1 Overview of JSON-Relational Duality Views

Duality views combine the advantages of using JSON documents with those of the relational model, while avoiding the limitations of each. JSON-relational duality underpins collections of documents with relational storage: active, updatable, hierarchical documents are based on a foundation of normalized relations.

  • A single JSON document can represent an application object directly, capturing the hierarchical relations among its components. A JSON document is standalone: self-contained and self-describing — no outside references, no need to consult an outside schema. There's no decomposition, which means that JSON is schema-flexible: you can easily add and remove fields, and change their type, as required by application changes.

    However, relationships among documents are not represented by the documents themselves; the application must code relationships separately, as part of its logic. In particular, values that are part of one document cannot be shared by others. This leads to data duplication across different documents (whether of the same kind or different kinds), which in turn can introduce inconsistencies when documents are updated.

  • The relational model decomposes application objects ("business objects") into normalized tables, which are explicitly related but whose content is otherwise independent. This independence allows for flexible and efficient data combination (joining) that is rigorously correct and reliable.

    This avoids inconsistencies and other problems with data duplication, but it burdens application developers with defining a mapping between their application objects and relational tables. Application changes can require schema changes to tables, which can hinder agile development. As a result, developers often prefer to work with document-centric applications.

A JSON-relational duality view exposes data stored in relational database tables as JSON documents. The documents are materialized — generated on demand, not stored as such. Duality views give your data both a conceptual and an operational duality: it's organized both relationally and hierarchically. You can base different duality views on data that's stored in one or more of the same tables, providing different JSON hierarchies over the same, shared data.

This means that applications can access (create, query, modify) the same data as a set of JSON documents or as a set of related tables and columns, and both approaches can be employed at the same time.

  • Document-centric applications can use document APIs, such as Oracle Database API for MongoDB and Oracle REST Data Services (ORDS), or they can use SQL/JSONFoot 1 functions. You can manipulate documents realized by duality views in the ways you're used to, using your usual drivers, frameworks, tools, and development methods. In particular, applications can use any programming languages — JSON documents are the lingua franca.

  • Other applications, such as database analytics, reporting, and machine learning, can make use of the same data directly, relationally (as a set of table rows and columns), using languages such as SQL, PL/SQL, C, and JavaScript. You need not adapt an existing database feature or code that makes use of table data to instead use JSON documents.

A JSON-relational duality view directly defines and reflects the structure of JSON documents of a given kind (structure and field types). The view is based on underlying database tables, which it joins automatically to realize documents of that kind.

Columns of SQL data types other than JSON in an underlying table produce scalar JSON values in the documents supported by the view. Columns of the SQL data type JSON can produce JSON values of any kind (scalar, object, or array) in the documents, and the JSON data can be schemaless or JSON Schema-based (to enforce particular document shapes and field types). See Car-Racing Example, Tables for the column data types allowed in a table underlying a duality view.

JSON fields produced from an underlying table can be included in any JSON objects in a duality-view document. When you define the view you specify where to include them, and whether to do so individually or to nest them in their own object. By default, nested objects are used.Foot 2

Note:

A given column in an underlying table can be used to support fields in different objects of a document. In that case, the same column value is used in each object: the data is shared.Foot 3

A duality view can be read-only or completely or partially updatable, depending on how you define it. You can define a duality view and its updatability declaratively (what/where, not how), using SQL or a subset of the GraphQL language.

When you modify a duality view — to insert, delete, or update JSON documents, the relevant relational (table) data underlying the view is automatically updated accordingly.

We say that a duality view supports a set of JSON documents of a particular kind (structure and typing), to indicate both (1) that the documents are generated — not stored as such — and (2) that updates to the underlying table data are likewise automatically reflected in the documents.

Even though a set of documents (supported by the same or different duality views) might be interrelated because of shared data, an application can simply read a document, modify it, and write it back. The database detects the document changes and makes the necessary modifications to all underlying table rows. When any of those rows underlie other duality views, those other views and the documents they support automatically reflect the changes as well.

