== Version 0.24.0

A snapshot of the latest developments. This version features 
tons of big and small features and bug fixes that were 
implemented during the last month. Special thanks fly to
Tom, Chris, Guill, Rob, Aleksi, and all others that sent
patches and code.

Most notable additions:


* Totaly recoded annotation / property system. The property
  system is now based on Facet annotations and inheritors.
  You can now annotate every object, attribute or method
  in Nitro. For example you can annotate your actions with
  routing rules or sitemap strings etc, etc. One unified
  system for annotations and metadata is used throughout
  the whole Framework.
  
  Here is an example:
  
  class A
    attr_accessor :val
    ann :val, :klass => String
    ann :self, :doc => 'Self annotation'
  end
  
  A.ann.val.class => String
  A.ann.self.doc =>
  
  if A.ann.val[:class] 
  
  etc, etc...

* Implemented one of the most requested features. An Og
  frontend for KirbyBase. The KirbyBase store does not yet
  support all features, but is really useful. For example
  it can power the Spark wiki example. Switching between
  KirbyBase, Mysql, Postgres, Sqlite, etc by changing
  one line of code is really cool ;-)
  
* Better Seperation of Concerns for Og managed classes. Og
  can now annotate and handle classes from other libraries.
  Lets say you have the following class:
  
  class User
    attr_accessor :name
    attr_accessor :body
  end
  
  Using Ruby's open classes and Nitro's advanced annotation
  system you can easily prepare this class for Og management
  
  class User
    ann :user, :klass => String
    ann :body, :klass => String
  end
  
  or even better:
  
  class User
    property :user, String
    property :body, String
  end
  
  This style promotes SOC: You define your classes in one
  place and annotate them for Og in another place.
  
* Introduced a simple code generation framework. Two example
  generators are included:
  
  * app: generates the directory structure for a new
    nitro application.
    
  * form: generates an xhtml form for a Ruby class.
    This generator leverages the annotations of the
    object class.

* Improved scaffolding code and django style auto 
  administration system. The new scaffolder supports all
  Og relation types and provides an intuitive DHTML 
  interface for editing.
  
* SCGI (http://python.ca/nas/scgi/protocol.txt) support. 
  Thanks to a refactoring of Nitro's cgi code this 
  version integrates with Zed Shaw's SCGI adapter. This 
  provides FastCGI's performance with an easy 
  installation.

* Experimental HTTP streaming support. Here is a 
  simple example:

  def index
    response.content_type = "text/plain"
    stream do
      5.times do |i|
        print "#{i}"*10000 + "\n"
        sleep 1
      end
    end
  end

* Simple Og automatic evolution system. Lets say you have a class Article

  class Article
    property :title, String
    property :nbody, String
    property :dumy, Fixnum
  end
  
  lets you want to change your 
  class to this one:

  class NewArticle
    property :ntitle, String
    property :nbody, String
    property :new, Float
  end

  First you export the database:
  
  og.export

  Then you import the database. Some rules
  are needed when renaming classes or properties. 
  New properties or deleted properties are handled
  automatically.

  rules = {
    :Article => {
      :self => :NewArticle, # rename the class
      :title => :ntitle,
      :body => :nbody
    }
  }
  og.import :evolution => rules

  Thats all. In a future version this will be integrated into
  the default runner scripts.

* Og helpers to create simple rdbms management scripts. Here 
  is an example:

  mysql "-u root -p", <<-END
    drop database if exists weblog_development;
    create database weblog_development;
    grant all on weblog_development.* to #{`id -un`.strip}@localhost;
  END
  
  At the moment this is only available for Mysql.

* Added support for dynamic CSS using Nitro's advanced templating
  system. Create the dynamic CSS file with an .csst extension in
  your template root.
  
  <?r
  green = '#79b500'
  background = '#fff'
  ?>
  
  .page {
    background: #{green}
    
    <?r 10.times do ?>
      padding: 5px;
    <?r end ?>
    
    // or even
    
    <% 3.times do %>
      margin: 5px;
    <% end %>
  }
  ..
  
  then, add this line in run.rb:
  
  Compiler.precompile 'style.css'
  
  The CSS file will be regenerated automatically whenever the
  template changes...

* Updated to support latest Prototype, Scriptaculous etc.

* Cleaned up Og implementation.

* Fixed minor Ruby 1.8.3 compatibility issues.

* Even better integration with Ruby Facets.

* Tons of bug fixes and small but useful features.


== Version 0.23.0

The summer vacations are over and there is a brand new Nitro 
release. There is a preview of the new Scaffolder (also handles
Og relations), support for Tagging (folksonomy), lots of small 
features and improvements and many bug fixes. Additionally, the 
code has been restructured to utilize the excellent Nano and Mega
support libraries.

Most notable additions:

* Scaffolding reloaded. The scaffolding infrastructure is 
reimplemented to generate more flexible code. The automatically
generated forms allow for visualization and editing of 
Og relations such as HasMany and BelongsTo.

For example when rendering a BelongsTo relation all possible 
parents are presented with a select element. When rendering a 
HasMany relation a list of all children is presented.

Moreover, an experimental admin component is provided. Just add the
line:

require 'part/admin'

and surf

http://www.mysite.com/admin

To enter a simple administration screen. 

WARNING: This feature is considered a preview and will be 
improved in a future version.

* Major cleanup in the Glue subproject. Some files are moved
to the nano/mega project. Nano/Mega is now used throughout 
Nitro and really makes development so much easier.

* Introduced Og Taggable mixin. It was never easier to add
tagging to your application.

  class Article
    include Og::Taggable
    ..
  end

  article.tag('navel', 'gmosx', 'nitro')
  article.tags
  article.tag_names
  Article.find_with_tags('navel', 'gmosx')
  Article.find_with_any_tag('name', 'gmosx')

  t = Article::Tag.find_by_name('ruby')
  t.articles
  t.articles.count

For an example usage of this Mixin, consult the Spark sources.

* Added support for relative and absolute URLs in redirects
and renders. This feature simplifies the creation of reusable
components.

