Rails - Record Handling Tricks learned the hard way

- use dynamic model query methods described in: http://api.rubyonrails.org/classes/ActiveRecord/Base.html
- Story.order("column_ord").find_all_by_state("B").size returns the collection
- Story.order("column_ord").find_by_state("B").size returns only one story
- render(@collection) throws error if said collection is empty, so check <% if @collection then render(@collection) end %>
- assigns[:variablename] is how to get hold of a page object in the controller_test

Ruby: Stringception - a string within a string

good reference on all the various ways to define strings in ruby
http://www.techotopia.com/index.php/Ruby_Strings_-_Creation_and_Basics

So that you can run testing like:

assert_difference( %[Story.count :conditions => 'state = "P"']) do
put :drop, :id => story.to_param, :state => 'P'
end

Where the query string contains a string.

jquery w/ rails : ajax update of drag & drop item

A script added to my html.erb file, at page load it establishes sortable columns, and ajax call to automatically update an item when it is dragged from one column to another:

$(function() {
$(".sortcol").sortable({
connectWith: ".sortcol",
receive: function(event, ui) {
$.ajax({
url: "<%= drop_url() %>",
type: 'PUT',
data: $.param({id : ui.item.attr('id'), state : ui.item.parent().attr('id')}),
success: function(d) { ui.item.effect("pulsate", {} , 500 ) }
})//end get
} //end receive function
}); //end sortable
}); // end function

The drop_url references as routes.rb entry /board/drop to a specific board_controller method. Upon retrospect, I'm thinking this ought to work with the default update method on my :story resource. Note to self, if your update method doesn't update... check that you haven't forgotten your parameter naming conventions. status != state no matter how you write the method to update.

Rails -migrations with column default changes

So wanted to add a default value for a column, but also wanted to do a proper rollback method on the migration. My problem may have been more with SQLite than rails, something about it not liking alter column commands.

Anyway:
change_column :table, :column, :type, default = 'value' did update the schema.

But also needed to add the date
Table.update_all( "column = 'value'", "column IS NULL") to fix unset values.


Didn't find a good syntax to remove the default setting from the schema.

Rails - Ruby: lesson in MVC

So turns out if you apply MVC correctly you save yourself from lots of unnecessary objects!

Originally I thought I needed a Board --> Panel --> Story. But as mentioned last post, using poro instead of ActiveRecord was getting 'interesting', as I had decided I didn't need any state for Board and Panel. As I was fixing tests... suddenly realized how all this was very unnecessary work:

The better solution is to create a BoardController, and then define the panels as simple attributes of that controller that reference the appropriate collection of stories:

class BoardController < ApplicationController
 def show
  @backlog = Story.find :all, :conditions => 'state IS NULL'
   respond_to do |format|
    format.html # show.html.erb
    format.xml { render :xml => @stories }
    end
  end
end

Now instead of needing to keep a bunch of keys, names in alignment between the view, board, panel objects... I just have simple associations. Each named collection I want is an attribute in the board controller, available to the view. The view simply needs to render each attribute appropriately. This approach also seem to have simplified my views, since now I just have the 'board' view, and a '_story' partial view to render each collection element.

Ruby-Rails : lessons on rails ActiveRecord vs plain old rails objects

Got distracted from doing the Agile-Web-Development-With-Rails examples, am now starting from scratch on an experiment with jquery and rails. I'm looking to see what I can achieve from a robust client app with rails based data & services via ajax.

1) Unit Tests throwing " ActiveRecord::StatementInvalid: Could not find table 'panels' "

Kept seeing this error for several basic assert statements in the model unit tests.
> assert p.is_a(Panel)
> assert p.display_text == 'backlog'

Noob mistake here... using 'rails g model' for objects that will not be stored in the database. Caused the objects to inherit from ActiveRecord class, and when ActiveRecord triggers automagic that expects a table to exist for the Panel class, and complains when it doesn't.

Instead, I really wanted to be dealing with plain old ruby objects, and viola! once I deleted the ActiveRecord reference in the model object.

> couldn't use is_a, instead had to use ruby method is_a?
> have to define the accessor methods for private instance variables.

Rails - Task I: Logging In - Noob notes on authlogic

There's a little auto-magic going on that as a rails/ruby noob I didn't expect.

1) The password, password_confirmation fields are not explicitly declared anywhere except for your registration, login views. This mean you have to add the "acts_as_authentic" to your User class BEFORE you can being using these fields in your views and tests.

2) Naming of your DB fields is important, password_confirmation != confirm_password. (Coming from a Java background, this feels weird, not having name control over model fields. Just go with it.)