Conversely, if you modify data in tables that underlie one or more duality views then those changes are automatically and immediately reflected in the documents supported by those views.

The data is the same; there are just dual ways to view/access it.

Duality views give you both document advantages and relational advantages:

  • Document: Straightforward application development (programming-object mappings, get/put access, common interchange format)

  • Relational: Consistency, space efficiency, normalization (flexible data combination/composition/aggregation)

See Also:

1.1 The Use Case for JSON-Relational Duality Views

The motivation behind JSON-relational duality views is presented.

Suppose the following:

  • You have, or you will develop, one or more applications that are document-centric; that is, they use JSON documents as their primary data. For the most part, you want your applications to be able to manipulate (query, update) documents in the ways you're used to, using your usual drivers, frameworks, tools, development methods, and programming languages.

  • You want the basic structure of the various kinds of JSON documents your application uses to remain relatively stable.

  • Some kinds of JSON documents that you use, although of different overall structure, have some parts that are the same. These documents, although hierarchical (trees), are interrelated by some common parts. Separately each is a tree, but together they constitute a graph.

  • You want your applications to be able to take advantage of all of the advanced processing, high performance, and security features offered by Oracle Database.

In such a case you can benefit from defining and storing your application data using Oracle Database JSON-relational duality views. You can likely benefit in other cases, as well — for example, cases where only some of these conditions apply. As a prime motivation behind the introduction of duality views, this case helps present the various advantages they have to offer.

Shared Data

An important part of the duality-view use case is that there are some parts of different JSON documents that you want to remain the same. Duplicating data that should always be the same is not only a waste. It ultimately presents a nightmare for application maintenance and evolution. It requires your application to keep the common parts synced.

The unspoken problem presented by document-centric applications is that a JSON document is only hierarchical. And no single hierarchy fits the bill for everything, even for the same application.

Consider a scheduling application involving students, teachers, and courses. A student document contains information about the courses the student is enrolled in. A teacher document contains information about the courses the teacher teaches. A course document contains information about the students enrolled in the course. The problem is that the same information is present in multiple kinds of documents, in the same or different forms. And it's left to applications that use these documents to manage this inherent sharing.

With duality views these parts can be automatically shared, instead of being duplicated. Only what you want to be shared is shared. An update to such shared data is reflected everywhere it's used. This gives you the best of both worlds: the world of hierarchical documents and the world of related and shared data.

There's no reason your application should itself need to manage whatever other constraints and relations are required among various parts of different documents. Oracle Database can handle that for you. You can specify that information once and for all, declaratively.

Here's an example of different kinds of JSON documents that share some parts. This example of car-racing information is used throughout this documentation.

  • A driver document records information about a particular race-car driver: driver name; team name; racing points earned; and a list of races participated in, with the race name and the driver position.

  • A race document records information about a particular race: its name, number of laps, date, podium standings (top three drivers), and a list of the drivers who participated, with their positions.

  • A team document records information about a racing team: its name, points earned, and a list of its drivers.

Stable Data Structure and Types

Another important part of the duality-view use case is that the basic structure and field types of your JSON documents should respect their definitions and remain relatively stable.

Duality views enforce this stability automatically. They do so by being based on normalized tables, that is, tables whose content is independent of each other (but which may be related to each other).

You can define just which document parts need to respect your document design in this way, and which parts need not. Parts that need not have such stable structure and typing can provide document and application flexibility: their underlying data is of Oracle SQL data type JSON (native binary JSON).

No restrictions are imposed on these pliable parts by the duality view. (But because they are of JSON data type they are necessarily well-formed JSON data.) The data isn't structured or typed according to the tables underlying the duality view. But you can impose any number of structure or type restrictions on it separately, using JSON Schema (see below).

An example of incorporating stored JSON-type data directly into a duality view, as part of its definition, is column podium of the race table that underlies part of the race_dv duality view used in the Formula 1 car-racing example in this documentation.Foot 4

Like any other column, a JSON-type column can be shared among duality views, and thus shared among different kinds of JSON documents. (Column podium is not shared; it is used only for race documents.) See JSON Data Stored in JSON-Relational Duality Views for information about storing JSON-type columns in tables that underlie a duality view.