* Support for assigning compound objects from the request. Here 
is an example:

  class Article
    property :title, String
    property :body, String
  end

  <form>
    <input type="text" name="article.title" />
    <input type="text" name="article.body" />
  </form>

  article = request.assign('article')

Alternatively you can use the article[title] article[body]
notation.

* Added simple Benchmarking mixin.

* Added support for 'evolving' a single Og managed class. Useful
when you are in development mode and change your schema.

* Added support for session garbage collection.

* Many many small bug fixes in Og and Nitro.


== Version 0.22.0

A snapshot of the latest developments. Many requested features
where implemented, and many reported bugs fixed.

* The much requested Og 'reverse mode' is implemented. Og's 
domain specific language and API is extended to allow fine-grained 
customization of the schema-to-objects mapping. Og can now handle 
most of the legacy schemas you can throw at it. Here is an 
example:

class User
  property :name, String, :field => :thename, :uniq => true
  property :password, String
  property :age, Fixnum, :field => :age3    
  has_many Comment, :foreign_field => :user
  set_table :my_users
  set_primary_key :name, String
end

class Comment
  property :cid, Fixnum
  property :body, String
  belongs_to User, :field => :user
  set_table :my_comments
  set_primary_key :cid
end

As you can see, even relation fields can be customized. For 
higher level customization you can overload methods like #table,
#field_for_property etc.

* Og now handles multiple connections to multiple stores even 
in thread safe mode. This means you can now have objects serialized
in different RDBM systems in the same application. Or you can have
some objects managed by an RDBMS store and others by a FAST 
in-memory store. You can still have relations between objects in
different stores. An example:

mysql = Og.setup(:store => :mysql, ..)
psql = Og.setup(:store => :psql, ..)

class User
  has_many Comment
end

class Comment
  belongs_to User
end

mysql.manage_class User
psql.manage_class Comment

user.comments << comment

* Greatly improved support for testing and test-driven-development.
A reworked testing infrastructure for controllers is in place.
Moreover, support for Og fixtures and automatic test database
setup is provided. Fixtures can be defined with yml or csv files.

For an example of the testing infrastructure check out the 
Spark 0.4.0 sources. Spark is a Wiki powered by Nitro.

* Reworked mailer implementation. The new mailer allows for 
incoming email handling. Here is an example:

class Mailman < Glue::Mailer
  def receive(mail)
    article = Article.new(mail.subject, mail.body)
    article.category = Category.find_by_name(mail.to.split('@').first)
    article.save
  end
end

* Added script/runner in proto as a convenience to running
ruby code in your application environment.

* Many fixes in the caching infrastructure.

* Many fixes in the XMLRPC service infrastructure.

* Many smaller changes and feature implementations that make
development with Nitro so much more pleasurable.


== Version 0.21.2

This is mainly a bug fix release. Some minor features are 
implemented:

* Pluggable dispatchers. A new dispatcher that implicitly handles
nice urls is enabled by default. You can easily change to the
old general dispatcher though.

* Better markup setup.

* Cookie Helpers.

* Important bug fixes in Og.


== Version 0.21.0 was released on 25-07-2005

Nitro goes from strength to strength as the community gets bigger.
Great new features were introduced and a lot of effort 
was spend on cleaning up Nitro's implementation and polishing
'little things'. Many patches were contributed by the community
to make this is a release you will love!

Some notable changes:

* New compiler pipeline replaces the obscure shader mechanism. A
  compiler object is responsible for transforming a controller action
  and or the associated xhtml templates to a special ruby method
  that will respond to given urls with the action output. Using
  the aspect oriented features of Nitro or the specialized pipeline
  setup method a developer can easily customize the compilation
  stage to meet his needs. Here are some examples:
  
  Compiler.setup_template_transformation do |template|
    template = Elements.transform(template)
    template = Localization.transform(template)
    template = Markup.transform(template)
    template = Template.transform(template)
  end

  or 
  
  Compiler.before "puts 'Compiling'", :on => :compile
  Compiler.after "puts 'Template processed'", :on => :transform_template
  Compiler.wrap Benchmark, :on => compile
  Compiler.before :my_method, :on => compile
  
  Special rendering effects where never easier. If you need run
  time effects, just intercept the Render object.
    
* Integrated the new configuration system that was previewed in 
  the last release. All the configuration settings are now accessible
  either through the owner class or from the global Configuration
  variable using Ruby's metaprogramming features.
  
  Moreover, Nitro now dynamically decides the configuration mode
  (debug, stage, live/production) from environment variables or
  command line arguments so depolying your application is easier
  than ever: just run svn update on the live server, the application
  will grab the correct configuration parameters automatically.
  Of course you can use svn's advanced features to rollback to 
  earlier versions if you experience problems. The mode specific
  configuration settings are defined either on ruby files or
  yaml files or both.
  
  Point your browser to http:://my.app.com/settings to browse
  the settings for your application.

* Og dynamic queries. You can inject caclulated or join attributes
  to your objects. Here is an example:
  
  class Item
    property :quantity, Fixnum
    property :unit_price, Float
    
    def initialize(quantity, unit_price)
      @quantity = quantity
      @unit_price = unit_price
    end
  end
  
  Item.create 2, 3.0
  item = Item.find_one :select => 'quantity*unit_price as total_price'
  item.total_price # => 6.0
  
  Please note that total_price is a dynamically injected 
  attribute. Now you can combine SQL's powerful query features
  with the simplicity and elegance of Og.

* Og customized join tables allows you to use any Ruby object
  as the join relation between other objects. Here comes an 
  example:
  
  class Category
    property :title, String
  end

  class Article
    property :title, String
  end
  
  class ArticleToCategory
    property :rate, Float
    has_one Article
    has_one Category
  end
  
  c1 = Category.create
  c2 = Category.create

  a = Article.create
  a.categories.add(c1, :rate => 2.3) 
  a.categories.add(c2, :rate => 1.2) 

  for c in a.categories
    p a.category_join_data(c).rate
  end
  
* Og collections size() is now optimized.

