The normal controller/view flow is to display a view template corresponding to the current controller action, but sometimes we want to change that. We use render
in a controller when we want to respond within the current request, and redirect_to
when we want to spawn a new request.
The render
method is very overloaded in Rails. Most developers encounter it within the view template, using render :partial => 'form'
or render @post.comments
, but here we'll focus on usage within the controller.
We can give render
the name of an action to cause the corresponding view template to be rendered.
For instance, if you're in an action named update
, Rails will be attempting to display an update.html.erb
view template. If you wanted it to display the edit form, associated with the edit
action, then render
can override the template selection.
This is often used when the model fails validation:
def update
@book = Book.find(params[:id])
if @book.update_attributes(params[:book])
redirect_to(@book)
else
render :action => :edit
end
end
When render :action => :edit
is executed it only causes the edit.html.erb
view template to be displayed. The actual edit
action in the controller will not be executed.
As of Rails 3, the same effect can be had by abbreviating to render :edit
.
Most commonly you want to render the template for an action in this controller. Occasionally, you might want to render an action from another controller. Use a string parameter and prefix it with the other controller's name:
render 'articles/new'
Even if this were executed in the CommentsController
it would pull the view template corresponding to ArticlesController#new
.
You can use render
to display content directly from the controller without using a view template.
You can render plain text with the :text
parameter:
render :text => "Hello, World!"
This can be useful for debugging but is otherwise rarely used.
You can render XML or JSON version of an object:
render :xml => @article
render :json => @article
Rails will automatically call .to_json
or .to_xml
on the passed object for you.
When using render
you can override the default layout with the :layout
option:
render :show, :layout => 'top_story'
Or turn off they layout system completely:
render :show, :layout => false
Use redirect_to
to spawn a new request.
Why care about a new request? When a user submits data it comes in as a POST
request. If we successfully process that data we likely next display them the data they just created. If they wrote an article and click SAVE, then we'd probably show them that article. We could display the article using render
in the same POST
that sent us the data.
But, what's going to happen if they hit refresh? Or click a link, then use their browser's BACK button? They'll get a pop-up from the browser: "Submit form data again?" Do they push yes? No? Will clicking yes create a duplicate article? Will clicking no somehow mess up the old article? It's confusing for the user.
Instead, when you successfully store data you want to respond with an HTML redirect. That will force the browser to start a new request. In our scenario, we'd redirect to the show
action for the new article. They could refresh this page, navigate forward then back, and it would all be normal GET
requests -- no warning from the browser.
The redirect_to
method is typically used with a named route helper:
redirect_to articles_path
If you're linking to a specific resource outside your application, you might use a full URL string:
redirect_to 'http://rubyonrails.org'
By default Rails will use the HTTP status code for "temporary redirect." If you wanted to respond with some other status code, you can add the :status
parameter:
redirect_to 'http://rubyonrails.org', :status => 301
The request would be marked with status code 301, indicating a permanent relocation.
You can set a flash message within your call to redirect_to
. It will accept the keys :notice
or :alert
:
redirect_to articles_path, :notice => "Article Created"
redirect_to login_path, :alert => "You must be logged in!"
Keep in mind that redirect_to
does not cause the action to stop executing. It is not like calling return
in a Ruby method.
Here's how that could go wrong. Imagine you have a delete
action like this:
def destroy
article = Article.destroy(params[:id])
redirect_to articles_path, :notice => "Article '#{article.title}' was deleted."
end
Then you begin adding security to your application. You've seen "guard clauses" used in Ruby code, where a return
statement cuts a method off early. You decide to imitate that here:
def destroy
redirect_to login_path unless current_user.admin?
article = Article.destroy(params[:id])
redirect_to articles_path, :notice => "Article '#{article.title}' was deleted."
end
When an admin triggers destroy
, here's what happens:
- The
unless
condition istrue
, so the firstredirect_to
is skipped - The article is destroyed
- The browser is redirected the the index
Then some non-admin user comes and triggers the destroy
action:
- The
unless
condition isfalse
, so theredirect_to
runs, a redirect response is set, and execution continues - The article is destroyed
- The second
redirect_to
runs, it sees that a redirect has already been set, and raises an exception (AbstractController::DoubleRenderError
)
The article gets destroyed either way. The redirect_to
does not stop execution of the method, it just sets information in the response. The correct way to achieve this protection would be:
def destroy
if current_user.admin?
article = Article.destroy(params[:id])
redirect_to articles_path, :notice => "Article '#{article.title}' was deleted."
else
redirect_to login_path, :notice => "Only admins can delete articles."
end
end
- RailsGuide on Layouts and Rendering: http://guides.rubyonrails.org/layouts_and_rendering.html
@raquelhortab you found a 12-year-old gist!! ;)
No, you typically would not want to execute a second action within a single request. I'd encourage you to think about it a bit differently. Like say you're running the
create
action but are thinking you also want to run thenew
action. Instead, what is common about them that you're wanting to run in both? Extract that to it's own method. Maybe that's something like anauthenticate
method that gets called from bothcreate
andnew
, then the actions don't need to rely on / call each other.Does that make sense?