JSON data can be totally schemaless, with structure and typing that's unknown or susceptible to frequent change. Or you can impose a degree of definition on it by requiring it to conform to a particular JSON schema. A JSON schema is a JSON document that describes other JSON documents. Using JSON Schema you can define and control the degree to which your documents and your application are flexible.

Being based on database tables, duality views themselves of course enforce a particular kind of structural and typing stability: tables are normalized, and they store a particular number of columns, which are each of a particular SQL data type. But you can use JSON Schema to enforce detailed document shape and type integrity in any number of ways on a JSON-type column — ways that are specific to the JSON language.

Because a duality view definition imposes some structure and field typing on the documents it supports, it implicitly defines a JSON schema. This schema is a description of the documents that reflects only what the duality view itself prescribes. It is available in column JSON_SCHEMA of static dictionary views DBA_JSON_DUALITY_VIEWS, USER_JSON_DUALITY_VIEWS, and ALL_JSON_DUALITY_VIEWS. You can also see the schema using PL/SQL function DBMS_JSON_SCHEMA.describe.

Duality views compose separate pieces of data by way of their defined relations. They give you precise control over data sharing, by basing JSON documents on tables whose data is separate from but related to that in other tables.

Both normalizing and JSON Schema-constraining make data less flexible, which is sometimes what you want (stable document shape and field types) and sometimes not what you want.

Oracle Database provides a full spectrum of flexibility and control for the use of JSON documents. Duality views can incorporate JSON-type columns to provide documents with parts that are flexible: not normalized and (by default) not JSON Schema-constrained. See JSON Data Stored in JSON-Relational Duality Views for information about controlling the schema flexibility of duality views.

Your applications can also use whole JSON documents that are stored as a column of JSON data type, not generated by a duality view. Applications can interact in exactly the same ways with data in a JSON column and data in a duality view — in each case you have a set of JSON documents.

Those ways of interacting with your JSON data include (1) document-store programming using document APIs such as Oracle Database API for MongoDB and Oracle REST Data Services (ORDS), and (2) SQL/JSON programming using SQL, PL/SQL, C, or JavaScript.

JSON-relational duality views are special JSON collection views. Ordinary (non-duality) JSON collection views are not updatable. JSON collection views, along with JSON collection tables (which are updatable), are JSON collections. You can use a JSON collection directly with a document API. In particular, the documents in duality views and the documents in JSON collection tables can have the same form and are thus be interchangeable.

This means, for example, that you could start developing an application using a JSON collection table, storing your JSON documents persistently and with no schema, and later, when your app is stable, switch transparently to using a JSON-relational duality view as the collection instead. Your application code accessing the collection can remain the same — same updates, insertions, deletions, and queries. (This assumes that the documents stored in the collection table have the same shape as those supported by the duality view.)

JSON duality views are listed in these static dictionary views, in order of decreasing specificity: *_JSON_COLLECTIONS, *_VIEWS, and *_OBJECTS.

Enforcing structural and type stability means defining what that means for your particular application. This isn't hard to do. You just need to identify (1) the parts of your different documents that you want to be truly common, that is, to be shared, (2) what the data types of those shared parts must be, and (3) what kind of updating, if any, they're allowed. Specifying this is what it means to define a JSON-relational duality view.

Existing relational data has already undergone data analysis and factoring, so it's straightforward to define duality views that are based on any existing relational data. This means it's easy to adapt or define a document-centric application that reuses existing relational data as a set of JSON documents. This alone is a considerable advantage of the duality between relational and JSON data. The wide world of relational data is available to you as sets of JSON documents.

See Also:

1.2 Map JSON Documents, Not Programming Objects

A JSON-relational duality view declaratively defines a mapping between JSON documents and relational data. That's better than mapping programming objects to relational data.

If you use an object-relational mapper (ORM) or an object-document mapper (ODM), or you're familiar with their concepts, then this topic might help you better understand the duality-view approach to handling the "object-relational impedance mismatch" problem.