* Og join code support refactoring. The new code is easier to read,
  more customizable and generates more efficient code by default.

* Nitro automatically assings reasonable template_roots on
  Controllers based on mount points. Moreover, the template_root
  overloading feature was improved. If you cant to reuse an
  existing controller but customize the view rendering by using
  different templates for some actions, just assign a different
  template root for this controller, and put the custom templates
  inside that dir. The custom templates 'override' the templates
  of the parent controller (templates for non customized actions
  come from the parent's controller template_root). Creating
  a library of reusable controllers (parts) was never easier.
  
* Major cleanup of the nitro source code, removed many files.

* Greatly improved Wee integration makes Nitro the premium 
  container for Wee components. Have a look at the updated
  wee example.
  
* Improved builder mechanism for programmatic rendering. The new
  builder resuses the controller helper mixins thanks to extensive
  refactoring of the implementation. Here is an example:
  
  class MyController
    def index
      build do 
        labels = ['George', 'Stella', 'Renos']
        html {
          head {
            title 'A simple test'
          }
          body {
            10.times {
              strong 'Hello World'
              i 'Hello World222'
            }
            select(:id => 'names') {
              options :labels => labels, :selected => 1
            }
          }
        }
      end
    end
    
* Improved helper integration. Helpers are now accessed either
  as mixins or through the builder mechanism.
  
  #{form entity}
  #{options :labels => [..], :values => [..], :selected => 1}
  
  or
  
  #{emit :form entity}
  #{emit :options :labels => [..] ...}

* Reimplemented the markup mixin to make this more flexible. You
  can easily customize the markup logic that is applied by the
  markup macros.

* Updated the documentation.
    
* Fixed all reported or discovered bugs, many smaller 
  improvements.

  
== Version 0.20.0 was released on 12-07-2005

Another superb release! State of the art AJAX/Javascript support,
Wee components / programmatic renderer integration, a beginners
tutorial, self documenting configuration and many important bug 
fixes. Don't forget to check out our new community site at
http://www.nitrohq.com

Some notable changes:

* Ajax is THE buzzword at the moment, and Nitro provides the 
  best support you can find. Nitro fully separates the behaviour
  from the template using the behaviour.js library and allowing
  for dynamic injection of ajax functionality. The generated code
  contains clean html and all the javascript organized in a
  single <script> block. Here is an example:
  
  in the Controller:
  
  def index
    # Inject functionality to the DOM elements of the template.
    
    behaviour '#alert', %{
      el.onclick = function() {
        alert('Hello world');
      }    
    }  
    auto_complete :country # convert to autocomplete field
    live :live # convert to async link!
    draggable :dragger, :revert => false
  end    

  def a_simple_action
    puts 'LIVE!'
  end
  
  def country_auto_complete # data for auto complete.
    %{
    <ul>
      <li>Greece</li>
      <li>England</li>
      ...
    </ul>
    }
  end
  
  in the Template:
  
  <label>Enter a country:</label>
  <input type="text" id="country" />
  <div id="country_auto_complete" class="auto_complete"> </div>
  
  Behaviour example:<br />
  <button id="alert">Click to alert</button>

  Drag and Drop example:<br />
  <div id="dragger">DRAG ME</div>

  Live/Asynchonous request (AJAX) example:<br />
  Here comes a <a id="live" href="a_simple_action">live link</a>.<br />
  (Check out the log to see the action called in the background!)

  All you have to do is define id's for your DOM elements. Here
  comes the generated code:
  
  <label>Enter a country:</label>
  ...
  (Check out the log to see the action called in the background!)

  <script type="text/javascript">
      var _behaviours = {
      '#alert': function(el) { 
      el.onclick = function() {
        alert('Hello world');
      }    
      },'#country': function(el) { 
      el.autocomplete = 'off';
     
      },'#live': function(el) { 
        el.onclick = function() {
          new Ajax.Request(el.href, {});
          return false;
        }
       }
      }
      Behaviour.register(_behaviours);    
        
      new Ajax.Autocompleter('country', 'country_auto_complete', 'country_auto_complete');
      new Draggable('dragger', {revert:false});
  </script>
  
  See how a normal <a> tag is converted transparently to an Ajax request.
  The prototype, scriptacoulous and behaviour js libraries are used.

  The ajax support in this release is a PREVIEW. Stay tuned for
  major improvements (and some surprises) in the next version.

* Wee Components integration. Nitro now transparently integrates
  Wee components. This is truly a win-win situation. Wee applications
  can use Nitro's infrastructure. Nitro applications can use 
  extremely reusable, stateful Components. For a demonstration 
  check out the new Wee example in the examples distribution! 
  Wee 0.9.0 is needed!
  
  Even better interoperability with Wee is coming soon. Nitro
  plays well with the others, and always provides more options.

* Mixins are modules of functionality that can be mixed into
  Controllers. The developer can explicitly include Mixins or
  use Nitro conventions to have helper Mixins automatically 
  included:
  
  class MyController
  ..
  end
  
  module MyControlllerMixin
    # gets automatically mixed in
  end
  
  This works just like Og Mixins.

* Brand new, self-documenting configuration system. There is a new
  keyword for defining configuration settings:
  
  Here is an example:
  
  class Render
    setting :template_extension, :default => 'xhtml', :doc => 'The default template extension'
  end

  class Session
    setting :store, :default => 'memory', :doc => 'The session store'
  end
  
  You can configure the Application using ruby code or yaml files:
  
  Render.template_extension = 'xhtml'
  Session.store = 'drb'
  
  or
  
  Render:
    template_extension: xhtml
  Session:
    store: drb
    
  You can access all defined settings:
  
  Configuration.settings.each { |s| ... }
  
  You can view the settings of the application along with documentation
  on the following url:
  
  www.myapp.com/settings
  
  This feature is also a PREVIEW. Will be used a lot more in the 
  next release.
  
