Remove NameError rescue in page base

[ndrluis]

Fixes #480

[ndrluis]

The rescue statement generates a false positive on payments feature specs, the timestamps migrations are created to fix this.

[nickcharlton]

This looks good! Could I get you to fix the Hound comment above?

[pablobm]

Stale :-( Happy to see this reopen if someone wants to take over.

[pablobm]

Actually, I thought better of this one. I'm going to try see it through.

[wkirby]

This PR appears to break any support for virtual fields by enforcing the the resource must define a method to respond to all keys in the Dashboard class. Maybe this is the desired behavior, but the following use case was possible in 0.12.0, and was very useful:

class ReportDashboard < Administrate::BaseDashboard
  ATTRIBUTE_TYPES = {
    id: Field::Number,
    filename: Field::String,
    mimetype: Field::String,
    s3_key: Field::String,
    download: S3DownloadField.with_options({ download_path: :download_lab_report_path }),
    page_count: Field::Number,
    created_at: Field::DateTime,
    updated_at: Field::DateTime
  }.freeze

  COLLECTION_ATTRIBUTES = [
    :id,
    :filename,
    :download,
    :mimetype,
    :created_at
  ].freeze

  SHOW_PAGE_ATTRIBUTES = [
    :id,
    :download,
    :filename,
    :mimetype,
    :s3_key,
    :page_count,
    :created_at,
    :updated_at
  ].freeze

  FORM_ATTRIBUTES = [
    :lab,
    :filename,
    :mimetype,
    :s3_key,
    :page_count
  ].freeze
end

class S3DownloadField < Administrate::Field::Base
  def to_s
    data
  end

  def downloadable?
    options.key?(:download_path)
  end

  def destroyable?
    options.key?(:destroy_path)
  end

  def destroy_path(field, attachment)
    path_helper(:destroy_path, field, attachment)
  end

  def download_path(field, attachment)
    path_helper(:download_path, field, attachment)
  end

  private

  def path_helper(key, field, _attachment)
    path_helper = options.fetch(key)
    Rails.application.routes.url_helpers.send(path_helper, field.resource)
  end
end

This allowed us to inject a download link for our S3 resource without editing the underlying views, or modifying the display for any of the actual attributes.

Is there any advice on how to fix this without defining useless methods on our models?

[sedubois]

@wkirby I also filed regressions from this PR here:
#1570

[pablobm]

@wkirby - Thank you for your report. I have filed a separate issue at #1586