3) After you do add acts_as_authentic, you have to update user_controller_test. Passing user.parameters doesn't work anymore, since that's the DB model and not what the user will send -- example of the changes that worked for me.

test "should create user" do
assert_difference('User.count') do
post :create, :user => {:email => 'test@functional.com', :password => 'godpassword', :password_confirmation => 'godpassword'}
end

4) Really enjoying the automated testing, they have been catching all kinds of mistakes I make in ignorance, and a great teaching tool. Am beginning to see how much you can skip explicit tutorials and APIs if the testing is robust.

Rails - Task I: Logging In w/ authlogic_example

Picking up on a new branch, and take a few integration shortcuts to get authlogic up and running.

> git pull git://github.com/trevmex/authlogic_rails3_example

Then working through the merges necessary to merge in authlogic code, with a few skips
* git checkout --ours db/schema.rb
* git checkout --ours db/seeds.rb
* git checkout --theirs Gemfile
* change config/initializers/rspec_generator.rb Replace AuthlogicRails3Example with Depot (or your app name)

Then after code looked reasonably merged

> bundle install
> rake

Get complaints about unknown email_format validator

Seemed to have missed an important set of entries in config/application.rb

#Custom directories with classes and modules you want to be autoloadable.
# config.autoload_paths += %W(#{config.root}/extras)
config.autoload_paths += %W(#{config.root}/lib)
config.autoload_paths += %W(#{config.root}/lib/**/)

This is required to reference the format validator introduced with the AuthLogic Example.

After this fix we get to something interesting... the rspec tests fail!

Rails - Task I: Logging In

For this task I decided to diverge from the book. From what I've heard attending Atlanta's local chapter meetings for OWASP (www.owasp.org), own-rolled authentication is at the root of far too many security flaws and breeches in web applications. Granted this book touches on some of the oft overlooked necessities (e.g. salted hash passwords), but for a real live website, I always push to use a robust, tested, peer-reviewed plugin over hand rolled code. At the very least, typical project timelines leave little time for developers and testers to stay current on the ever evolving threat landscape. If your shop is different, please, please help out the rest of us.

I looked around a bit and decided to go with authlogic. Trevmex has provided a fully Rails3 ready example found at https://github.com/trevmex/authlogic_rails3_example . http://www.dixis.com/?p=352 rewrites the tutorial for a few other steps you need to then incorporate the module into the depot code. (I did start first trying more commonly referenced authlogic_example by binarylogic at https://github.com/binarylogic/authlogic_example, but all I succeeded at was proving what I complete noob I am)

What follows is a combination of various sources. Comments indicate what prompted each step to be taken, but not the rationale or theory behind it.

Pre-Iteration Work: Adding authlogic to depot
$ sudo gem install authlogic #binarylogic

To depot Gemfile add:
# Add support for Authlogic authentication
gem 'authlogic', :git => 'git://github.com/odorcicd/authlogic.git', :branch => 'rails3' #trevmex
gem "rails3-generators" #dixis
> bundle install #'cause rake prompted me to run this after the changes.

> rails generate authlogic:session UserSession #dixis

>rake test

And viola! everything now fails because of 'ActiveRecord::StatementInvalid: SQLite3::SQLException: no such table: user_sessions: DELETE FROM "user_sessions" WHERE 1=1'

bummer!

Grails with Aptana Studio 3 on Mac OSX

Setting JAVA_HOME: (http://www.oreillynet.com/mac/blog/2001/04/mac_os_x_java_wheres_my_java_h.html )
setenv JAVA_HOME /System/Library/Frameworks/JavaVM.framework/Home

Setup Grails (http://www.grails.org/Installation)
git clone git://github.com/grails/grails-core.git
Set the GRAILS_HOME environment variable to CHECKOUT_LOCATION/grails-core
Add the $GRAILS_HOME\bin directory to your PATH environment variable
Go to the GRAILS_HOME directory and run: ./gradlew libs

In Aptana Studio Workspace
grails create-app grails-demo
New Project 'grails-demo'

Then in the grails-demo project
> grails run-app

Note that grails commands feel much slower than the equivalent ruby commands, and initial application startup is significantly slower. I actually was beginning to wonder if everything was working as the browser spun.. patience paid off and I was rewarded with the initial welcome page.

Also attempted to add some sort of groovy support, but without any real success. Java editor helps with syntax highlighting of terms in the groovy files, but is cluttered reporting other syntax and compile errors 'cause doesn't understand groovy.

Note that 'natures' do report several groovy natures, but these do not appear to be associated with any file types.

Rails - Iteration E3 -Fixture tests for prices

Need further study: http://api.rubyonrails.org/classes/Fixtures.html

When I tried to follow simple example e.g.

one_quantity_two:
product_id: one
cart_id: 2
quantity: 2

line_item.product was a nil object.

The format that did work

one_quantity_two:
product_id: <%= Fixtures.identify(:one) %>
cart_id: 2
quantity: 2

Most likely culprit --> rails/ruby version?