* CherryPy style published objects. Nitro allows you to publish
  any Ruby object. Here is the new hello world example:
  
  class HelloWorld
    def index 
      print 'Hello World'
    end
    
    def add(val)
      print "added: #{val + 2}"
    end
  end
  
  App.start(HelloWorld)
    
  Now, point your browser to localhost:/ or localhost/add?val=2
  
  If you need the advanced controller functionality just extend
  your published object from the Controller base class. The normal
  heuristics to decide which method is safe to publish are aplied.

* CherryPy-style dispatcher configuration, provides another way
  to define mounting points for Controllers and published objects.
  Here is an example:
  
  server = Server.new
  server.root = MainController              # /
  server.root.fora = ForaController          # /fora
  server.root.wiki = WikiController          # /wiki
  server.root.really.deep = DeepController  # /really/deap

* Improved pager interface. Thanks to the new PagerMixin, presenting
  paged lists is easier (and as efficient) than ever:
  
  items, pager = paginate(Article, :per_page => 5, :order => 'date DESC')
  
* Added better sql injection protection in Og sql stores.

* Fixed Mysql store reconnect bug.

* Og falls back to pure ruby adapters for Mysql and Postgres, to 
  make it easier to run out of the box. Please, don't forget to 
  switch to the natively compiled adapters for production sites.

* This is surely the most request feature: Nitro Step by Step
  by James Britt, a beginers guide to Nitro. Available in the
  brand-new, Nitro-powered, www.nitrohq.com Community wiki.

* New examples: The totaly recoded and ultra cool ajax example,
  a Wee/Nitro example and the new Hello world example.

* Cleaned up a lot of source files.
  
* Many, many, many bug fixes and small improvements. This release
  fixes all reported bugs in the spirit of out zero-bug tolerance
  philosophy.
  

== Version 0.19.0

Og reloaded part 2: Another superb release introducing a balanced 
mix of innovative new features, common but useful stuff and bug 
fixes. 

Some notable changes:

* Og polymorphic relations. A groundbreaking feature made possible
  by Og's unique design and Ruby's power. Let's use an example 
  to explain the concept:
  
  class Comment
    ...
    belongs_to Object # polymorphic marker
  end
  
  class User
    ...
    has_many Comment
  end
  
  class Article
    ...
    has_many Comment
  end
  
  u = User.new
  u.comments << User::Comment('Hello')
  
  a = Article.new
  a.comments << Article::Comment('Wow!')
  
  User::Comment and Article::Comment where automatically created
  by Og and are serialized in different tables (also automatically
  created by Og). This is the  next step in DRY!

* Og now supports inheritance using the well known Single Table
  Inheritance pattern. Thanks to Og's advanced design the pattern
  is fully encapsulated:
  
  class Document
    ...
    schema_inheritance
  end
  
  class Article < Document
    ..
  end
  
  class Photo < Document
    ..
  end
  
  Document.all # => includes Articles and Photos
  Article.all # => only Articles
  
  User.documents # => Articles and Photos
  User.documents(:type => Photo) # => only photos.

  Btw, this feature is orthogonal to the polymorphic relations
  feature just described, giving the developer great 
  flexibility.

* Added SUPERB support for auto reloading, the new system 
  can detect source changes everywhere! Based on original
  code by Michael Neumann.
    
* Introduced the concept of Flash, a temporal store where
  objects that should be kept alive for the next request are
  saved.

* Recoded the Blog example to use the Elements system for 
  shading instead of XSLT. The new code runs easier under Windows
  (so the no_xsl_blog example is now removed). Here comes an 
  example:
  
  <Page>
    <Box>
      Please login as an author by entering the blog password.
      <br />
      The password is: <b>#{Blog.password}</b>.
    </Box>
    
    <Error>#@error</Error>
    ...
  </Page>

  The <Page> tag is supported by the Page Ruby class to do the
  programmatic rendering of templates or logic code:
  
  class Page
    def render
      ...
    end
  end

  This system can be used to create a library of useful 
  components to abstract common tasks.
  
* The updated Blog example demonstrates how easy it is to write
  webservices using Nitro and the experimental Service feature.
  The code is so short, so lets paste it here:
  
  module BloggerApi
    SBlog = Struct.new(:url, :blogid, :blogName)

    def blogger__getUsersBlogs(appkey, username, password)
      Blog.all.collect { |b| SBlog.new('http://www.gmosx.com', b.oid, b.title) }    
    end
  end

  module MetaWeblogApi
    def metaWeblog__newPost(blogid, username, password, content, publish)
      entry = BlogEntry.new(content['title'], content['description'])
      entry.blog = Blog[blogid]
      entry.save
      entry.oid
    end  
  end

  module MovableTypeApi
    SCategory = Struct.new(:categoryId, :categoryName)

    def mt__getCategoryList(blogid, username, password)
      blog = Blog[blogid]
      blog.categories.collect { |c| SCategory.new(c.oid, c.title) }
    end

    def mt__setPostCategories(postid, username, password, categories)
      cid = categories.first['categoryId']
      BlogEntry.update(postid, "category_oid=#{cid}")
      true
    end
  end
  
  class ApiController < Nitro::XmlRpcService
    include BloggerApi
    include MovableTypeApi
    include MetaWeblogApi
  end

  That's all!
  
* Integrated an SQLite3 patch by Ghislain Mary.

* Integrated PostgreSQL binary data patch by Michael Neumann.

* Fixed all reported bugs.


== Version 0.18.0 was released on 01/06/2005. 

Mainly a bug fix release. Many small improvements were 
implemented. All reported bugs were fixed, as well as bugs found 
during the deployment of a live application. For more information
consult the detailed changelog. Thanks to Julien Perrot for small
patches.

Some notable changes:

* Some changes to make Webrick a valid solution for powering
  a production server. The full page output caching was improved 
  and support for daemonizing the server was added. Support
  for running as a proxy target (behind an apache server) was also
  added. 
  
* Thread safe mode was added again in Og. This works nice with
  the Webrick server.
  
