Valentino L. wrote:
In the MVC architecture framework, the Model is actually communicate
with database. In my rails project, there is a model/ssh.rb that is
not
connect with database. Is it a good practice for this arrangement in
my
project? Am I violate the rules of MVC architecture? Or is it better
to
move all the codes from models to controllers? What do you think?
Thank
you.
MVC has two meanings. The most important meaning is a set of Guidelines
(not
“Rules”) for where to put what logic. For example, if you put HTML code
in the
Controller or Model, you had better be prepared to answer why.
MVC’s other meaning is the answer to these questions for most GUIs:
- do all your tests pass?
- is your code simple and easy to read?
- are you as DRY as possible without violating 2 (or 1)?
- is your code minimal, without run-on modules?
The game is to apply each item, in order, while not breaking the
previous items.
Don’t DRY your code (meaning Don’t Repeat Yourself) if you break tests,
or make
your code unreadable. Don’t remove lines of code if tests break? (That
is
“refactoring”.)
The ideal of MVC is that a good design that follows those guidelines
ought to
automatically fall out as three layers, with the data accessors in one
layer;
the “switchboard operators” in another layer, plugging everything
together, and
the views in another layer, stretching it all out for the users. So, in
theory,
if you apply those Simplicity Guidelines an MVC pattern ought to emerge
for you.
Put anything you like in the models folder, so long as it’s something
that
controllers mix and match together.
models/ssh.rb
$LOADED_FEATURES[$LOADED_FEATURES.length] =
‘net/ssh/authentication/pageant.rb’
require ‘net/ssh’
class Ssh
def self.execute(cmd)
ssh=Net::SSH.start(“ahnibm71”, “testlib”, :port => 19207,
:paranoid
=> false, :auth_methods => [“publickey”], :passphrase =>
“hellokitty”,
:keys => [“C:/testing/id_rsa”])
output = ssh.exec!(cmd)
ssh.close
if output.nil?
return “NULL”
else
return output.gsub("\n", “
”)
Firstly, this mixes HTML and Model. That
is not DRY, because it
repeats the
concept of a view down here. Remove that gsub, and render your output
like this:
<%=h output %>
Nextly, all your code should be pure XHTML (google assert_xpath for more
on this
subject!), so you should have used
. XHTML makes your pages
super-easy to
test, and this, in turn, keeps your code quality extremely high. Lack of
tests
is a much, much bigger problem than any variations in MVC ideals!
end
end
end
controllers/ssh_controller.rb
class SshController < ApplicationController
Does this control SSHs? What is the concept on the SSH wire that’s so
important?
Is this a “RemoteCommandController”?
def main
If you have a main.html.erb file, and if it takes care of itself, you
don’t need
this action in here.
end
def test
render :text => Ssh.execute(params[:command])
Exactly. Models are typically things that respond to Hash options, and
yours
does. But put the
tag up here, like this:
render :inline => "<pre><%=h Ssh.execute(params[:command])
%>
"
Why did I use :inline, instead of :text with a #{} masher? Because h()
is not
visible at this level, and :inline opens up a little bitty View inside
your
Controller.
Yet your existing design was already very DRY, so these are only
suggestions.
The only technical fix is the h().
Further, shouldn’t some other Model object intermediate between the raw
SSH
layer and the Controller? What concept is is the SSH object?
–
Phlip