Musings about Entrepreneurship, Technology and Software Development

Accidental Technologist

When You Can’t Check a Checkbox using Capybara

In my last post I documented an issue I was having using Capybara on a client’s Ruby on Rails application with a ReactJS frontend.  I wanted to share another issue I came  across during my feature testing escapades.

The Problem

During my recent feature testing project I had a form which has a checkbox on it. The checkbox had a label with it. Did I mention this is a ReactJS frontend? I’m not sure if this is specific to ReactJS, but I suspect it isn’t. I think other frontend JavaScript frameworks may exhibit the same problem.

The Ruby code for my feature test is dead simple:

  check “English”

That’s it. The test should run and when it finds the checkbox with a label of English, the checkbox should be checked. But, it doesn’t work. After many attempts at making this work and more Google searches than I can remember..I ended up at the Capybara mailing list.

The Solution

Thomas Walpole was kind enough to reply with his thoughts on the matter:

99.9% sure your checkbox isn’t actually visible on the screen.  

What you’re describing as the “checkbox” is probably an image (probably added via CSS pseudo elements) being shown in place of the actual checkbox input element to ensure the same styling across different browsers.  If the checkbox has a label element correctly attached to it you can use `check(‘whatever’, allow_label_click: true)` – https://www.rubydoc.info/ github/jnicklas/capybara/ Capybara/Node/Actions#check- instance_method –  to tell Capybara to click the label element instead of the hidden checkbox input to toggle state.  If you don’t have an associated label then you’ll need to find whatever element is actually visible on the page and click it instead.

Changing my test to include this for the checkbox, worked perfectly. 

  check(“English", allow_label_click: true)

I hope someone finds this valuable and will save them some time and hair pulling.

Fixing StaleElementReferenceError When Using Capybara

I’ve been spending an extended period of time lately writing Feature specs for a Ruby on Rails project using Capybara. Overall it’s been a fun project to work on but I had run into a few issues along the way. One in particular is the StaleElementReferenceError exception that get thrown when you are trying to find an element on your page that you know is there but Selenium can no longer find.

The error appears to be a timing issue between the page loaded in the browser, it being rendered and Capybara and Selenium trying to find the element. After much trial and error, along with many Google searches, I found no solutions that were reasonable. Even though many people had the same problem.

The solution was pretty simple actually. I added a rescue to my test and simply slept for 1 second and retry the failure. I admit, this seems like a hack but it works and that’s all I really cared about at the time. I see no ill effects from this, no measurable delay.

RSpec.feature "Listings", type: :feature, js: true do
   scenario "can edit common area - step 3" do
      begin
        post_a_space_step_1_with
        click_on 'Next'
        post_a_space_step_2_with
        click_on 'Next'
        click_on 'Edit Common Area'
        expect(page).to have_content "Step 1"
      rescue Selenium::WebDriver::Error::StaleElementReferenceError
        sleep 1
        retry
      end
    end
end

I’m certainly open to a better solution or an explanation as to why this might be bad. Please leave comments. Hopefully I help someone get around this really annoying problem.

Ruby on Rails Testing Resources

When taking the plunge into Ruby on Rails it’s really easy to get carried away with learning all about the framework. It’s easy to learn the fundamentals and later realize the Rails community is a community of testers. It’s a strange world when you set out to learn about testing, TDD (test-driven development), BDD (behavior-driven development) and other acronyms and phrases relating to testing Ruby on Rails applications.

I decided to put together a short list of helpful resources to get started. If you have suggestions that would be useful to be added to this list, please add a comment or email me directly and I’ll update this post.

Books

  • Everyday Rails Testing with RSpec – this is a great, hands-on, roll-up your sleeves and get-to-work book. If you want to use RSpec on a daily basis, this book gives great advice on how to use RSpec in your day-to-day routine. It’s kept up-to-date with latest RSpec too.
  • Rails 5 Test Prescriptions – I use this book as a reference I often go to. It’s been updated to from previous versions to now Rails 5 and is a great tool to have on the shelf.
  • Effective Testing with RSpec 3 – if you decide you’d rather start without worrying about all the details around Rails you can start with learning RSpec with plain Ruby and help yourself. I’ve been through this one cover-to-cover and it’s a great tutorial.
  • The Minitest Cookbook – if you decide RSpec isn’t for you, this is probably the ultimate resource for Minitest. Well-written and kept up-to-date.