* New order macro in Og to set default ordering for each
  entity. The has_many collections respects the order setting.
  
* Improved error reporting.

* Support for template_root overloading. Lets say you have
  the following Controller hierarchy:
  
  Controller > MyController > SpecificController.
  
  The template roots of the 3 controller are searched in reverse
  order. So by adding a template file in the SpecificController
  template root you can override a default action in the base
  Controller. For convienience the base Controller points to
  the Proto directory. This is experimental.

* Provide the examples as a separate distribution to make it
  easier for Ruby newbies to find them.

* Bug fixes in the Elements system.

* Improved the nitrogen generator command. Just run

  nitrogen app ~/the/target/path
  
  to create a skeleton application.
  
* Cleaned up some source files.


== Version 0.17.0 was released on 16/05/2005. 

The biggest release yet, featuring the brand new implementation
of Og. Due to the many changes this should be considered a preview
release, a stabilized version is expected shortly. Bug reports and
comments on the new features are especially appreciated. Many 
thanks to Michael Neumann, Dan Yoder and other hackers on the mailing 
list for ideas and suggestions that made this version possible.

Most notable Og additions:

* Extremely clean source code. Better names are used thorougout.
  Extra care was taken to make all features more orthogonal.
  
* Brand new relation mechanism. The 'enchanting' of entities 
  (managed classes) happens in multiple passes to be more
  flexible. Totaly separated graph/metadata creation and serialization
  code generation. The Graph metadata can be used for advanced
  scaffolding, testing and more.

* Support for fully customizable primary keys. You are no longer
  forced to use xxx_oid primary keys. Appart from the extra 
  flexibility this feature provides, this is an essential step
  towards the planed 'reverse engineering' mode that will allow the
  use of existing schemas with Og.
  
* More elegant inspection mechanism. Example:

  Article.relation(:user) # => Og::BelongsTo(...)
  Article.relations # => [...]
  Article.properties # => [...]

* Joins_many relation, as an alias for one way, join table relations.

* Support for 'active' collections. Active collection are
  synchronized with the backend Store and provide a more elegant
  interface and the opportunity for 'session' caching:
  
  article.comments << Comment.new
  
  instead of 
  
  article.add_comment(Comment.new) # this is also allowed though.
  
  p article.comments
  p article.comments.size # the results of the first query is cached

* Eager relations.

  comments = Article.comments(:include => User)

  for comment in comments
    p comment.user.name
  end
  
  Elegantly solves the N+1 query problem by using one join
  query.

* No need for forward references when defining relations. Now,
  the following code magically works:
  
  class User
    has_many Comment # works even though Comment is not defined!
  end
  
  class Comment
    belongs_to User
  end
   
* Use inflection where possible to infer missing configuration
  options. For example
  
  class Article
    belongs_to User # infects relation name :user
    ...
    
* New, lean and mean Store interface. The code needed to teach 
  Og how to  serialize objects to backend store is dramatically
  reduced. The new interface is SQL agnostic, so non SQL-RDBM's 
  stores are possible.
  
* SQL agnostic querying interface, compatible with non-sql 
  Stores. Here is an example:
  
  Article.find(
    :condition => 'hits > 2 AND rate > 3', 
    :order => 'title',
    :offset => 30,
    :limit => 10
  )

* More elegant (and non-sql store compatible) way for selective 
  updates:

  article.title = 'Changed'
  article.hits += 1
  article.update(:title, :hits)

* New, in-memory store that support all features. This is a pure
  ruby solution useful for experimentation. It will also serve
  as the base for the forthcoming madeleine Store implementation.

* Allow for multiple stores in one application. A great example,
  mysql_to_psql is provided. This example uses Og's powerfull
  features to automatically convert a Mysql database to a 
  PostgreSQL database. Database migration was never easier.
  
* Uses the excellent Facets utility collection to further
  clenup and minimize the code.

* Managed classes or Entities should include the EntityMixin
  or extend the Entity class. Example:
  
  class Article < Entity
    ..
  end
  
  class Article 
    include EntityMixin
  end
  
  This is done to avoid the Module module like in earlier
  versions of Og. However, Og is can infer the need to include
  the Managed mixin in typical cases:
  
  class Article
    property :title, String 
    # when a property is defined Og automatically converts the
    # class to an Entity
  end
  
  class Article < AnExistingManagedEntity
    # also includes the mixin
    ...
    
  class Article
    include AModuleThatDefinesProperties
    ...

* Improved support for og_delete interception.

* Support for nested transactions.

Check out the file test/og/tc_store.rb for a demonstration of
the new features.

And some Nitro additions:

* Integrated the Facets library, and started donating some Glue
  code to this project.
  
* Integrated the Prototype object oriented javascript library.

* Included a preview implementation of the new tag library 
  system, called Elements. This is based on an original idea
  by Dan Yoder. This feature is expected to change substantially
  in the next version. In the meantime here is an example:
  
  <Page>
    Here is a <b>nice</b> page.
    <br />
    <Box>
      This is #{variable}.
      <?r if admin ?>
        hello admin
      <?r end ?>
    </Box>
  </Page>
  
  The tags <Page> and <Box> automatically instantiate objects
  of class Page and Box that handle the rendering. This is an
  alternative to the XSLT templating system and can also be used
  to implement some reusable components. The hierarchical structure
  of the elements is available to each object to facilitate 
  advanced rendering tricks. Just like the XSLT shader the Elements
  transformation is preapplied to the page to avoid the overhead
  at runtime.
  
* Fixed many reported bugs.


== Version 0.16.0 was released on 16/04/2005.

A snapshot of the latest developments. Many, many subtle improvements,
new features and a major cleanup of the source code. Thanks to 
James Britt for significantly contributing to this release.