Duality views could be said to be a kind of ORM: they too map hierarchical object data to/from relational data. But they're fundamentally different from existing ORM approaches.

Duality views centralize the persistence format of application objects for both server-side and client-side applications — all clients, regardless of language or framework. The persistence model presents two aspects for the same data: table and document. Server-side code can manipulate relational data in tables; client-side code can manipulate a collection (set) of documents.

Client code need only convert its programming objects to/from JSON, which is familiar and easy. A duality view automatically persists JSON as relational data. There's no need for any separate mapper — the duality view is the mapping.

The main points in this regard are these:

  • Map JSON documents; don't map programming objects!

    With duality views, the only objects you map to relational data are JSON documents. You could say that a duality view is a document-relational mapping (DRM), or a JSON-relational mapping (JRM).

    A duality view doesn't lock you into using, or adapting to, any particular language (for mapping or for application programming). It's just JSON documents, all the way down (and up and around). And it's all relational data — same dual thing!

  • Map declaratively!

    A duality view is a mapping — there's no need for a mapper. You define duality views as declarative maps between JSON documents and relational tables. That's all. No procedural programming.

  • Map inside the database!

    A duality view is a database object. There's no tool-generated SQL code to tune. Application operations on documents are optimally executed inside the database.

    No separate mapping language or tools, no programming, no deploying, no configuring, no setting-up anything. Everything about the mapping itself is available to any database feature and any application — a duality view is just a special kind of database view.

    This also means fewer round trips between application and database, supporting read consistency and providing better performance.

  • Define rules for handling parts of documents declaratively, not in application code.

    Duality views define which document parts are shared, and whether and how they can be updated. The same rule validation/enforcement is performed, automatically, regardless of which application or language requests an update.

  • Use any programming language or tool to access and act on your documents — anything you like. Use the same documents with different applications, in different programming languages, in different ways,….

  • Share the same data in multiple kinds of documents.

    Create a new duality view anytime, to combine things from different tables. Consistency is maintained automatically. No database downtime, no compilation,.... The new view just works (immediately), and so do already existing views and apps. Duality views are independent, even when parts of their supported documents are interdependent (shared).

  • Use lockless/optimistic concurrency control.

    No need to lock data and send multiple SQL statements, to ensure transactional semantics for what's really a single application operation. (There's no generated SQL to send to the database.)

A duality view maps parts of one or more tables to JSON documents that the view defines — it need not map every column of a table. Documents depend directly on the mapping (duality view), and only indirectly on the underlying tables. This is part of the duality: presenting two different views — not only views of different things (tables, documents) but typically of somewhat different content. Content-wise, a document combines subsets of table data.

This separation/abstraction is seen clearly in the fact that not all columns of a table underlying a duality view need be mapped to its supported documents. But it also means that some changes to an underlying table, such as the addition of a column, are automatically prevented from affecting existing documents, simply by the mapping (view definition) not reflecting those changes. This form of table-level schema evolution requires no changes to existing duality views, documents, or applications.

On the other hand, if you want to update an application, to reflect some table-level changes, then you change the view definition to take those changes into account in whatever way you like. This application behavior change can be limited to documents that are created after the view-definition change.

Alternatively, you can create a new duality view that directly reflects the changed table definitions. You can use that view with newer versions of the application while continuing to use the older view with older versions of the app. This way, you can avoid having to upgrade all clients at the same time, limiting downtime.

In this case, schema evolution for underlying tables leads to schema evolution for the supported documents. An example of this might be the deletion of a table column that's mapped to a document field. This would likely lead to a change in application logic and document definition.

1.3 Duality-View Security: Simple, Centralized, Use-Case-Specific

Duality views give you better data security. You can control access and operations at any level.

Security control is centralized. Like everything else about duality views, it is defined, verified, enforced, and audited in the database. This contrasts strongly with trying to secure your data in each application. You control access to the documents supported by a duality-view the same way you control access to other database objects: using privileges, grants, and roles.

