On 10 March 2010 23:33, Josh Cheek <[email protected]> wrote: > On Wed, Mar 10, 2010 at 11:48 AM, Robert Walker <[email protected]> > wrote: >> >> 1. What about methods on models that change themselves in some way? >> >> <%= @post.last_viewed_at %> >> >> Suppose the last_viewed_at method returned a previously stored time, >> then updated the model to store a new current time. Maybe a bad example, >> but I hope you get my meaning. >> > > Don't worry about it. What the method does itself shouldn't be your concern, > in the view. You want your code to be orthogonal, it shouldn't matter how > the variable returns returns the date, or what it does when you request it, > that is the model's prerogative. Trying to keep track all over your > application of what your model methods are doing requires you to know and > consider their internal plumbing, this couples your code that you write to > the model's implementation. Something that won't bite you on a small app, > but will likely turn into a nightmare on a large app. > > If your view has some object that the controller gave you, just consider it > as an object you can access in whatever way that is necessary to perform the > responsibilities of the view. Saving data is not a responsibility of the > view, so that should not happen there. Displaying the date it was modified > may possibly be a responsibility of the view, so you can display that. If > the model decides that it needs to do something every time someone asks when > it was saved, your view shouldn't know or care or change it's behaviour > accordingly. > > Really, your view shouldn't even know it is an ActiveRecord object, it > should just be some object that has the information necessary to get things > done. Then you can swap it out with other variables later, maybe a struct or > an object pulled from a yaml file, or whatever. > >> 2. What about aggregating class methods like count, sum or avg? >> >> <%= Person.count %> >> >> Obviously a class methods and does touch the database. I assume it would >> be better to let the controller deal with stuff like this. >> >> Controller >> �...@person_count = Person.count >> >> View >> <%= @person_count %> >> >> Thoughts anyone? > > A variable is better here, because your view shouldn't know how to tabulate > the size, that is business logic. What happens if you later add another > another type of user, and it should treat them as the same? What happens if > different controllers want to render that same view to show their data? Your > view knows too much about the data it is serving, it works for Person.all, > but what if you add another type of person with different attributes, stored > in the model OtherPerson ? Then you cant use that view (or you will have to > change it, and change the controller for it). > > To keep your code robust, keep your views stupid. To keep them stupid, make > sure they are agnostic towards the implementation of displaying the data. > Let the controller worry about that, that is why the controller exists. > > http://www.youtube.com/watch?v=ku3QkWcPSEw > > On Wed, Mar 10, 2010 at 4:36 PM, Andy Jeffries <[email protected]> > wrote: >>> >>> > It's dirty, horrible, bad form, breaks the separation of layers... >>> >>> I don't know what you mean by dirty, it saves several lines of code >>> and when looking at the view code it is easier to see what is >>> happening than to see a variable that has to be hunted for in a filter >>> somewhere to find out what it is. >> >> It saves 4 lines of code, but breaks one of the principles of MVC layered >> separation. I'd say the 4 lines is worth it for keeping the application >> clean. > > But think how many lines of code you are going to have to go edit when you > realize that you need to change it. > > Also, I wouldn't consider Person.all to be more clean than @people. What if > you need to exclude some? Person.all :conditions => {whatever}, if you are > just using a before filter, it is easy to override, you can override it for > any given controller, and for any given controller method. If it's hard > coded into the view, then that view has to serve everybody's wishes, it ends > up having to know how it is to be used, and having lots of brittle > conditional code for each of these situations. > > This is why the controller must be responsible for supplying the appropriate > data to the view, not the view being responsible for creating it's own data. > > It might start as innocently as Person.all, it can easily turn into > if this > Person.all > elsif that > OtherPerson.all > else > Person.all + OtherPerson.all > end
An excellent post if I may say so that brings out the salient points I think. Can the issues be summarised as follows? 1. The controller should provide all the data that the view should display in instance variables (@person for example). 2. The view is expected to understand the structure of the objects and so can access attributes (virtual or otherwise) of the objects. 3. If the model needs to access the db in order to provide an attribute value, or accessing the attribute has some side effect that affects the db, then this is ok, providing the view does not 'know' that the side effect or db access is happening. (Not very well written but I hope you know what I mean). 4. The view must not call any method of the model who's purpose is to perform an action rather than return a value. 5. The view should not make any explicit use of model classes. For example there should be no reference to Person or any other model class Colin -- You received this message because you are subscribed to the Google Groups "Ruby on Rails: Talk" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/rubyonrails-talk?hl=en.
