Scalable Web Framework aimed at Speed and Simplicity
$ [sudo] gem install e
require 'e' class App < E map '/' def index # ... end end
Espresso in the wild?
Yes, CIBox service is built entirely on Espresso Framework.
Route | Params | Passing Control | Fetching Body | Halt | Redirect | Reload | Streaming | Error Handlers | Hooks | Authorization | Sessions | Flash | Cookies | Content Type | Charset | Cache Control | Expires | Last Modified | Accepted Content Type | Cache Manager | Send File | Send Files | Attachment | Headers
Highlights / Motivation
In terms of performance, the only really important thing for any framework it is to add as low overhead as possible.
The overhead are the time consumed by framework to accept the request then prepare and send response.
The tests that follows will allow to disclose the overhead added by various frameworks.
The overhead are calculated by dividing 1000 milliseconds to framework’s standard speed.
The framework’s standard speed are the speed of HelloWorld app running on top of given framework.
The framework’s standard speed means nothing by itself. It is only used to calculate the framework’s overhead.
Tested apps will run on Thin web server and will return a trivial "Hello World!" response.
Processor Name: Intel Core i5 Processor Speed: 3.31 GHz Number of Processors: 1 Total Number of Cores: 4 Memory: 8 GB
To run tests on your hardware, clone Espresso Framework repository and execute
rake overhead inside it.
--- Speed Overhead 1ms-app 5ms-app 10ms-app 20ms-app 50ms-app 100ms-app espresso 5518 0.18ms 847 193 98 49 19 9 sinatra 3629 0.28ms 783 189 97 49 19 9 rails 792 1.26ms 442 159 88 47 19 9 ---
1ms-app shows your app speed when your actions takes 1ms to run.
10ms-app shows your app speed when your actions takes 10ms to run.
The app speed are calculated as follow:
1000 / (time taken by action + time taken by framework)
So, if your actions takes about 1ms and you use a framework with overhead of 0.18ms, the app speed will be:
1000 / ( 1 + 0.18 ) = 847 requests per second
However, if framework's overhead is of 1ms or more, the app speed will decrease dramatically:
1000 / ( 1 + 1.26 ) = 442 requests per second
The framework speed matter only if your code matter.
If you develop a site aimed to serve a serious amount of requests, you should write actions that takes insignificant amount of time.
Only after that it make sense to think about framework speed.
These tools may help you to dramatically reduce the time consumed by your actions.
I never understood why should i create actions in some file,
then open another file and directing requests to created action.
Even worse! To use params inside action, i have to remember how i named them in another file. And when i want to change a param name i have to change it in both files?
What about consistency?
A good tradeoff would be to use some DSL.
get '/book/:id' do params[:id] end
Looks much better.
But! Strings/Regexps as action names? No, thanks.
What if i need to remount a bunch of actions to a new root? Say from /news to /headlines? Refactoring? Using vars/constants in names? No, Thanks.
How do i setup multiple actions?
How do i find out the currently running action?
What if i do a request like "/book/100/?id=200"? What? Should i use unique param names? No, thanks.
And why should i remember so many non-natural stuff?
Is not Ruby powerfull enough? I guess it is:
def book id end
That's a regular Ruby method and a regular Espresso action.
That's also an Espresso route. Yes, the app will respond to "/book/100"
And of course action params are used naturally, through method arguments(
id rather than
And all this offered by Ruby for free! Why to reinvent the wheel?
Usually you do not want to instruct each action about how it should behave.
It would take eras to define inside each action what content type should it return or what layout should it render.
Instead, you will use few lines of code at class level to write instructions that will be followed by all actions.
Example: Instruct all actions under
App controller to return JSON Content-Type
class App < E content_type :json # ... end
But what if you need to setup only specific actions?
Simple! Put your setup, well, inside
setup block and pass action names as parameters.
Example: Instruct only
feed actions to return XML Content-Type
class App < E setup :rss, :feed do content_type :xml end # ... end
Well, what if i need some setup for some 10 actions ad another setup for another 20 actions?
Should i pass 30 arguments to
setup? I do not want to buy a new keyboard every month...
That's simple too. Use regular expressions.
Ex: setup only news related actions:
class App < E setup /news/ do # some setup end # ... end
Portability and DRY done right and easy.
With Espresso, any controller can be mounted under any app.
Even more, any set of controllers - a.k.a. slices - can be mounted under any app.
To create a slice simply put your controllers under some module.
Then you can mount that module under any Espresso app.
Even more, when mounting you can easily setup all controllers(or some) at once.
And of course when mounting, you can give a mount point.
module Cms class Articles < E # ... end class News < E # ... end class Pages < E # ... end end app = Cms.mount do # some setup that will run inside each controller end # or app = Cms.mount do |ctrl| # some setup that will run inside controllers that match `ctrl` param end app.run
By default, verbless actions will respond to any request type.
To make some action to respond only to some request type, simply prepend the corresponding verb to the action name.
# will respond to any request type def book # ... end # GET def get_book # ... end # POST def post_book # ... end # PUT def put_book # ... end # etc.
With Espresso built-in rewriter you can redirect any requests to new URL.
However, beside trivial redirects, rewriter can also pass the control to an arbitrary controller#action or simply halt the request and send the response.
For most web sites, most time are spent at templates rendering.
When rendering templates, most time are spent at reading and compiling.
Espresso allow to easily skip these expensive operations by keeping compiled templates in memory and just render them on consequent requests.
If you have some expensive operations that basically return static data,
simply put them inside
The result will be cached and returned on consequent requests.
To clear cache, simply call