Podcasts

You can’t really learn testing from a podcast but you can learn how others approach the craft. The first is a podcast dedicated to testing Ruby applications. The rest is a list of a few episodes of podcasts that discussed testing.

I’ve been listening to The Ruby Testing Podcast and picked up some nice tidbits so far.

Training

I love Pluralsight.

Xavier Shay has long been involved in the Ruby community and well-known for discussions around testing. One of his best blog posts explains his approach to testing.

I’ve taken several courses on Udemy and they are one of my favorite places for training. The prices are low and there are many courses, so you have to do a bit of work to see which course is right for you but well worth the effort.

Ruby on Rails – Supporting SSL for PostgreSQL on Heroku

For anyone who received this notice from Heroku:

Since 2016, all newly-provisioned Heroku Postgres databases have enforced the use of SSL to keep your data safe. However, one or more of your Postgres databases are running on legacy infrastructure, which does not enforce the use of SSL. In order to update your database to our security standards, and in response to potential impacts caused by Spectre and Meltdown, all databases – including those on legacy infrastructure – will be moved to our new Heroku PGX plans in a set of maintenances starting in March 2018 and concluding by April 2018.

What Do I Need to Do
In preparation for these maintenances, please check that your applications are using SSL to connect to your Postgres database and enable SSL connections if needed. Instructions on how to perform these steps are available in Dev Center.

If you’re using Rails 4.1+ there is a support article on the Heroku Dev Center, that helps clarify making updates to our database.yml file. This allows customizing some connection behavior to PostgreSQL.

Some parts of the database.yml file that cannot be changed include:

You cannot use the config/database.yml to set any values found in ENV[‘DATABASE_URL’]. This is a list of attributes you cannot change:

  • adapter
  • database
  • username
  • password
  • host
  • port

But, what can be changed include sslmode.

production:
 sslmode: require (disable|allow|prefer|require)
 pool: 15

I decided it would be helpful to reach out to Heroku to understand their guidance with regard to their notice. Their response:

If you’re using the pg gem, the default sslmode setting (and for libpq, the library that underpins it), is prefer this means that should the server have SSL support, it will be used when the connection is established. This means there should be no action required, though if you wish, it’s worth a test with spinning up a staging environment with a non-legacy Postgres instance.

It seems if you’re using Ruby on Rails with the pg gem, you should be OK doing nothing but with brownout period scheduled, it’s probably a good idea to test during one of those times.

Heroku Support also indicated setting the environment variable PGSSLMODE would also override the default behavior for sslmode used by libpq.

It seems this is a notice which doesn’t effect a majority of Heroku customers and is a necessary and worthwhile upgrade. Hopefully this helps others as the public information available for this from Heroku is minimal.

Fixing Your Puma-dev SSL Problems on Google Chrome

Ruby on Rails developers have it made in many ways. We rely on and take advantage of great software created by the community. Puma-dev is just one of those great pieces of software.

Puma-dev has some nice improvements over Pow, which Basecamp had promoted for years but has seemed to stop development.

Puma-dev allows developers to better mirror their local development environment to that of production. Prior to Puma-dev or Pow, developer would have to access their Ruby on Rails applications with something like localhost:3000 in their browsers. It works but having a real URL to visit, like mygreatapp.dev is better.

This worked well until a recent update to Google’s Chrome browser:

A lot of (web) developers use a local.dev TLD for their own development. Either by adding records to their/etc/hosts file or by using a system like Laravel Valet, which runs a dnsmasq service on your system to translate *.dev to 127.0.0.1.

In those cases, if you browse to http://site.dev, you’ll be redirect to https://site.dev, the HTTPS variant.

That means your local development machine needs to;

  • Be able to serve HTTPs
  • Have self-signed certificates in place to handle that
  • Have that self-signed certificate added to your local trust store (you can’t dismiss self-signed certificates with HSTS, they need to be ‘trusted’ by your computer)

I’ve faced this myself and Chrome refuses to serve the site, only showing security errors.

