Data Division

A library that generates data-holding structures with validation, error maps, and compatibility with Phoenix form_for.

Let's Get Started

1. Add the dependency:

   def deps() do
     [
       :data_division, ">= 0.0.0",
       . . .
    ]

2. Define the fields in the structure:

    defmodule Planet do
      use DD
   
      deffieldset do
        string :name,      min: 4
        float  :mass
        bool   :habitable, default: false
        int    :moon_count
      end
    end

3. Create and populate structures based on this structure:

   neptune = Planet.new(
       name:       "Neptune", 
       mass:       1.024e26, 
       moon_count: 14)

4. They found another moon:

   new_neptune = neptune |> Planet.update(moon_count: 15)

5. Is the structure valid? If not, what are the errors?

   if !Planet.valid?(neptune) do
     for { field, error_msg } <- neptune.errors do
       IO.puts "#{field}: #{error_msg}"
     end
   end

6. Let's use it in Phoenix:

   controller action:

   render conn, "edit.html", planet: neptune

   template:

   <%= form_for @planet, ..... 

Why?

If you want to use Phoenix forms to create and update data accessed using Ecto, then you have to have ecto running in the Phoenix application.

That's not how I want to design applications. I want my business logic in its own applications, and I want Phoenix to be one of potentially many frontends to it.

For example, an online store might perform overnight accounting functions. These functions will want to use the same business logic used by the store (for example, to access orders), but it doesn't have a web UI. The accounting code should be able to make the same calls into the business layer as the frontend, but without having the frontend in the middle.

The Data Division library lets you do this. It lets you define things that are like Ecto schemas, and populate from from an Ecto changeset.

You can then pass the result to other applications that are not running Ecto. Because a Data Division fieldset implements the Phoenix Form.Data protocol, you can use them with Phoenix forms.

The net result is that you can decouple presentation from business logic.

deffieldset

deffieldset is a bit like Ecto's schema. It defines a struct that contains a place for data (like the planet information above) and a place for metadata on field types, options, and so on.

It should be used in a module, just like defstruct:

    defmodule Planet do
      use DD

      deffieldset do
        string :name,      min: 4
        float  :mass,      min: 0.0
        bool   :habitable, default: false
        int    :moon_count
      end
    end

This code defines a structure called Planet with 4 fields. Each field definition starts with the field type, followed by the field name (an atom). The rest depends to some extent on the type of the field, although all fields support default values.

In this example, the name field has a validation: it must be at least 4 characters long. Similarly the mass field has a validation: it cannot be less that zero. Note that although the option is named the same in both cases min:) its interpretation depends on the field type: for strings it is the length, for floats the value.

A list of the available types and their options is below.

Using the Fields

You create new instances of a structure using Name.new(values). The values you pass in can be a keyword list, map, or struct. If values is a struct of type Ecto.Changeset, then values and errors are copied directly from it into the record.

new returns a structure containing three entries:

• values

  A map, keyed by the field name, containing that fields current value.

• errors

  A map, keyed by the field name. If an entry exists for a field, its value is the first validation error associated with that field.

• fields

  A reference to a field definition structure

So, we could do something like:

   neptune = Planet.new(
       name: "Neptune", 
       moon_count: 14)

   IO.inspect neptune.errors     #=> %{ mass: "must be present" }

   neptune = Planet.update(neptune, mass: 1.024e26

   IO.inspect neptune.errors     #=> %{ }
   IO.inspect neptune.valid?     #=> true

Built-in Types

• All types accept the options:

  • default:a type appropriate value

  • optional: _defaults to true unless a default was provided

    In this world, an optional field is one that may have a nil value. Think of it as corresponding to the database not null constraint.

  • validate_with:a Module or a function (or a list of them)

    See Custom Validations below.

    Note that validation for a particular field stops on the first validation failure—once a field has been found to be invalid, no more checks are done on that particular field.

• (string)[...]:name

  Options:

  • min:min_length
  • max:max_length
  • matches:string or regex

  Conversion:

  • incoming values: nonstrings are converted using inspect
  • outgoing values: none

• (int)[...]:name

  Options:

  • min:min_value
  • max:max_value

  Conversion:

  • incoming values: incoming strings are converted using String.to_integer
  • outgoing values: none

• (float)[...]:name

  Options:

  • min:min_value
  • max:max_value

  Conversion:

  • incoming values: incoming strings are converted using String.to_float or String.to_integer. Incoming integers are converted by adding 0.0
  • outgoing values: none

• (bool)[...]:name

  Options:

  • show_as{ true-values, false-values }_

    true-values is a string or a list of strings. It this field is set to one of these, it will have an internal value of true. false-values work the same way for false.

    If not specified, show_as: defaults to [ "true", "false" ].

  Conversion:

  • incoming values: incoming strings are converted using show_as: to either true or false.
  • outgoing values: the first true or false value in ahow_as: is used.

• (id)[...]:name

  Options:

  • none

  Conversion:

  • none

A field of type id is byu default both hidden and optional. It is recognized as the primary key when generating Phoenix forms.

• etc

Custom Validations

You can define your own field validators. Each is a function that takes a value and returns either

nil

if the value is valid, or

{ msg_with_placeholders, p1: v1, … }

if it is invalid. In the latter case the first field in the returned tuple is a string containing optional placeholders. The values that are to be substituted for each placeholder are given in the subsequent keyword list:

{ "%{value} must have exactly %{n} factors", value: 30, n: 2 }

Add a custom validator to a field using the validate_with: option. This takes either a single validator or a list of validators.

If a validator is the name of a module, then the field is validted by calling the function validate/1 in that module.

If a validator is a function, then it is called withe the value to validate.

For example:

# This is a validation module
defmodule EvenValidator do
  require Integer
  def validate(value) when Integer.is_even(value), do: nil
  def validate(_), do: { "must be even..", [] }
end

# and this module contains validation functions
defmodule Validations do
  require Integer
  def is_even(value) when Integer.is_even(value), do: nil
  def is_even(_), do: { "must be even!!", [] }
end


defmodule A do
  use DD
  deffieldset do
    int(:even1, default: 2, validate_with: EvenValidator)
    int(:even2, default: 2, validate_with: &Validations.is_even/1)
  end
end

Adding Your Own Types

A type is simply an Elixir module that:

1. is named DD.Type.YourType

2. uses the behaviour DD.Type.Behaviour

3. implements the handful of functions required by that behaviour.

If this module is loaded into your project, then the type becomes available in deffieldset as if it was a function named using the lowercase form of the last part of the module name. So, in this example, you could have

deffieldset do

 string(:name)
 your_type(:orbit_parameters)

end

See the module doc for Data.Type for details.

Working with Ecto

The MyFieldset.new/1 function normally takes a map, struct, or keyword list containing key value pairs. It copies the values into the fieldset, then performs validations and sets any errors.

The Repo functions return a struct, so DD will copy in any values from this that it has a field definition for:

def find_by_id(id) do
  Repo.get(Table, id)
  |> Notebook.Fields.new
end

You can also pass it an Ecto changeset. In this case, it copies the values from the changeset data, and copies the errors from the changeset errors.

ToDo: add support for deffieldset based_on My.Schema