database to your project or starting a new project, I am here to help.

Check out the Couchbase developer portal for more details.

Please contact me at matthew.groves@couchbase.com, ask a question on the Couchbase Forums, or ping me on Twitter @mgroves.

Before we get started, I want to spend a little bit of time on motivation. There are 3 main reasons why one might consider using a document data store instead of (or in addition to) a relational database. Your motivation may be one or all three:

For these reasons and others, Gannett switched from SQL Server to Couchbase Server. If you are considering this, definitely check out Gannett’s full presentation.

It should be noted that document databases and relational databases can be complimentary. Your application may be best served by one, the other, or a combination of both. In many cases, it simply is not possible to completely remove relational databases from your design, but a document database like Couchbase Server can still bring the above benefits to your software. The rest of this blog series will assume you have a SQL Server background and are either replacing, supplimenting, or starting a new greenfield project using Couchbase.

The ease or difficulty of transitioning an existing application varies widely based on a number of factors. In some cases it may be extremely easy; in some cases it will be time-consuming and difficult; in some (shrinking number of) cases it may not even be a good idea.

The first step is to understand how data is modeled in a document database. In a relational database, data is typically stored flat in a table and it is given structure with primary and foreign keys. As a simple example, let’s consider a relational database for a web site that has a shopping cart as well as social media features. (In this example, those features are unrelated to keep things simple).

In a document database, data is stored as keys and values. A Couchbase bucket contains documents; each document has a unique key and a JSON value. There are no foreign keys (or, more accurately, there are no foreign key constraints).

Here’s a high-level comparison of SQL Server features/naming as compared to Couchbase:

These comparisons are a metaphorical starting point. Looking at that table, it might be tempting to take a simplistic approach. "I have 5 tables, therefore I’ll just create 5 different types of documents, with one document per row." This is the equivalent of literally translating a written language. The approach may work sometimes, but it doesn’t take into account the full power of a document database that uses JSON. Just as a literal translation of a written language doesn’t take into account cultural context, idioms, and historical context.

Because of the flexibility of JSON, the data in a document database can be structured more like a domain object in your application. Therefore you don’t have an impedence mismatch that is often addressed by OR/M tools like Entity Framework and NHibernate.

There are two main approaches you can use when modeling data in Couchbase that we will examine further:

Let’s consider the "shopping cart" entity.

To represent this in a relational database would likely require two tables: a ShoppingCart table and a ShoppingCartItem table with a foreign key to a row in ShoppingCart.

When creating the model for a document database, the decision has to be made whether to continue modeling this as two separate entities (e.g. a Shopping Cart document and corresponding Shopping Cart Item documents) or whether to "denormalize" and combine a row from ShoppingCart and row(s) from ShoppingCartItem into a single document to represent a shopping cart.

In Couchbase, using a denormalization strategy, a shopping cart and the items in it would be represented by a single document.

{
  "user": "mgroves",
  "dateCreated": "2017-02-02T15:28:11.0208157-05:00",
  "items": [
    {
      "name": "BB-8 Sphero",
      "price": 80.18,
      "quantity": 1
    },
    {
      "name": "Shopkins Season 5",
      "price": 59.99,
      "quantity": 2
    }
  ],
  "type": "ShoppingCart"
}

Notice that the relationship between the items and the shopping cart is now implicit to being contained in the same document. No more need for an ID on the items to represent a relationship.

In C#, you would likely define ShoppingCart and Item classes to model this data:

public class ShoppingCart
{
    public Guid Id { get; set; }
    public string User { get; set; }
    public DateTime DateCreated { get; set; }
    public List<Item> Items { get; set; }
}

public class Item
{
    public Guid Id { get; set; }    // necessary for SQL Server, not for Couchbase
    public string Name { get; set; }
    public decimal Price { get; set; }
    public int Quantity { get; set; }
}

These classes would still make sense with Couchbase, so you can reuse them or design them this way. But with a relational database, this design does not match up in a direct way.

