Kellogg Associates: Category Ruby on Rails

ActiveWarehouse

gregg@kellogg-assoc.com (Gregg Kellogg) — Sat, 19 May 2007 18:10:00 -0500

Related to the interesting talk at a recent NBRug meeting on Ruport, I’ve been looking at doing data warehousing in Ruby. Fortunately, Anthony Eden has beaten me to the punch with ActiveWarehouse.

For those unfamiliar with Data Warehousing, the concepts are basically to create a series of facts that are indexed buy multiple dimensions. A fact is typically an integer (eg, sales amount) with data relating to the fact expressed in dimensions. (Note that in some cases, you can have a Factless Fact table where the information is entirely in the dimensions). Dimensions provide different types of data relating to the fact (e.g., date & time of entry, user, product, location, etc.), so a fact table has a column for each ’’dimension’’ with a column for the fact scalar value itself.

The ’’dimension’’ tables contain an id index column, a column for the fact value (e.g., timestamp for date & time) and columns for each ’’summary’’ to be associated with the dimension (e.g., day of week, day of month, month, year, hour minute, ...).

At this point, queries can be performed by joining the dimension and fact tables and summarizing (or counting) the scalar fact value against conditions against the dimension tables (e.g., sales per year by person and region).

Typically, this data is pre-summarized in ’’cubes’’ with summary tables that contain the results of these summary queries.

The ActiveWarehouse plugin contains everything necessary to define, populate and report on this data.

I expect to be using this to hold information on user listening habits for MP3 files.

Gregg Kellogg

RailsConf Sessions

gregg@kellogg-assoc.com (Gregg Kellogg) — Thu, 26 Apr 2007 15:05:00 -0500

Here are the sessions Ill be attending at RailsConf. Hope to see you there.

Gregg

Observers in Rails 1.2

gregg@kellogg-assoc.com (Gregg Kellogg) — Fri, 26 Jan 2007 21:13:00 -0600

Thanks to Rick Olson for providing assistance with the proper syntax for instantiating observers in Rails 1.2. The old way was to add an observer statement to the appropriate controler, such as:

observer :user_observer

Hidden in the ignored parts of conf/environment.rb was the proper notation:

# Activate observers that should always be running
# config.active_record.observers = :cacher, :garbage_collector

This makes the last of my deprecation messages go away!

Gregg

Eager Finder SQL

gregg@kellogg-assoc.com (Gregg Kellogg) — Sun, 05 Nov 2006 19:31:00 -0600

EagerFinderSql allows custom SQL to be specified when doing eager loading of associations through the :include option to find. This allows for purpose-constructed queries to be used and still result in a fully linked object model.

Background

ActiveRecord constructs SQL to satisfy the requirements of a find request. Associations allow for customized SQL to be specified, using the :finder_sql option, but this has not been available when performing eager loading using the :include option. The result is that a standardized query is constructed to bring in the associated tables using LEFT OUTER JOIN. For some queries, this can be result in expensive queries and potentially very large result sets.

Custom SQL

EagerFinderSql addresses this problem by allowing :finder_sql to be added to find options when the :include option is also specified.

Columns in the result set are mapped to attributes in the resuting object model through the :column_mapping option.

The :column_mapping option specifies a hash containing the following entries:

primary_key – indicates the column alias associated with the classes id attribute.
columns – is a hash of attribute to column alias associations.
associatios – is a hash of association mappings for each directly included model. The key for each entry is the name of the association, while the value is a hash similar to the hash for the parent class.

Example

Consider the following query relating authors with many books:


Author.find(:all, :include => :books)

This would likely produce the following SQL:


SELECT
    authors.id AS t0_c0, authors.name AS t0_c1,
    books.id AS t1_c0, books.author_id AS t1_c1, books.title AS t1_c2
  FROM authors
  LEFT OUTER JOIN books ON books.author_id = authors.id

The query might be written with books as the driver, rather than authors as follows:


 SELECT
    authors.id, authors.name,
    books.id AS book_id, books.name AS book_name
  FROM books
  JOIN authors ON authors.id = books.author_id

The Rails query would then be written as follows:


Author.find(:all, :include => books,
           :finder_sql => " 
                SELECT
                    authors.id, authors.name,
                    books.id AS book_id, books.name AS book_name
                  FROM books
                  JOIN authors ON authors.id = books.author_id".
           :column_mapping => {
             :primary_key => 'id',
             :columns => {
                  'id'  => 'id',
                  'name => 'name'
              },
              :associations => {
                :books => {
                  :primary_key => book_id,
                  :columns => {
                      'id'        => 'book_id',
                      'author_id' => 'id',
                      'name'      => 'book_name
                  }
                }
              }
            })

This is more verbose, but allows for absolute control of the SQL used to return results across multiple model associations.

A more complicated example would be the following:


QUERY = " 
SELECT
    a.id AS author_id,
    a.name AS author_name,
    p.id AS post_id,
    p.title AS post_title,
    p.body AS post_body,
    p.type AS post_type,
    c.id AS comment_id,
    c.body AS comment_body,
    c.type AS comment_type
  FROM authors a
  LEFT JOIN posts p ON p.author_id = a.id
  LEFT JOIN comments c ON c.post_id = p.id" 

MAPPING = {
  :primary_key => 'author_id',
  :columns => {
    'id' => 'author_id',
    'name' => 'author_name'
  },                         
  :associations=> {
    :posts => {
      :primary_key  => 'post_id',
      :columns => {
        'id' => 'post_id',
        'title' => 'post_title',
        'author_id' => 'author_id',
        'body' => 'post_pody',
        'type' => 'post_type'
      },
      :associations => {
        :comments => {
          :primary_key  => 'comment_id',
          :columns => {
            'id' => 'comment_id',
            'post_id' => 'post_id',
            'author_id' => 'author_id',
            'body' => 'comment_body',
            'type' => 'comment_type'
          }
        }
      }
    }
  }
}
Author.find(:all,
            :include=>{:posts=>:comments},
            :order=>"authors.id",
            :finder_sql => QUERY,
            :column_mapping => MAPPING)

This example shows the recursive nature of associations. It can also be used to render deeper structures that may reference the same model multiple times, such as the following:


Firm.find(:all,
          :include=>{:account=>{:firm=>:account}},
          :order=>"companies.id")

(Here, the query and mapping is left to the reader; or, you can look at the unit test which documents many more possibilities).

Install using


script/plugin install svn://rubyforge.org/var/svn/eagerfindersql

The plugin is managed on RubyForge. Rdoc is available.

Button Labels

gregg@kellogg-assoc.com (Gregg Kellogg) — Mon, 23 Oct 2006 21:22:00 -0500

I checked in the button-labels plugin. I was frustrated when trying to create a form with multiple radio-buttons and no way to tell them apart. Basically, if the :label option is passed to radio_button, it calls content_tag() instead of tag() and uses the label as the content.


radio_button("post", "title", "Goodbye World", :label => "Goodby World")

<input id="post_title_goodbye_world" 
     name="post[title]" 
     type="radio" 
     value="Goodbye World">
  <label for="post_title_goodbye_world>
    Goodby World
  </label>
</input>

The plugin is managed on ruby_forge. Rdoc is available.

Entered some time ago as a bug #4627.

Gregg Kellogg