Skip to content

Latest commit

 

History

History
551 lines (366 loc) · 17.3 KB

README.md

File metadata and controls

551 lines (366 loc) · 17.3 KB

Innate

Innate is the core of Ramaze, but useful on its own.

Philosophy

The philosophy behind Innate is to provide a simple web framework that:

  • Stays below 2000 easily readable lines of code
  • Has Rack as the only dependency
  • Provides the default helpers
  • Is easy to encapsulate and reuse in other contexts as a simple rack app
  • Has a rock-solid and fast implementation
  • Scores at least 95% in rcov
  • Is fully documented using YARD

Innate vs. Ramaze

Numerous people asked where Innate fits into the picture of Ramaze and when it would be appropriate to use Innate alone.

The answer to this is not as simple as I thought, given that Ramaze improves on it a lot by taking full advantage of the encapsulation.

Innate started out as I grew increasingly annoyed by the many hacks and add-ons that diluted the original simplicity of Ramaze as we discovered ways to do things better, so I went on to create something that is impossible to dilute and simply works as a multiplier.

It does work standalone, and it might be suitable for applications that just need the little extra over plain old Rack, making it easy to deploy and rely on the stability that comes with simple code. It's also the starting point if you want to study the inner workings of Ramaze without being distracted, all you need to keep in mind is Rack.

If all you need is a couple of ERB templates and a few nodes then Innate is the way to go. Upgrading to Ramaze later is quite easy, just change a couple of lines and the power of your code multiplies (again).

The split between Innate and Ramaze is clear cut and gives Ramaze freedom to expand in any direction (and I have quite some ideas where to take it).

Ramaze adds things that require other dependencies, caching with memcached or sequel, logging with analogger or growl, templating with haml or erubis, just to name a few.

Ramaze also adds Apps to the mix, and it is easier to understand them when the concepts of what's going on inside an App are completely abstracted by Innate.

To summarize, Innate really keeps it simple, acts as a learning tool and encourages everybody to build on top of it whatever they want.

Features

  • Powerful Helper system
  • Nodes[1] simply include Innate::Node, so class inheritance is your choice.
  • The only hard dependency is Rack
  • Easy to get started
  • Compatible with major Ruby implementations[2].
  • Usage of Fiber[3] instead of Thread if possible.
  • Namespaced serializable configuration system
  • Simple testing without need of a running server
  • No clutter in your application directory structure, scales from a single file upwards
  • Seamless integration with Rack middleware
  • No patching of ruby core or stdlib.
  • Direct access to the current Request, Response, and Session from anywhere via Trinity
  • Works out of the box with ERB the templating engine.
  • Support for all rack-compatible web-servers.
  • Dynamic content-representation.

[1]: What you may think of as Controller. [2]: This includes: Ruby 1.8, Ruby 1.9.1, JRuby, Rubinius [3]: Fiber is available on 1.9 only at this point.

Usage

A simple example of using Innate that also shows you how to add your custom middleware, write specs and the overall concept:

require 'innate'

Innate.setup_middleware

Innate.map('/') do |env|
  Rack::Response.new(['Hello, World!']).finish
end

Innate::Mock.get('/')

And another example, using Node with a normal server:

require 'innate'

class Hi
  include Innate::Node
  map '/'

  def index
    "Hello, World!"
  end
end

Innate.start :adapter => :mongrel

Installation

Via git (recommended)

Installing Innate from git is highly recommended, since it gives you easy access to alternate branches, bugfixes, and new features.

git clone git://github.com/Ramaze/innate.git

Do not forget to add the innate/lib directory to your RUBYLIB environment variable.

For unixish systems you may want to add it to ~/.bashrc or the equivalent for your shell:

export RUBYLIB="~/path/to/innate/lib:$RUBYLIB"

Via gem install

gem install innate

Direct download

Download the latest tarball from here.

Concepts

First let's see about the good old MVC:

Model

Innate doesn't have any ties to models, it does not offer helpers or rake tasks or whatever you may be expecting, there is no M, use whatever you like. Innate is, however, known to be compatible with the ORMs listed below:

  • ActiveRecord
  • DataMapper
  • M4DBI
  • Og
  • Sequel

Please consider giving us a heads up about what worked for you if it isn't in the list yet.

View

Innate supports multiple templating engines and it is very easy to add your own. At the moment we offer following engines out of the box:

  • ERB
  • Etanni

Controller

Innate follows a different approach than most frameworks, making the controller subclassing obsolete. To make an object accessible from the outside simply include Innate::Node and map it to the location you would like to serve.

Differences from Ramaze

Innate throws a lot overboard; it doesn't provide all the bells and whistles that you may be used to. This makes Ramaze the way to go for larger applications.

For this reason, Innate is not only a standalone framework, it is also the core of Ramaze.

Ramaze started out without any of the benefits that Rack gives us these days, especially regarding the server handlers, request/response, and middleware.

