Dead Simple Model Diagrams for Your Rails Project

While working on Rails project I often find myself wanting a visual representation of my model classes.  I usually grab a notebook and manually write them out. Depending on the project, it can take time.

I started searching for a diagramming tool that might be easier and faster than writing out by hand, there are a bunch of them out there.  Most have a steep learning curve and are expensive.

A bit of searching around the web for a Ruby-specific tool lead me to a gem named rails-erd.  Maybe you have heard of it, maybe I’m the last to know, but regardless it is a nice find.

Installing

The gem relies on GraphViz to do it’s drawing magic.  There are a multitude of ways to install it, I used Homebrew:

brew install graphviz

Add the gem to your development group in your Gemfile:

group :development do
 gem 'rails-erd'
end

Don’t forget to run the bundle command.

When everything is install, from the root of your Rails project simple run:

rake erd

When the rake task runs, watch the output from the tool. It tells you items you won’t find on the diagram either because it’s not used or a relationship isn’t right.

The Output

The result will be a PDF file in the root of your project that looks something like this: 

Erd

As you can see, it gives a very nice model diagram with all the relations and properties. Just what I was looking for.

The tool is very customizable and the web site outlines everything that can be changed.  I haven’t looked very much at this aspect since it produced everything I needed the first time.

Minitest Not Running My Unit Tests

I have started using minitest for Rails 3.x unit testing and since it comes with Ruby 1.9.x by default, it seems like a good direction to take my testing.

I am using the great gem by Mike Moore called minitest-rails which has a minitest dependency so I can easily have the latest version of minitest.  This gem helps to more cleanly integrate and use minitest in our Rails 3.x projects.

As part of my setup I use Autotest so my tests run automatically when files change.  I noticed none of my unit tests were running and upon further inspection, minitest was looking in test/unit and test/unit test directories but not in the test/models folder, which seems to be the minitest default.  

Opening up my .autotest configuration file and adding the following:

Autotest.add_hook :initialize do |at|
 at.add_mapping(/test\/models\/.*\.rb/) do |_, m|
  at.files_matching %r%^test/(models)/.*\.rb$%
 end
end

After restarting autotest, my unit tests are now being tested.  I hope this saves someone some time.

Damn You Rails Multiparameter Attributes

Boy with a headache MG 0599

I came across an interesting problem that was driving me crazy when using Ruby on Rails 3.2 Date types in an application I am working on for client.

Problem

I have a date property that is virtual and not backed by a column in a database. When trying to create a new object from a date select on a form, I was being greeted by the following error:

ActiveRecord::MultiparameterAssignmentErrors in Users::MembershipsController#create

1 error(s) on assignment of multiparameter attributes

After some searching around the web for a solution, as any self-respecting developer does, and came up with many others facing the same problem.  It was suggested this was a problem with Rails, a quick check on the Github Rails project revealed something similar reported, but no solid fix I could find.  It may be out there and if someone is aware, please let me know.  I am using Rails 3.2.8 so any fixes that exist, should be in there. 

This works great when using the date select and storing to a database, Rails takes care of processing multiparameter attributes and pushing into the date field.   We are talking about virtual attributes here, no database field to store the data.

Solution

Please don’t comment how bad this solution is..it’s a hack, I know, but it works. I’m never too proud to share a hack. 

The goal here is to end up with and expiration date in a virtual attribute on my model.  To accomplish this I construct a plain Ruby Date class from the components of the date from the date select form helper.  Ruby Date expects parameters; Date.new(Year, Month, Day)

NOTE: if you try this look at the parameter values for each component of your date to make sure choose the right values.  I have changed the default order of the date select on the form.

params[:user][:membership_attributes][:expiration_date]=Date.new(
params[:user][:membership_attributes][:"expiration_date(1i)"].to_i,
params[:user][:membership_attributes][:"expiration_date(2i)"].to_i,
params[:user][:membership_attributes][:"expiration_date(3i)"].to_i)

Now remove the individual date components from the parameter hash:

params[:user][:membership_attributes].delete(:"expiration_date(1i)")
params[:user][:membership_attributes].delete(:"expiration_date(2i)")
params[:user][:membership_attributes].delete(:"expiration_date(3i)")

In my case this is strictly for a Ruby Date type in Rails but the problem and solution is the same with a Ruby DateTime type.  The date and time are broken down more, having 4i and 5i representing the time.

Finally

This little hack works great and hopefully helps those using a version of Rails 3 that is not patched..or heck, maybe it will never be.

I’d be happy to learn this was fixed or how I could have handled this better.  Please add some details in the comments.