Hence the need for OR/Ms like NHibernate or Entity Framework. The way the above model can be mapped to a relational database is represented in Entity Framework* like this:

public class ShoppingCartMap : EntityTypeConfiguration<ShoppingCart>
{
    public ShoppingCartMap()
    {
        this.HasKey(m => m.Id);

        this.ToTable("ShoppingCart");
        this.Property(m => m.User);
        this.Property(m => m.DateCreated);
        this.HasMany(m => m.Items)
            .WithOptional()
            .HasForeignKey(m => m.ShoppingCartId);
    }
}

public class ShoppingCartItemMap : EntityTypeConfiguration<Item>
{
    public ShoppingCartItemMap()
    {
        this.HasKey(m => m.Id);

        this.ToTable("ShoppingCartItems");
        this.Property(m => m.Name);
        this.Property(m => m.Price);
        this.Property(m => m.Quantity);
    }
}

*Other OR/Ms will have similar mappings

Based on these mappings and an analysis of the use cases, I could decide that it would be modeled as a single document in Couchbase. ShoppingCartItemMap only exists so that the OR/M knows how to populate the Items property in ShoppingCart. Also, it’s unlikely that the application will be doing reads of the shopping cart without also needing to read the items.

In a later post, OR/Ms will be discussed further, but for now I can say that the ShoppingCartMap and ShoppingCartItemMap classes are not necessary when using Couchbase, and the Id field from Item isn’t necessary. In fact, the Couchbase .NET SDK can directly populate a ShoppingCart object without an OR/M in a single line of code:

public ShoppingCart GetCartById(Guid id)
{
    return _bucket.Get<ShoppingCart>(id.ToString()).Value;
}

This isn’t to say that using Couchbase will always result in shorter, easier to read code. But for certain use cases, it can definitely have an impact.

It’s not always possible or optimal to denormalize relationships like the ShoppingCart example. In many cases, a document will need to reference another document. Depending on how your application expects to do reads and writes, you may want to keep your model in separate documents by using referencing.

Let’s look at an example where referencing might be the best approach. Suppose your application has some social media elements. Users can have friends, and users can post text updates.

One way to model this:

With two users, two updates, we would have 4 documents in Couchbase that look like this:

[
  // Key: "7fc5503f-2092-4bac-8c33-65ef5b388f4b"
  {
    "friends": [
      "c5f05561-9fbf-4ab0-b68f-e392267c0703"
    ],
    "name": "Matt Groves",
    "type": "User"
  },

  // Key: "c5f05561-9fbf-4ab0-b68f-e392267c0703"
  {
    "friends": [ ],
    "name": "Nic Raboy",
    "type": "User"
  },

  // Key: "5262cf62-eb10-4fdd-87ca-716321405663"
  {
    "body": "Nostrum eligendi aspernatur enim repellat culpa.",
    "postedDate": "2017-02-02T16:19:45.2792288-05:00",
    "type": "Update",
    "user": "7fc5503f-2092-4bac-8c33-65ef5b388f4b"
  },

  // Key: "8d710b83-a830-4267-991e-4654671eb14f"
  {
    "body": "Autem occaecati quam vel. In aspernatur dolorum.",
    "postedDate": "2017-02-02T16:19:48.7812386-05:00",
    "type": "Update",
    "user": "c5f05561-9fbf-4ab0-b68f-e392267c0703"
  }
]

I decided to model 'friends' as a one-way relationship (like Twitter) for this example, which is why Matt Groves has Nic Raboy as a friend but not vice-versa. (Don’t read too much into this, Nic :).

The way to model this in C# could be:

public class FriendbookUser
{
    public Guid Id { get; set; }
    public string Name { get; set; }
    public virtual List<FriendbookUser> Friends { get; set; }
}

public class Update
{
    public Guid Id { get; set; }
    public DateTime PostedDate { get; set; }
    public string Body { get; set; }

    public virtual FriendbookUser User { get; set; }
    public Guid UserId { get; set; }
}

The Update to FriendbookUser relationship can be modeled as either a Guid or as another FriendbookUser object. This is an implementation detail. You might prefer one, the other, or both, depending on your application needs and/or how your OR/M works. In either case, the underlying model is the same.

Here’s the mapping I used for these classes in Entity Framework. Your mileage may vary, depending on how you use EF or other OR/M tools. Focus on the underlying model and not the details of the OR/M mapping tool.

public class UpdateMap : EntityTypeConfiguration<Update>
{
    public UpdateMap()
    {
        this.HasKey(m => m.Id);

        this.ToTable("FriendBookUpdates");
        this.Property(m => m.Body);
        this.Property(m => m.PostedDate);
        this.HasRequired(m => m.User)
            .WithMany()
            .HasForeignKey(m => m.UserId);
    }
}

public class FriendbookUserMap : EntityTypeConfiguration<FriendbookUser>
{
    public FriendbookUserMap()
    {
        this.HasKey(m => m.Id);

        this.ToTable("FriendBookUsers");
        this.Property(m => m.Name);
        this.HasMany(t => t.Friends)
            .WithMany()
            .Map(m =>
            {
                m.MapLeftKey("UserId");
                m.MapRightKey("FriendUserId");
                m.ToTable("FriendBookUsersFriends");
            });
    }
}

If, instead of storing these entities as separate documents, we applied the same denormalization as the shopping cart example and attempted to store a user and updates in one document, we would end up with some problems.

