Reusable schemas - Featurevisor

Recently, Featurevisor introduced more complex data structures in variables supporting nested objects, and also arrays of objects with full type safety.

Now those schemas (optionally) can be defined once separately, and referenced from anywhere by name.

TL;DR: read full API for reusable schemas here.

How variables are defined today#

Each feature maintains its own variables' schema definitions.

The types of variables supported today include:

string
boolean
integer
double
array
object
json (any valid JSON in stringified form)

Here's a simple example of a string variable

features/my_feature.yml

# ...
variablesSchema:
  my_variable:
    type: string
    defaultValue: red

Variables may also be complex data structures in the form of:

objects (flat or nested), or
arrays (of strings, objects, or more):

Here's an example of a nested object variable:

features/my_feature.yml

# ...
variablesSchema:
  my_nested_object_variable:
    type: object
    properties:
      my_property:
        type: string
      another_property:
        type: object
        properties:
          another_nested_property:
            type: string
      required:
        - my_property
        - another_property
    defaultValue:
      my_property: my value
      another_property:
        another_nested_property: nested value

Challenges with current approach#

When your application configuration grows, you may find yourself repeating the same schema definitions across multiple features, and also across multiple variables.

This also means your individual feature definitions become harder to read and maintain as they grow longer and longer.

Solution: reusable schemas#

Instead of having to inline the schema in each feature, we can optionally choose to extract them to separate files and reference them by name.

First we create a schema in schemas directory, following a simplified flavour of JSON Schema:

schemas/my_schema.yml

description: My schema
type: object
properties:
  my_property:
    type: string
  another_property:
    type: object
    properties:
      another_nested_property:
        type: string
required:
  - my_property
  - another_property

And then we can reference it from any feature:

features/my_feature.yml

# ...
variablesSchema:
  my_nested_object_variable:
    schema: my_schema
    defaultValue:
      my_property: my value
      another_property:
        another_nested_property: nested value

Advanced usage#

Several other JSON Schema inspired properties are supported, including:

Property	Applies to	Description
`required`	`object`	Define the properties that are required in the schema
`minimum`	`integer`, `double`	Define the minimum value for a particular property in the schema
`maximum`	`integer`, `double`	Define the maximum value for a particular property in the schema
`minLength`	`string`	Define the minimum length of a string
`maxLength`	`string`	Define the maximum length of a string
`pattern`	`string`	Define a regular expression pattern that the string value must match
`minItems`	`array`	Define the minimum number of items in an array
`maxItems`	`array`	Define the maximum number of items in an array
`uniqueItems`	`array`	Require that all items in an array are unique (no duplicates)
`const`	any	Define a constant value for a particular property in the schema
`enum`	any	Define a list of allowed values for a particular property in the schema
`oneOf`	any	Allow a value to match exactly one of several schemas

Learn more here.

Full linting support#

As complexity grows, and you define the schemas more precisely, Featurevisor's linting will help you find issues early if you mistakenly provided any wrong value for the variables anywhere in your feature definitions.

Command

$ npx featurevisor lint

Looking forward#

All of this paves the way for more complex variable definitions with full type safety and code generation.

More advanced use cases utilizing these new functionalities are coming soon as guides.

Fun fact: if I didn't build Eventvisor last summer, I probably wouldn't think of bringing schemas into Featurevisor like how it's done here now.

How variables are defined today#

Challenges with current approach#

Solution: reusable schemas#

Advanced usage#

Full linting support#

Looking forward#

Further reading#