 

Rails Views and Backbone.js with David Heinemeier Hansson

This is a great discussion with DHH about the Basecamp rewrite, not going crazy using Backbone.js, like so many developers are doing these days, for everything.  One technology used is PJAX, for updating small bits of a view to give small, fast updates.  PJAX combines HTML5 pushState and AJAX to make the fast magic happen.

PJAX is a great example how to use the views we know and the fast updates we need instead of completely changing the way Rails developers approach views using tools like Backbone.js, Knockout and others.

The discussion is about 1 hr 40 min long but well worth it.  You get some good views of new Basecamp Rails code and the usual colorful dialog.  

David had a great post on the 37Signals blog that serves as some background to the above talk titled How Basecamp Next got to be so damn fast without using much client-side UI.

RailsConf 2012 Wrap Up

IMG 0370

I was fortunate enough to be able to work out attending RailsConf 2012 in Austin, TX.   This was the first time on many years that the conference was not organized by O’Reilly but rather Ruby Central, Inc.

I have to go on record and say I usually avoid cities but the city of Austin is a great place and would not hesitate to return.  The people are friendly and there is so much diversity in the city that there is something new on each corner.  I noticed an abundance of restaurants with so many different types of food.  I can’t say I had a single bad meal during my journey.  Everyone I spoke with about the trip said I had to try the BBQ, and they were right…it was fantastic.

IMG 0379

Many of the sessions overfilled the room.  This on in particular exemplifies what I’m talking about.  I bet the fire marshal wasn’t aware of these.  Overall the floors were pretty comfortable.

It was often difficult to decide which sessions to attend, with 3 full-tracks there always seemed to be two talks during the same time slot I wanted to take.  I usually decide which sessions to attend by how applicable they are to current work.

One of my favorite sessions was by Obie Fernandez about using Redis with Rails.  Although the examples of the talk were from his recent startup, they were excellent and showed integrating Redis into a Rails application not to remove ActiveRecord but to compliment it.  Obie discussed a gem he released to help the integration called redis_props along with sample code used in the talk.  The code is clean and concise…great stuff.

Another talk I found personal value in was the Semi Automatic Code Review by Richard Huang.  Richard is the creator and maintainer of the Rails Best Practices gem.  In the talk he discusses another related open source project called Railsbp.com which allows for your code to be reviewed when committing to Github.  The results will be displayed on the Railsbp.com site where you can change the code right there and commit back to your repo.  Very informative details produced from the site, GitHub allows hooks into the service and thoughtfully open sourced.  I wasn’t aware of the site before but now I am using it regularly.

The other talk which I took a lot away from was Digging Deep with ActiveSupport::Notifications by Matt Sanders.  This talk when into great detail with many examples of using notifications in your applications.  It is similar to the event publishing and subscriber model from other platforms such as .NET.  Having spent many years writing .NET applications this talk brought back many memories of this pattern.  The techniques exemplified here I had never used in Rails but do need this functionality on a new project.  

UPDATE (05/03/2012): One talk that was intended to be included here, is from Lori Olson.  Her talk titled, Mobile Rage – What causes it & how to fix it (Confreaks), takes the view of web application use on a mobile device from the user’s perspective and how developers can implement very simple techniques to ease the pain.  I recommend this one highly, good stuff and some tips I was not aware of.  I admit I have some sites that can take advantage of this.  

IMG 0387

The final keynote of day one was from a non-Ruby developer, Rich Hickey, which seemed to be out of the ordinary.  Maybe he was there to pull some Ruby developer to the Clojure world.  It appears Rich is trying to convince these two Rubyist that LivingSocial would be better with Clojure.  I wish I could have overheard the conversation.

There were three very large, two-sided, white boards used for companies to post jobs, and they were pretty full of opportunities.  I noticed there were far too many companies attempting to make the next Facebook or Twitter and not enough companies creating really useful applications.  There were exceptions from what I could see, but too few.  I remember the same thing happening around 2000 and then the bubble burst.  Apparently we are not better from this event in history because we have not learned from it.

I finally met face-to-face many friends I only knew from various social networks with lively hallway track discussions.  I think this is the #1 reason to attend conferences.  The materials from the talks are available everywhere and with Confreaks recording all the sessions, you can watch the show later.  You can’t however, experience meeting new friends and seeing old ones without attending.

I recommend every Rails developer attend just one of these events, well worth the time and effort.  The next on is in Portland, OR from April 29 to May 2, 2013.

Palliative Symptoms Survey Hits the Apple App Store

I have been working diligently on a project for some time now and it’s finally available.  My company, Still River Software, received approval from Apple last week for Palliateive Symptoms Survey to help doctors and caregivers provide better and faster care to their patients.