Still it tried to provide everything one might need with the least effort, leading to a lot dependencies (we had things like bacon, simple_http, gettext, mime types, ...)

Configuration

Innate provides the Innate::Options DSL, similar in some aspects to the old Ramaze::Global, but a lot more flexible.

Options has namespaces, inheritance, defaults, triggers, documentation, and a sane name.

The definition syntax roughly resembles Ramaze::Global.

We break with the tradition where all options one would ever need were provided in one file. It made maintenance rather difficult and the code hard to follow. So the new approach is to put options where they belong, alongside the class or module they are used in.

There are some things still in Innate.options, but they have large impact on the whole system.

Options doesn't do things like merging env variables or parsing ARGV, these are things that Ramaze adds.

What makes Options especially useful is that you can use it in your own application to configure it without using diverse routes like putting config into counter-intuitive yaml files, using global variables, or relying on yet another dependency.

A small example:

module Blog
  include Innate::Optioned

  options.dsl do
    o "Title of the blog", :title, "My Blog"
    o "Syntax highlighting engine for blog posts", :syntax, :coderay
  end

  class Articles
    Innate.node '/'

    def index
      "Welcome to #{Blog.options.title}"
    end
  end
end

Controller

Away with controllers, long live the Node.

Nodes are objects that include Innate::Node and are then considered rack applications that get automatically hooked into the routing.

Since every existing Ramaze application relies on Ramaze::Controller and normal subclassing is not without merits as well we keep that part entirely in Ramaze, Controller simply includes Innate::Node and makes it suitable for the usage within Apps.

Layouts

Since layouts were an afterthought in Ramaze, they were made normal actions like every other on the respective controllers, leading to lots of confusion over the correct way to use layouts, the Controller::layout syntax in respect to the real location of layouts, how to exclude/include single actions, where layouts should be stored, how to prevent lookup from external requests, ...

I made layouts just as important as views and methods for the Action in Innate, and they have their own root directory to live in and will not be considered as a normal view template, so they cannot be accidentally be rendered as their own actions.

This strikes me as important, because I consider layouts to be superior to Ezamar elements and equal to render_partial or render_template, just about every application uses them, so they should be handled in the best way possible.

The root directory for layouts is in line with the other default locations:

proto
|-- layout
|-- model
|-- node
|-- public
`-- view

While none of these directories is necessary, they are the default locations and should be included in a new proto for Ramaze (except that ramaze uses /controller instead of /node.

Innate will not have project generating capabilities, so we just have to document it very well and provide some example apps.

Provides

This is a major new feature stolen from Merb and adapted for the ease of use of Innate/Ramaze. It won't have all the capabilities one might be used to out of the box, but extending them is easy.

Having "provides" means that every Action has different ways of being rendered, depending on so-called wishes, that's what people usually call content-representation.

A wish may be anything related to the request object, and by default it will trigger on the filename extension asked for. This enables you to create a single action that can be rendered in json/rss/atom/yaml/xml/xhtml/html/wap or different languages...

The dispatching in Node depends on the filename extension by default, but more decision paths can be added to Action by overriding some defaults.

Layouts map to wishes just like views, so every content representation can have a suitable layout as well.

This is very alien to old Ramaze, which always has a 1:1 mapping between actions and their views and how they are rendered, which made people wonder how to serve sass as css or how to respond with json for a AJAX request until they finally were pointed to setting content-type, using respond and setting up custom routes.

I hope this feature makes things simpler for people who care about it while it can be ignored by people who don't.

More specifics

Here I try to list the most important things that Ramaze will offer but that are not included in Innate itself in terms of globs:

  • cache.rb and cache/*.rb
  • current/response.rb
  • tool/{create,mime,localize,daemonize,record,project_creator}.rb
  • spec/helper/*.rb
  • snippets/**/*.rb
  • gestalt.rb
  • store/default.rb
  • contrib.rb or any contribs
  • adapter/*.rb (superseded by a lightweight adapter.rb)
  • template/ezamar*/*
  • bacon.rb
  • dispatcher.rb
  • dispatcher/*.rb

There might be a couple of things I've forgotten, but that's what a quick glance tells me.

Let's go through them one by one and consider what's gonna happen to them:

Cache

Caching is a very important concern and one of the most difficult things to get right for any web application.

Innate defines the caching API that enables everybody to add caches with very little code.

Innate provides caching by the following means:

  • DRb
  • Hash
  • Marshal
  • YAML

Ramaze adds:

  • Memcached
  • Sequel

And as time goes on there will be many more.

Response

Very little code, it just provides some options regarding default headers and easy ways to reset the response. Ramaze adds some more convenient methods.

Tools

Innate doesn't provide any of the stuff in ramaze/tool and it was removed from Ramaze as well, some of the reasoning is below.

Tool::Create

This has been used by bin/ramaze --create, bougyman is working on a new bin/ramaze which won't use it, so we removed it.