Duality-view security is use-case-specific. Instead of according broad visibility at the table level, a duality view exposes only relevant columns of data from its underlying tables. For example, an application that has access to a teacher view, which contains some student data, won't have access to private student data, such as social-security number or address.

Beyond exposure/visibility, a duality view can declaratively define which data can be updated, in which ways. A student view could allow a student name to be changed, while a teacher view would not allow that. A teacher-facing application could be able to change a course name, but a student-facing application would not. See Updatable JSON-Relational Duality Views and Updating Documents/Data in Duality Views.

You can combine the two kinds of security control, to control who/what can do what to which fields:

  • Create similar duality views that expose slightly different sets of columns as document fields. That is, define views intended for different groups of actors. (The documents supported by a duality view are not stored as such, so this has no extra cost.)

  • Grant privileges and roles, to selectively let different groups of users/apps access different views.

Contrast this declarative, in-database, field-level access control with having to somehow — with application code or using an object-relational mapper (ORM) — prevent a user or application from being able to access and update all data in a given table or set of documents.

The database automatically detects document changes, and updates only the relevant table rows. And conversely, table updates are automatically reflected in the documents they underlie. There's no mapping layer outside the database, no ORM intermediary to call upon to remap anything.

And client applications can use JSON documents directly. There's no need for a mapper to connect application objects and classes to documents and document types.

Multiple applications can also update documents or their underlying tables concurrently. Changes to either are transparently and immediately reflected in the other. In particular, existing SQL tools can update table rows at the same time applications update documents based on those rows. Document-level consistency, and table row-level consistency, are guaranteed together.

And this secure concurrency can be lock-free, and thus highly performant. See Using Optimistic Concurrency Control With Duality Views.

Particular Oracle Database security features that you can use JSON-relational duality views with include Transparent Data Encryption (TDE), Data Redaction, and Virtual Private Database.

1.4 Oracle Database: Converged, Multitenant, Backed By SQL

If you use JSON-relational duality views then your application can take advantage of the benefits of a converged database.

These benefits include the following:

  • Native (binary) support of JavaScript Object Notation (JSON) data. This includes updating, indexing, declarative querying, generating, and views

  • Advanced security, including auditing and fine-grained access control using roles and grants

  • Fully ACID (atomicity, consistency, isolation, durability) transactions across multiple documents and tables

  • Standardized, straightforward JOINs with all sorts of data (including JSON)

  • State-of-the-art analytics, machine-learning, and reporting

Oracle Database is a converged, multimodel database. It acts like different kinds of databases rolled into one, providing synergy across very different features, supporting different workloads and data models.

Oracle Database is polyglot. You can seamlessly join and manipulate together data of all kinds, including JSON data, using multiple application languages.

Oracle Database is multitenant. You can have both consolidation and isolation, for different teams and purposes. You get a single, common approach for security, upgrades, patching, and maintenance. (If you use an Autonomous Oracle Database, such as Autonomous JSON Database, then Oracle takes care of all such database administration responsibilities. An autonomous database is self-managing, self-securing, self-repairing, and serverless. And there's Always Free access to an autonomous database.)

The standard, declarative language SQL underlies processing on Oracle Database. You might develop your application using a popular application-development language together with an API such as Oracle Database API for MongoDB or Oracle REST Data Services (ORDS), but the power of SQL is behind it all, and that lets your app play well with everything else on Oracle Database.



Footnote Legend

Footnote 1: SQL/JSON is specified in ISO/IEC 9075-2:2016, Information technology—Database languages—SQL— Part 2: Foundation (SQL/Foundation). Oracle SQL/JSON support is closely aligned with the JSON support in this SQL Standard.
Footnote 2: You use keyword UNNEST in the SQL view definition, or directive @unnest in the GraphQL view definition, to include fields directly. See Creating Duality Views.
Footnote 3: For a flex column (see Flex Columns: Duality-View Schema Flexibility and Evolution), the column value provides field names as well as values. In that case, both are shared across the different objects supported by the column.
Footnote 4: See Example 2-4 and Example 3-5.