Palliative Symptoms Survey is an application based on the Edmonton Symptom Assessment Survey (ESAS-r) which was developed to assist in the assessment of nine symptoms that are common in palliative care patients: pain, tiredness, drowsiness, nausea, lack of appetite, depression, anxiety, shortness of breath, and wellbeing There is also a blank scale for patient-specific symptoms.

The application is a native iPad application written in Objective-C with a back-end using Ruby on Rails.   Please read a little bit about the project on my company web site.

When Pow Eats Up the Clock Cycles

I have been a huge fan of Pow to serve my Rails applications for development on my Mac.  I wrote my love for Pow a while back and have been using it ever since, but all has not been a perfect experience along the way.  After upgrading to the 0.3.2 version I started to have my applications appear to lock up in the browser and finally timing out.

The only solution I could reliably make work was to open up Activity Monitor and kill the pow process manually.  It worked..until the next time and I would have to do it again and again.  It seemed like I was doing this many times a day.

It turned out this was a bug reported to the development team and is a confirmed issue.  The cause of the problem is determined to be the dns resolver, ndns.    A pre-release version, 0.4.0-pre is reported to resolve the issue.  Install it with the command:

curl get.pow.cx | VERSION=0.4.0-pre sh

I did this and have not had to manually restart pow once.  I’m not sure why the official release has not been updated to include the fix to this issue which seems to effect many users.

 

ASP.Net MVC vs Ruby on Rails Smackdown Results

I’m sure everyone saw this already, right?  In case you didn’t then it’s worth a look.

It amazes me (though it shouldn’t) how far a community of developers can take something vs. a software giant like Microsoft.  It looks like open source wins against corporate.  Sure, the results are not exactly scientific but interesting all the same.

Mixing Secure and Non-Secure Assets in Your Web Application

The life of the web developer never seems to be easy, always a new problem cropping up.

The Problem

One such issue surfaced when a client wanted to begin accepting credit cards.  As most Internet users assume, they land on a web page asking for credit card information and it’s secure, I wouldn’t enter my credit card information without seeing the friendly little lock.  Another telltale sign is seeing the https: in the browser address bar.

Many sites today make use of outside CSS and JavaScript files host on a content delivery network (CDN) somewhere on the interwebs.  This has tremendous benefit for web developers and users alike, giving applications better performance.  The problem arises when we have a secure page (https) which pulls in assets from non-secure CDNs, where requesting assets securely will fail to return successfully and ruins the user’s experience.

This application happens to be a Ruby on Rails application but that fact is irrelevant.  The scenario is likely common today; we have a secure checkout page but our site contains menus and links to pages which are not sure but just plain http.  When the users visit the site with their browser of choice they are presented with various messages or maybe none.

  1. Firefox 6/7, no message..just no indication the page is using SSL.
  2. Chrome, no message but a red line through the “https:” in the browser address bar.  This does not give confidence to the user, I would not put my credit card information in this page.
  3. Safari, no messages and everything looks good with the exception of the missing tiny lock icon indicating a secure page.
  4. Internet Explorer, well this is the least friendly of the browsers telling the user there is mixed content and prompted with how to proceed.

The problem was mainly centered-around the Yahoo YUI JavaScript and CSS assets and how they were included.  This application uses the Yahoo content delivery network (CDN) to serve the assets, which is a great way to serve the assets.

The Solution

I decided to do what every self-respected web developer does when facing a problem, Google for someone else who had the same problem and successfully solved it.  I ran into one very insightful post from Dan over at CollectiveIdea.  The post lays out a very similar problem with some good ideas for the solution.

Dan points out a URL like this:

<script src="http://ajax.googleapis.com/ajax/libs/jquery/1.4.4/jquery.min.js" type="text/javascript"></script>

Will cause Mixed Content warnings when included from a secure page.  Some of the suggested solutions include downloading all of the assets locally and the problem goes away.  Although this is true, we lose the benefits of using a CDN.

What works is both elegant and simple; two qualities that make my day.  Referencing your CDN-based assets this way is only a slight change:

<script src="//ajax.googleapis.com/ajax/libs/jquery/1.4.4/jquery.min.js" type="text/javascript"></script>

Notice how we are referencing the GoogleAPI URL, we leave off the http: and the https:.  By using two forward slashes only the request will resolve itself and work brilliantly.

Maybe I am the last web developer to find out this tidbit of information but I wanted to document it so the next one faced with this could find the answer here.

The Simplicity that is Pow

Logo pow

Simplicity is a beautiful thing.  I love a simple tool which makes my life as a developer easier.