Iteration E2: Functional Test for bad cart

test "should redirect on invalid cart" do
get :show, :id => 'wibble'
assert_redirected_to store_path
assert_match flash[:notice], "Invalid cart"
end

TBD: use internationalized error message?

FYI - cause of the critical bug?

bad --> @line_item = @cart.add_product(:product.id)
good --> @line_item = @cart.add_product(product.id)

First case not passing id, instead passing (I think) an object reference! gotta love those one character typos!

Rails - Iteration E1 - line_item aggregation starts new

Take 2 on smarter cart!

First: add a unit test for the cart.add_product method

test "cart aggregates line items" do
c = carts(:one)
assert_equal 0, c.line_items.count, "cart items not empty at start of test"
p = products(:ruby)
item = c.add_product(p.id)
assert_equal 1, c.line_items.count, "first product add did not create new line item"
assert_equal 1, item.quantity, "new item quantity not 1"
item = c.add_product(p.id) #hold reference for future testing
assert_equal 1, c.line_items.count, "readd product caused increase in line items count"
assert_equal products(:ruby).id, item.product.id, "a problem with product id between line item and orig"
end

(Note to self: originally tried to do above like:
- assert_difference('c.line_items.count') do
- p = products(:ruby)
- c.add_product(p.id)
- c.save
- end
Note that the assert_difference requires a DB save event. Decided I didn't like that extra dependency on a unit test. Also note that the assert_difference requires a string not an object expression.)

Second: modify line_item_controller test to show aggregation of line items

test "should create/update line_item" do
#original test when just creating new items (iteration D)
assert_difference('LineItem.count') do
post :create, :product_id => products(:ruby).id
end
assert_redirected_to cart_path(assigns(:line_item).cart)
#new test - just repeat old and this time no_difference
assert_no_difference('LineItem.count') do
post :create, :product_id => products(:ruby).id
end
assert_redirected_to cart_path(assigns(:line_item).cart)

now when I complete the line item tests everything (including listing of cart after add) works as intended.

Rails - Iteration E1 - recovery from new defect introduced

Interesting, at end of last class all of my tests passed, but today I test my app and find that I have a critical defect!

Post /lines_items?product_id=4 fails horribly when trying to show the cart page with updated list.
Cart page complains about no 'title' attribute for a nil object
/line_items shows that my cart now has a line item for a product id 123456 <- which of course doesn't exist.

Old instincts: well time to add a check for null products and skip processing anything won't work.
New TDD instincts: Whoa! how can all my tests be passing? Why is the product id changing on add_product?

Step 1: Recreate the defect in a test.
Looks like my line_items_controller_test only confirms line item addition and redirect to cart page. But defect manifest at cart page....
Decided to try my hand at 'integration' test, since now testing flow from one controller to another.
> rails generate integration_test user_cart_session_test

Then add test method to the resulting skeleton:

test "line item adds to cart" do
item = products(:ruby)
post "/line_items", :product_id => item.id
follow_redirect!
assert_response :success
end

> rake test:integration --> Same problem as reported in the app!

Last Step: Rollback to a clean branch of code and start lesson over.

> git commit -a -m "test for my broken code"
> git checkout master
> rake test --> lots of problems? Wait the DB was mucked with
> rake db:drop
> rake db:migrate
> rake test --> ah! that's better!
> git checkout class.chapter10 test/integration/ --> pull over my new fangled test
rails test --> look now it's passing, got rid of my bug!
git add test/integration/
git commit -m "adding a functional test for cart and line items"

Note To Self: never ever forget to branch when starting new development! Love this cycle
1) break code unexpectedly
2) build test to reproduce break
3) revert code, import test and start dev over again (and always have access to previous work!)

Rails4 - Iteration D3 - functional tests for new cart button

When building the store page, we went through several examples of adding functional tests to confirm the store index page contained expected page elements. Seemed logical to extend these existing tests to track the addition of the cart button. In doing so learned a couple of interesting features of assert_select

1) use a block to test that each .entry now displays button "Add to Cart"
assert_select '.entry' do
assert_select '.button_to input[value=?]', /Add to Cart/
end

2) check that each form for Add to Cart contains the product_id information, in this case using a substitution marker ? to check the action contents.
assert_select '.entry form[action=?]', /.*product_id=\d+.*/

Note: the substitution regex did not recognize partial matches. For example, /product_id=\d+/ failed to find any elements because 'product_id' is only a partial match to the actual action value /line_items?product_id=4 needed to add .* at start and end to pad out for the remaining action text.

Note for future study: How do I put variables into the regex? for example might be nice to check that the action also contains correct target 'line_items' but I don't really want to hard code that. Instead want to test with line_items_path variable?