Most notable additions:

* Aspect Oriented Programming support. This new system
  is used to reimplement features such as Controller filters,
  Og callbacks and Og observers. By using this unified
  system you can now add Observers to controllers and use
  a metalanguage for wraping Og object callbacks:

  class Controller
     pre :force_login, :where => :prepend
     wrap Benchmark, :on => :index
     post :taraa, :on => login
  end
 
  module Timestamped
     pre :on => :og_insert { |this| this.create_time = Time.now }
    pre :on => :og_update { |this| this.update_time = Time.now }
    pre :on => [:og_insert, :og_update] { |this| this.create_time = Time.now }
  end

  This feature will be used extensivelly  in future versions
  to improve logging, the shaders and more.

* Added support for Test Driven Development. Many helpers
  and utility methods are added to the standard TestCase
  class, more will come in a future version:

  handle(
    '/login', 
    request = { 'password' => Blog.password }
  )
  assert_redirect
  assert_session_has(:owner)
  assert_session_equal(:username, 'George Moschovitis')
  assert_has_cookie('nauth')
  assert_has_no_cookie('wow')
  assert_cookie_equal('nauth', 'just an example, not used :)')

* CGI adapter (by James Britt)

* Major cleanup of the source. Converted the N namespace to
  Nitro, to be more compatible with other Ruby projects. 

* Add Og Timestamped mixin.

* Og improvements.

* Improved runner helper.

* Improved reloading support.

* Improved the Gem installation process.

* Fixed all reported bugs.


== Version 0.15.0 was released on 04/04/2005.

A great release. Many cool new features and tons of subtle 
improvements. We also welcome a new core developer, Anastastios 
Koutoumanos, who's first contribution is the new SqlServer adapter.

Most notable additions:

* Advanced localization support:

  locale_en = {
    'See you' => 'See you',
    :long_paragraph => 'The best new books, up to 30% reduced price',
    :price => 'Price: %d %s',
    :proc_price => proc { |value, cur| "Price: #{value} #{cur}" }
  }

  locale_de = {
    'See you' => 'Auf wieder sehen',
    ...
  }

  lc = Localization.get
  lc['See you'] -> See you
  lc[:price, 100, 'euro'] -> Price: 100 euro
  lc = Localization.get[:de]
  lc['See you'] -> Auf wiedersehen
 
  Using the LocalizationShader you can have templates like this:

  <h1>[[This is a localized]] String</h1>
  <p>
     do you [[:like]] this?
  </p>

  All strings enclosed in [[ ]] are replaced with localized
  versions (honouring the session locale). Check out the updated
  blog example.

* Dynamic/Parametrised mixins. A great extension to Ruby's mixin
  feature. The Og::list implementation is replaced with the new
  Orderable dynamic mixin, here is an example:

  class Comment
    property :body, String
    belongs_to :article, Article
    include Orderable, :scope => article
  end

  c.move_higher

  The Orderable mixin uses the :scope parameter to dynamically alter
  the methods appended to the Comment class. This new feature will be 
  used throughout the platform.

* NestedSets mixin:

  class Comment
    include NestedSets
  end

  or

  class Comment
    include Hierarchical, :method => :nested_sets
  end

  c.add_comment(child_comment)
  c.full_children
  c.direct_children
  c.children
 
  this is a reimplementation of the SqlTraversable mixin
  available in older versions.

* Improved templating system. Now allows <% %> intrerpolators
  and provides a number of html morphing effects:

  <div if="admin">
    admin interface
  </div>

  <ul>
    <li each="u in @users">#{u.first_name} #{u.last_name}
  </ul>

  and more...

* Og provides an SqlServer adapter out of the box.

* Improved scaffolding code (still a lot more to come).

* Build default environment, introduced nitro and nitrogen 
  commands to help new users.

* Many, many small improvements and fixes.


== Version 0.14.0 was released on 28/03/2005.

This release fixes *important* bugs amd improves various
aspects of the platform. Moreover, we included some 
great *new* features for your pleasure.

Most notable additions:

* Fixed IMPORTANT bug in property inheritance.

* Fine grained caching. Nitro allows you to cache
  whole pages (output caching), actions and fine-grained
  fragments. 

  class MyController < Controller
    cache_output :my_action

    def my_action
    end
  end

  Stores the whole page created by the my_action method
  to the disk to be displayed by the web server thus completely 
  Nitro and Ruby.

  or 

  <strong>Here is some cached code</strong>
  <cache>
    <ul>
    <?r for a in Article.all ?>
      <li>#{a.title}: #{a.body}</li>
    <?r end ?>
    </ul>
  </cache>

  or

  <strong>Another one</strong>
  <?r cache('variant', :admin => session[:admin]) do ?>
    ...
  <?r end ?>

  Cached fragments can be stored in memory, filesystem.

  While this feature is fully operational, the API will be finalised in 
  the next version.

* Introduced support for Og mixins. In this version a List
  mixin is provided:

  class Article
    has_many :comments, Comment, :order => 'position DESC'
  end

  class Comment
    belongs_to :article, Article
    acts_as_list :scope => :article
  end

  An AR compatible API is provided. An alternative
  API is planned for the near future to give you more choice.

* Reimplemented filtering infrastructure, allows
  for inheritance, conditional application of filters
  (:only/:except) modifiers, more performant Filters
  as Strings and more.

* Fixed multipart support in fastcgi, added file upload
  example (tiny example).

* The webrick adapter reuses the fastcgi infrastructure, making
  the adapters more compatible with each other.

* Added many useful Og enchant methods.

* Cleaned up configuration files for lighttpd/apache.

* More compatible with win32.

* Fixed examples and all reported bugs.


== Version 0.13.0 was released on 17/03/2005.

A snapshot of the latest code. The Nitro project is now
split in three gems (nitro, og, glue) and the code is 
better separated. Many changes were made to make programming
with Nitro more similar to Rails, to make the platform
more accessible to newcomers. A conversion of ActionMailer
and an AJAX example are also included.

* Introduced Mailer subsystem:

  class MyMailer < Mailer
    def registration_email(to, username, password)
      @from = 'system@navel.gr'
      @to = to
      @subject = 'Registration information'
      @body.username = username
      @body.password = password
    end
  end

  the tempate (registration_email.xhtml):

  Hello #{username}

  Thanks for registering, your password is '#{password}'

  Then use it like this:

  MyMailer.deliver_registration_email('gmosx@navel.gr', 'gmosx', 'rulez')

  For more information, check out the updated blog example.