Ruby on Rails has made the life of the web developer much more pleasurable over the years but even so there are some things which could be made better.  One such thing is running your Rails applications locally when developing.  Every Rails developer is familiar with the script/server command if you are in Rails 2.3 and earlier or the rails server command if you are using Rails 3.  The next step would be to fire up the browser and enter localhost:3000 into the address bar.  Most of the time this works fine, but a bit tedious.  The real problems appear when your application supports subdomains like myaccount.myapp.com, which has been hard to do up until this point.

Enter Pow

Pow is a Rack server developed by the folks at 37Signals to help alleviate the pain of serving local Rails applications.  Using Pow allows the developer to go from accessing their local Rails application using localhost:3000 to something like myrailsapp.dev.  For example, my expense tracking software, Expens’d uses subdomains quite a bit, so we get URLs like rbazinet.expensd.com and when using Pow I can simply use the URL rbazinet.expensd.dev.  This is perfect and simplifies the process.

Setting Up Pow for Serving Rails Applications

Installing Pow is pretty easy and shown on the Pow web site but for those not interested in heading over there you can open up a terminal session and enter the following:

$ curl get.pow.cx | sh

Each application has to have a symlink defined.  The Pow web site says to:

$ cd ~/.pow$ ln -s /path/to/my/app

When I setup Expens’d to use Pow setting up the symlink this way didn’t work for me.  I had to add the application name after the path, like this:

$ ln -s ~/rails_apps/expensd expensd

Since Expens’d is currently a Rails 2.3 application, a config.ru file is needed and placed in the application root folder.  The file should contain the following:

# RAILS_ROOT/config.ru
require "config/environment"
use Rails::Rack::LogTailer
use Rails::Rack::Static
run ActionController::Dispatcher.new

If Expens’d was a Rails 3.x application, I would not have had to create the config.ru file.

Your mileage may vary.  Once the symlink is done heading to the browser you can just enter the domain for the application with the .dev extension, like expensd.dev.  This works just perfectly.

Restarting Your Application

One of the first things I wondered about when using Pow was a need to restart the “server” when I make changes to a routes.rb file.  It turns out we treat this the same way we restart Passenger.

$ [APP_ROOT]/touch tmp/restart.txt

Log File Monitoring

Using Pow provides us with the typical development.log file in the [APP_ROOT]/log directory.  Keeping an eye on the log can be done from a terminal window.

$ [APP_ROOT]/tail -f log/development.log

This provides  a nice way to see what’s going on.  There is also a raw log file produced from Pow that gives some additional details.

$ tail -f ~/Library/Logs/Pow/access.log

Potential Issue

When I setup Pow on my Mac Pro it all worked perfectly from the get-go but on my MacBook Pro I ran into a problem when I tried to browse a URL served by Pow such as expensd.dev.  DNS seemed to think I wanted to go to the Internet to find the site and I received a 404 error when I tried.    The problem was known and is outlined on Rob Conery’s site.  The first part of the solution involved running the scutil to see if .dev resolver was being used:

$ scutil --dns

You should see a bunch of entries and one should look something like this:

resolver #8
  domain : dev
  nameserver[0] : 127.0.0.1
  port    : 20560

I don’t know if Pow uses the same port all the time, so that may change.  The key here is the domain, indicating dev.  If this resolver is missing the solution is pretty simple, open the file /etc/resolver/dev and simply save it.  Worked like a charm for me.  Run the scutil –dns command from above and see if the resolver is now listed.

A Better Pow?

I think Pow is pretty awe some just as it is but it seems someone has stepped up to make it even better with a gem named Powder.  I haven’t had the chance to play around with this tool yet but a blog post by one of it’s creators says they wanted to make Pow “ridiculously easy”.  The commands supported help make Pow that much easier:

$ powder applog
=> tail the log of the current app
$ powder config
=> Get Pow's current configuration information
$ powder list
=> List all the current apps linked in ~/.pow
# aliased as powder -l
$ powder log
=> Tails the pow log.
# Not the application log, but the pow log, available at
# ~/Library/Logs/Pow/apps/#{app-directory}.log
$ powder open
=> Opens the pow link in a browser
# aliased as powder -o
$ powder open bacon
=> Opens http://bacon.dev in a browser
# if you have set up alternative top level domains in .powconfig,
# then the first listed domain will be opened.
$ powder restart
=> Restart the current app
# aliased as powder -r
$ powder status
=> Get Pow's current status information
$ powder version
=> Returns the current powder version
# aliased as powder -v

More Than Rails Applications

Since Pow is serves up Rack apps the possibilities are pretty endless.  I found one bit to share where someone is using Pow to serve his PHP apps, pretty clever.  I would imagine this technique could be used in many applications like this.

Finally

The only thing I can say is I love Pow.  It has made my life so much easier.  Thank you 37Signals.