Note: There’s an N+1 problem here too (friends of friends, etc), but I’m not going to spend time on addressing that. It’s a problem that’s not unique to either database.

Additionally, it may not be the case that when the application reads or writes a user that it will need to read or write friends & updates. And, when writing an update, it’s not likely that the application will need to update a user. Since these entities may often be read/written on their own, that indicates that they need to be modeled as separate documents.

Note the array in the Friends field in the user document and the value in the User field in the update document. These values can be used to retrieve the associated documents. Later in this post, I’ll discuss how to do it with key/value operations and how to do it with N1QL.

To sum up, there are two ways to model data in a document database. The shopping cart example used nested objects, while the social media example used separate documents. In those examples, it was relatively straightforward to choose. When you’re making your own modeling decisions, here’s a handy cheat sheet:

public List<Update> GetTenLatestUpdates()
{
    var n1ql = @"SELECT up.body, up.postedDate, { 'id': META(u).id, u.name} AS `user`
        FROM `sqltocb` up
        JOIN `sqltocb` u ON KEYS up.`user`
        WHERE up.type = 'Update'
        ORDER BY STR_TO_MILLIS(up.postedDate) DESC
        LIMIT 10;";
    var query = QueryRequest.Create(n1ql);
    query.ScanConsistency(ScanConsistency.RequestPlus);
    var result = _bucket.Query<Update>(query);
    return result.Rows;
}

[
  {
    "body": "Autem occaecati quam vel. In aspernatur dolorum.",
    "postedDate": "2017-02-02T16:19:48.7812386-05:00",
    "user": {
      "id": "c5f05561-9fbf-4ab0-b68f-e392267c0703",
      "name": "Bob Johnson"
    }
  },
  {
    "body": "Nostrum eligendi aspernatur enim repellat culpa eligendi maiores et.",
    "postedDate": "2017-02-02T16:19:45.2792288-05:00",
    "user": {
      "id": "7fc5503f-2092-4bac-8c33-65ef5b388f4b",
      "name": "Steve Oberbrunner"
    }
  },

  // ... etc ...
]

CREATE INDEX IX_Name ON `SocialMedia` (Name) USING GSI;

In relational databases, rows of data (typically, not always) correspond to a primary key, which is often an integer or a Guid, and sometimes a composite key. These keys don’t necessarily have any meaning: they are just used to identify a row within a table. For instance, two rows of data in two different tables may have the same key (an integer value of 123, for instance), but that doesn’t necessarily mean the data is related. This is because the schema enforced by relational databases often conveys meaning on its own (e.g. a table name).

In document databases like Couchbase, there isn’t anything equivalent to a table, per se. Each document in a bucket must have a unique key. But a bucket can have a variety of documents in it. Therefore, it’s often wise to come up with a way to differentiate documents within a bucket.

{
    "address" : "1800 Brown Rd",
    "city" : "Groveport",
    "state" : "OH",
    "type" : "address"
}

SELECT d.*
FROM `default` d
WHERE d.type = 'address'

{
    "address" : "1800 Brown Rd",
    "city" : "Groveport",
    "state" : "OH",
    "documentInfo" : {
        "type" : "address",
        "lastUpdated" : "1/29/2017 1:31:10 PM",
        "lastUpdatedBy" : "mgroves"
    }
}

This blog post covered data modeling using denormalization, data modeling using referencing, key design, and discriminating fields. Modeling data in a document database is a thought process, something of an art form, and not a mechanical process. There is no prescription on how to model your data in a document database: it depends greatly on how your application interacts with your data.

You can get the source code for the entire blog series on GitHub now, parts of which were featured in this blog post. If you have questions about various parts of that code, feel free to leave a comment below, or open an issue on GitHub.

Stay tuned for the next blog in the series, where data and data migration will be discussed.

If you have any questions, please leave a comment below, contact me on Twitter, or use the Couchbase Forums.