* AJAX example, demonstrates how to use ajax techniques with 
  Nitro. A simple Google Suggest style UI is implemented.

* Use a Rails compatible directory structure. Check out
  the updated blog example. Please note that Nitro does not
  force a directory structure (see an alternative structure
  in the no_xsl_blog example).

* Optional separate template_root/public_root for extra security.

* Further code cleanup and bug fixes.

 
== Version 0.12.0 was released on 07/03/2005.

A careful blend of new features and subtle improvements
to the existing infrastructure. Some important bugs where
fixed aswell.

Most notable additions:

* Nitro allows the definition of metadata for each action.
  Routing (rewrite) rules, parameter constrains, hierarchy information
  and custom data can easily be attached to actions:

  def view
    @entry = Article[@oid]
  end
  action :view, :route => /view\/(.*)/, 'oid' => 1

  just browse

  view/1

  and @oid is automatically initialized with request['oid']

  Browse the source to see how to add additional constrains,
  more is comming in the next version. This feature replaces 
  the non portable ParseTree implementation. The scaffolder
  is updated to generate routings for nice urls.

* Og automatically generates finders for all properties, for
  even easier (and portable) querying:

  class Article
    property :title, :body, String
    property :hits, Fixnum
    property :create_time, Time
  end

  you get the finders:

  Article.find_by_title
  Article.find_by_body
  Article.find_by_hits
  Article.find_by_create_time

  The finders take into account the unique constrain, to return
  an array or just an object as needed.

* Og introduces lifecycle observers to avoid 'poluting' the model
  objects with excess functionality. You can use every object
  as observer (duck typing) or extend from an AR style Observer
  class. The observer callbacks are precompiled in the lifecycle
  methods only if defined, so the perfomance is not affected
  in the general case.

* Factored out templating engine, can now be used in stand-alone
  mode. Usefull for example to render email templates etc:

    template = %q{
      Hello #{user}

      dont forget the following facts:
      
      <?r for item in items ?>
        <li>#{item}</li>
      <?r end ?>
    }

    user = 'gmosx'
    items = %w{ nitro is really great }
    out = '' # the rendered template comes here.

    Template.process(template, :out, binding)

* New options in the default runner: 
  
  --render crawls a web application and renders all pages
  as static html files. Allows you to leverage Nitro's 
  advanced templating features to generate static sites.

  --crawl spiders the application, useful for testing.

* Better error page, with more information (see blog example).

* Fixed Og bug: multiple many_to_many relations with the
  same target class.

* further code cleanup, improved examples and more.


== Version 0.11.0 was released on 28/02/2005.

The platform continues to evolve and now supports the 
Apache web server and the Oracle database out of the box.
This version features improved documentation, important
bug fixes  and many subtle improvements to make programming
even more enjoyable. Many thanks to Matt Bowen for his help
with this release.

Most notable additions:

* Configuration files for Apache/FastCGI.

* Og documentation (doc/og_tutorial.txt, doc/og_config.txt)
  
* Actions with parameters:

  def list(username, oid)
    ...
  end

  http://www.mysite.com/list?username=tml;oid=2

  calls list() with the correct parameters, no need to 
  use request['oid'], just use oid.

  def list(username, oid)
    puts oid # oid == request['oid']
  end

  WARNING: this feature requires the ParseTree library which
  is not compatible with Windows. For the moment, this feature 
  is dissabled by default. To enable this feature, set 
  Nitro.resolve_action_arguments = true.

* New session store system, DRb sessions (stored in an independent
  DRb server) are reimplemented.

  Session.store_type = :drb

* TableBuilder helps in creating html tables.

* Integrated builders with the programmatic render. For example:

  <div class="mytable">
    #{o.build_table(:id => 'my_tab', :headers => headers, :values => values)
  </div>

  or
  
  <p>
    <b>Object Form</b>
    #{o.build_form(obj)}
  </p>

* Og Oracle adapter.

* Og provides advanced metadata for the managed objects

  class Article
    property :title, String
    property :body, String
    has_many :comments, Comment
  end

  par = Article.properties_and_relations
  => [Property(:title), Property(:body), Og::HasMany(:comments)]

  par[2].klass
  => Comment

  par[2].meta[:linkback]
  => :article_oid

* Og Typemacros, here is an example:

  def VarChar(size)
    return String, :sql => "NOT NULL VARCHAR(#{size})"
  end

  property :title, VarChar(30)

* Option for faster startup, skip schema check.

* Many small Og improvements and fixes.

* FastCGI adapter bug fixes.

* New example, why's wiki, from a cool post at
  http://redhanded.hobix.com

* IMPORTANT security fixes.


== Version 0.10.0 was released on 15/02/2005.

Another strong release featuring a completely recoded Og
implementation and redesigned Og adapter (formely backend) 
system. An SQLite3 adapter is also provided. Moreover the
Nitro installation process is improved, and many small
changes make the framework more elegant and easier to 
use: the updated examples reflect this!

Most notable additions:

* Improved Og implementation (cleaner code) and new Og 
  adapter subsystem.

* New SQLite3 Og adapter, improvements in MySQL and PostgreSQL 
  adapters (WARNING: needs version 1.1.0 of Sqlite3-Ruby).

* Added install.rb for easier installation of the tar.gz distribution.

* Better GemSpec for easier installation by RubyGems.

* Action/template/xsl auto-reloading system in debug mode,
  new implementation, works again.

* New Nitro configuration system, with rational default
  parameters.

* --console option attaches an irb session to a running
  instace of an application (works again).

* Og supports optional typechecking by using property metadata.

* request alias for context to be compatible with older
  versions of nitro and Webrick/jsp and other frameworks.

* Improved the examples, cleaner code, work from any
  directory.
 
* Removed more obsolete code and improved directory
  structure.

* and many more smaller fixes.