Tool::ProjectCreator

Dependency for Tool::Create, removed as well.

Tool::Daemonize

Nothing but issues with this one although it is just a very thin wrapper for the daemons gem. Nobody has contributed to this so far despite the issues and it seems that there are a lot of different solutions for this problem.

This was removed from Ramaze and Innate. You may use the daemonize functionality from rackup.

Ramaze::Record

Well, this might be the most obvious candidate for removal, maybe it can be revived as middleware. The functionality itself was in the adapter and even that's only a few lines. But so far I have never seen any usage of it.

Tool::Localize

Despite being quite popular it has many issues and is totally unusable if you don't have full knowledge about what is going to be served.

I and a lot of other people have used this over time and it has proven itself to be a very easy and straightforward way of doing localization.

It think it is better suited as middleware which can be included into rack-contrib and doesn't rely on the normal session but a simple unobfuscated cookie.

Innate does not attempt to do anything related to localization. Ramaze builds localization on top of Helper::Localize which does a much better job.

Tool::MIME

This one was removed, Rack::Mime is the way to go.

Spec helpers

Over the years, Ramaze has collected a wide variety of spec helpers that are not really compatible with each other and rely on real request/response with a running server.

Innate provides a better alternative via Innate::Mock for specs, the output formatting is done in a rake task.

There is some work in progress to integrate Innate and Ramaze with the rack-test library which will allow us to run specs with webrat (which is able to run using selenium and possibly watir in a DRY way).

Rack-test will provide us also with xpath checks, multipart requests, digest authorization, etc.

Snippets

Innate abandons the snippets, keeping your core clean.

Ramaze has still a lot of these snippets and will continue to, although I will constantly strive to reduce them slowly.

Gestalt

Gestalt has been the first "templating engine" for Ramaze and is still used in some fractions of the code and various applications. There are a lot of other html/xml builders out there these days so I think this is no good choice for inclusion into Innate. I will keep it inside Ramaze.

Ramaze::Store::Default

This has been removed from Innate and Ramaze. It started out as a simple wrapper for YAML::Store to make the tutorial easier, but I think it gives a wrong impression of being anything else.

It's very slow, might raise exceptions under heavy load and a plain YAML::Store or PStore or any other persistence mechanism is generally a better choice, so there is no need to keep this around.

Contrib

Ramaze has some things in contrib, some of them used widely, others not at all.

I'm still going through these, putting what is suitable for a wider audience into rack-contrib, including what fits into the Ramaze concept into Ramaze, but none of these will find its way into Innate.

One exception, the file cache has been added to Innate already and the sequel cache is in Ramaze proper.

The gzip filter has a more powerful and better maintained equivalent in Rack::Deflater.

The profiling hook is obsolete as well, there is an equivalent in rack-contrib.

The emailer is really light-weight and but lacks support for many things you will need, I'm working on a new Helper::Email that integrates with Mailit (a fork of MailFactory).

Things I'll not touch for now are facebook and gettext.

The sequel related stuff will have to be removed and might find a place somewhere else, (somebody start a sequel-contrib already!).

The gems was refactored and put into ramaze/setup.rb, where it provides you with painless first-start experience.

Adapters

These are entirely the responsibility of Rack/Innate now, Ramaze doesn't need to worry about that. WEBrick will remain the default adapter since it is in the Ruby stdlib.

Templating

The basic functionality for templating is provided by Innate, it only provides the None, Etanni and ERB templating engines. The other engines are in Ramaze.

Ezamar

Ezamar has become a standalone project. It has been stable since a long time and is suitable for other uses. The battle for the place of the default templating engine in Ramaze is still going on, competitors are ERB, Ezamar, and Nagoro.

Bacon

Bacon is still a dependency for specs, but we don't ship it anymore, the stable release includes all features we need.

Dispatcher

Innate uses a stripped down version of the Ramaze dispatcher. The Ramaze dispatcher was strongly influenced by Nitro, but proved to be a difficult part. We are now using Rack's URLMap directly, and have a minimal dispatching mechanism directly in Node (like we used to have one in Controller).

A lot of the functionality that used to be in the different dispatchers is now provided by Rack middleware.

The Dispatcher itself isn't needed anymore. It used to setup Request/Response/Session, which was superseded by Current, this again is now superseded by STATE::wrap.

We are going to remove all the other dispatchers as well, providing default ways to provide the same functionality, and various middleware to use.

Dispatcher::Action

This dispatcher was used to initiate the controller dispatching, this is now not needed anymore.

Dispatcher::Directory

This will also be removed. There is a directory listing middleware already.

Dispatcher::Error

There's middleware for this as well, and a canonical way of routing errors to other actions. This used to be one of the most difficult parts of Ramaze and it was removed to make things simpler.

Dispatcher::File

This is a combination of the ETag and ConditionalGet middlewares, ergo Innate and Ramaze will not serve static files themselves anymore, but leave the job to Rack or external servers.