Specification

Validation messages reside in the separate schema, decoupled from the validation logic. The reasons for the decoupling are the following:

• Establish a clear separation of concerns.

• Prevent from garbaging complex validation rules with the messages.

• Have a single validation schema associated with any amount of localized messages schema.

Resolver

Any message can be a plain string, or a resolver function that returns a string.

Resolvers are useful when a validation message derives not only from the field's validity, but also from its state, fields or a form. This way you can include the current value of a field in the error message, for example.

Definition

type MessageResolver = string | (params) => string

Parameters

Message types

Validation messages schema supports the following message types:

Selectors

Selectors are used to associate message resolvers to the relevant field(s).

Each selector expects the map of the selector values and its resolvers:

type ValidationMessages = {
    [selectorValue: string]: MessageResolver,
  },
}

Named resolvers

It is possible to associate validation rules with the specific validation messages. To do that, provide the { ruleName: message } map as the value of the rule key in the respective selector:

export default {
  type: {
    password: {
      missing: 'Please provide the password',
      invalid: 'The passwords is invalid',
      rule: {
        minLength: 'Password must be at least 6 characters long',
      },
    },
  },
}

Note: Named resolver must have the corresponding validation rule with the same name in order to resolve. Otherwise, the closest validation message will be returned by the resolver. Considering the example above, if minLength rule doesn't exist and the type=["password"] field is invalid, the closest invalid resolver will be used (which is type.password.invalid in this case).

Fallbacks

Whenever the resolver cannot resolve it would attempt to grab the closest resolver recursively and resolve it instead. Fallback resolvers are iterated by their specificity, which can be presented as follows:

Fallback sequence

Here is the resolver paths (listed by priority, from highest to lowest):

1. messages.name[fieldName][ruleName]

2. messages.name[fieldName].invalid / messages.name[fieldName].missing

3. messages.type[fieldType][ruleName]

4. messages.type[fieldType]invalid / messages.type[fieldType].missing

5. messages.general.invalid / messages.general.missing

Example

Consider the following validation messages:

export default {
  general: {
    invalid: 'General invalid message',
  },
  type: {
    email: {
      invalid: 'E-mail is invalid',
    },
  },
  name: {
    userEmail: {
      invalid: 'User e-mail is invalid',
      rule: {
        includesAt: 'E-mail must include "@" character',
      },
    },
  },
}

And the following component layout:

<Form>
  <Input
    type="email"
    name="userEmail"
    value="foo" />
</Form>

With the given scenario the includesAt validation rule would reject, marking the field as unexpected. This must be reflected in the UI using the validation messages schema.

This is the priority sequence in which resolvers will attempt to resolve the validation message:

1. name.userEmail.rule.includesAt

2. name.userEmail.invalid

3. type.email.invalid

4. general.invalid

The validation message resolves as soon as the resolver returns the value. The same sequence applies for the type-specific named resolvers.

​