WARNING: If you used an earlier version of Og you
may need to drop your database and let Og recreated it
automatically.


== Version 0.9.5 was released on 04/02/2005.

A bug fix release.

* Fixed nasty Windows deadlock bug.

* Experimental Wee-style example (will be updated in the next version).

* Changed default port to 8069 (8080 was to common).

* Small fixes and improvements.


== Version 0.9.3 was released on 01/02/2005.

A *very* important release. A new abstract rendering
pipeline is introduced. This pipeline allows for multiple
adaptors (WEBrick, fastcgi are included). 

Many classes have be rewritten using *clean* code, 
and the  directory structure has improved again. All 
unused and obsolete files have been removed. A new 
version of the Blog example was included. This version 
does not use XSLT and runs on Windows. 

If you havent seen Nitro or you had problems with an 
earlier version, this is the time to revisit this 
project!

Most notable additions:

* Drastically improved source code readability.

* New abstract rendering pipeline. Benefits:
  
  * multiple adaptors.
  * better unit tests.
  * optimized memory allocation.
  * more modular desing.

* FastCGI support (tested with Lighttpd).

* no_xsl_blog example.

  This new example is easier to setup, even on Windows.
  This also demonstrates that Nitro does NOT force
  the use of XSLT.

* No global variables used for nitro configuration.

* Updated the docs to better reflect the rappidly
  evolving codebase.

* Og metalanguage relations insert metadata into 
  the target class, useful for advanced scaffolders.

* Og refer_to meta-language command. 

* Many bug fixes and small optimizations.

The next version will stabilize the Nitro api.


== Version 0.8 was released on 12/01/2005.

A snapshot of the latest code. Cool new features,
many fixes and improvements in older features. Many thanks
to Michael Neumann for giving intelligent suggestions
and finding small bugs.

Most notable additions:

* New automatic validation system:

  class User
    prop_accessor :name, :password, String
    validate_confirmation :password
    validate_length :name, :range => 2..12
  end

  u = User.new(...)
  unless u.valid?
    p u.errors.on(:name)
    p u.errors[:password]
  end
  
* Programmatic xhtml rendering. This is an alternative
  method of rendering that is usefull in building 
  components and helpers. This is a preview implementation.

  options = ['Male', 'Female']
  o.select(:name => 'sex') { 
    o.options(options, selected = 1) 
  }

  options = { 'Male' => 'm', 'Female' => 'f' }
  o.select(:name => 'sex') { 
    o.options(options, selected = 1) 
  }

  o.html {
    o.p(:class => 'header') { o.b('Hello') }
  }

* No global variables in Og. 

* Recoded Og to allow for future support of multiple databases 
  (even on different RDBMS systems) on a single application.

* Improved Blog example demonstrates latest advancements.
  (You have to drop the database of earlier versions!)

* More unit tests.

* Supports Ruby 1.8.2

* Integrated extensions/dev-utils by Gavin Sinclair.

* Integrated blankslate by Jim Weirich.

* And many IMPORTANT bug fixes.

  
== Version 0.7 was released on 27/12/2004.

A snapshot of the latest code. Many fixes and new features result
in a more mature product. Many thanks to the ruby hackers that sent
suggestions and patches used in this release!

Most notable additions:

* Improved win32 compatibility (Tiny example runs out of the box).
* Binary content example (Flash, needs ruby-ming installed to run).
* Totaly recoded prop_accessor mechanism, avoids polution of the Module
  class.
* prop_accessors for Modules, allows synthesizing of managed objects
  from Mixins.
* new automatically generated methods in Og.
* MockDatabase leverages the FlexMock object for easier unit testing.
* Integrated the cool Breakpointer library by Florian Gross.
* Markup mixin to automatically expand/compact generalized markup code.
* new_form wizard.
* Statically (compile time) included subscripts.
* Important bug fix in redirects.

and many many bugfixes! :)

== Version 0.6 was released on 13/12/2004.

This is a preview release, the api for the new features is not 
finalized. This early release gives other developers to offer suggestions
on the final form of those features.

Most notable additions:

* Og many_to_many relations with auto generation of the join table.
* Og has_one relation.
* PHP-style nested output buffering.
* Rails-style Filters.
* autoreload services.
* complile time evaluation of ruby code in templates.
* improved pager ui.
* initial version of FormBuilder.
* initial version of new_app wizard.
* Improved Blog example.

and... the nitro logo :)


== Version 0.5 was released on 23/11/2004.

Many new features:

* Stand-alone Og distribution.
* Scaffolding.
* Og RDBMS index support (works again).
* improved render/dispatcher.
* REST/XML dspatcher.
* RssBuilder.
* Improved BLOG example.
* Many fixes.


== Version 0.4.1 was released on 15/11/2004.

A MAJOR revision! The rendering engine was recoded from 
scratch and now features Rails-style actions. Some older features
are not yet converted to the new engine though. Other new stuff 
includes a fully working / cool looking Blog example, a programmable 
shader pipeline, automatic object manager in Og, improved directory 
structure, bug fixes, many changes to make it easier to switch from 
Rails, and many many more cool improvements for you to find out in 
the detailed ChangeLog.


== Version 0.3.0 was released on 01/11/2004.

An important revision! Nitro now includes the first version
of a brand new ObjectRelational mapping library called 
Og (ObjectGraph) that combines the best features of Active
Record and NDB. A fully working MySQL adapter is also provided.
Moreover the code base is further cleaned up. Small improvements 
to the application configuration system.


== Version 0.2.0 was released on 25/10/2004.

Greatly improved RDoc documentation. Cleaned up many source files,
and improved the directory structure to be more compatible with
other Ruby projects. Introduced a test suite. Important bug fixes
in NDB. Tiny example no longer requires apache so should run out
of the box.


== Version 0.1.2 was released on 21/10/2004.

The first public version. Features 2 examples and limited
documentation. The aim of this release is to make the project known
to the Ruby Community. We hope that the Ruby hackers will respond
with valueable suggestions and ideas.
