Bulk Operations

When dealing with truly large amounts of data, you may want to skip ActiveRecord instantiation — its overhead is usually minimal for most Web access patterns, but can be very significant when loading many thousands of records.

flex_columns lets you create and save flex-column objects without requiring an instance of the enclosing ActiveRecord model to be present:

class User < ActiveRecord::Base
  flex_column :details do
    field :max_emails_per_week
    field :background_color
  end
end

raw_data = User.connection.select_all("SELECT id, details FROM users")
json_blobs = raw_data.map { |r| r['details'] }
flex_objects = User.create_flex_objects_from(:details, json_blobs)

flex_objects.each do |flex_object|
  puts "max_emails_per_week: #{flex_object.max_emails_per_week}"
  flex_object.max_emails_per_week += 1
end

new_json_blobs = flex_objects.map(&:to_stored_data)

The method #create_flex_objects_from accepts the name of the flex column for which you want to create objects, and an array of Strings that should be valid data for that column; it returns an array of flex objects for that column, in the same order. (#create_flex_object_from is identical, but takes a single String, rather than an array.) The strings can be JSON strings, or, if the column is binary, flex_columns binary storage, compressed or uncompressed for that column.

The returned flex objects have a #to_stored_data method, which will return the proper format for storage in the database. They also have a #to_json method, which will always return pure, uncompressed JSON without a header, if you want that instead.

Flex objects do not deserialize their JSON until required, so they maintain as high performance as possible — if you don't read or write any of their fields, you will not pay this performance penalty. (The #touch! method that each flex-column object has will force it to deserialize internally, if you just want to validate the object or delete unknown fields from it.)