[rspec-users] RSpec makes me want to write better code

[Ashley Moran]

On 26 Sep 2008, at 12:49, Matt Wynne wrote:

> Would you mind elaborating on why you don't like these? I'm pretty  
> new to rails (but not programming generally) and rather naive about  
> such things!

It's quite hard to explain briefly, but basically it makes the  
predicate (interpretation of the table) extremely difficult.

A normal table like

  cars
   ---------
   reg_plate
   owner_id
   purchase_date

Could be interpreted as "The car with reg plate <reg_plate> was bought  
by the person with ID <owner_id> on <purchase_date>" (ok, they might  
have sold it since, but...)

However, from the Rails wiki[1]:

  addresses
   ---------
   street
   city
   country
   addressable_id
   addressable_type # <- is a string

How do you interpret this?  The relationship is fundamentally  
different depending on the value of addressable_type, and is much  
harder to enforce as an integrity constraint*.

There's another angle you can take, based on data types - you can  
define a type PERSON_ID for the attribute 'owner_id' (and re-use it in  
the 'departments' table as 'manager_id'), but the type of  
'addressable_id' depends on the value of 'addressable_type'.  How can  
a value not know its own type?

Easier solution:

  people
   ---------
   name
   delivery_address_id
   billing_address_id

"The person called <name> wants items delivered to the address with ID  
<delivery_address_id> and invoices delivered to the address with ID  
<billing_address_id>".



> Also why is the article so down on STI? What are the drawbacks? What  
> do people use instead?
>
> [1]http://www.matthewpaulmoore.com/articles/1276-ruby-on-rails-code-quality-checklist

I'll reply to Mark...

TBC

Ashley

[1] http://wiki.rubyonrails.org/rails/pages/UnderstandingPolymorphicAssociations

* pah, who needs their data integrity protected anyway?

-- 
http://www.patchspace.co.uk/
http://aviewfromafar.net/