When trying to fix this problem I search a lot around the web and came up with very little. There were plenty of acknowledgements that since .dev is now a top-level domain (TLD) and Chrome 63 treats it as such and forces SSL, it looked like moving away from .dev would be needed.

This was my assumption until I discovered a recent post from Barry Woolgar of Storm demonstrating setting up Puma-dev. He addresses troubleshooting .dev TLD issues specifically:

If your browser complains about an untrusted root certificate, please do the following:

  1. Open Keychain Access
  2. Click the login keychain in the left pane, then find the Puma-dev CA certificate in the right pane
  3. Double-click it and expand the Trust section, and make sure it says Always Trust
  4. Drag it into the System keychain in the left pane
  5. Restart your computer
  6. Try https://my-project.dev again!

Although I didn’t need to restart my Mac, this worked perfectly! I can now serve .dev domains in Chrome and latest Firefox.

Tip: Save Typing ‘bundle exec’

If you’re a Ruby on Rails developer, you probably type the words bundle exec numerous times a day. I finally got tired of it and decided to do something about it.

In the context of keeping it simple, I use the ‘alias’ command and add to my .profile. I know there are more Rails recommended ways of solving this problem via binstubs but I rather not use that approach.

My solution is simple, add an alias to my .profile like this:

alias be='bundle exec '

Keen observers will notice the trailing space after the command. This space allows for alias chaining and can be helpful.

The resulting shortcut allows this:

bundle exec rspec

To this:

be rspec

The ‘alias’ command is super useful and has many applications to help remove repetitive typing tasks.

#

Life Beyond Rails: A Brief Look at Alternate Web Frameworks for Ruby →

I wrote up a list of alternative Ruby web frameworks some time ago and now Engine Yard has published a new list. Some of my original picks are in there, so it’s great to see they’ve continued.

A couple notables include Volt, for running Ruby on the server as-well-as the client, and Cramp that helps with needing a lot of open connections and bi-directional communication.

A great list to check out.

Introducing Rails Rescues

RR01

Today I?m happy to introduce Rails Rescues.

Rails Rescues?is?a service representing years of Ruby on Rails experience organized to help companies who have a Rails application but may be having some problems.

Rails Rescues aren?t limited to a fixed set of services but can include:

  • Scaling your website
  • Resolving site stability problems
  • Upgrading from an old version of Rails to current versions
  • Fixing Broken Code
  • Simplifying an out-of-control code base.
  • Finishing up after your contractor or employee left the project.
  • ?.we?ll fix anything holding you back from a successful Rails web site.

I?m happy to offer a $500 finders fee for any referral sent that results in a sign Rails Rescues contract. ?Please spread the word and visit the site. I will make myself available via live chat to answer any questions.

 

 

Speeding up Heroku Deploys

Anyone who deploys their Rails 3.x or 4.x utilizing the asset pipeline and doesn?t precompile those assets yet deploys to Heroku, knows it can take a really long time for your deployment.

I searched around a bit and found a great article on how to shave some time off my Heroku deployments. ?Alex MacCaw has a nice write up about the process:

If you?re using Heroku, the first step is enabling a Memcache addon. I?ve gone with the?memcachier?service, as they?ve got a generous free plan (which is all we need at this stage).

heroku addons:add memcachier:dev

Then we need to make sure the environmental variables are available to your app during the pre-compilation stage. Usually this isn?t the case on Heroku, but they?ve got a new labs feature called?user-env-compile?which will do the trick.

heroku labs:enable user-env-compile

Next you?ll need to add the?dalli?and?memcachier?gems to your Gemfile. Finally, the last step is to configure Sprockets.

Since I am using Rails:

With Rails

With Rails, just configure the assets cache store inconfig/environments/production.rb.

config.assets.cache_store =:dalli_store

And the time savings would be?.

?An example of time saving with a relatively small project:

Not using the speed-up method, deploy time:?2 minutes 40 seconds

Using the above method:?47 seconds

It is definitely worth the little effort.

Errors Installing the pg Gem When Using Heroku Postgres.app

I?ve been using the PostgreSQL Mac OS X app from Mattt Thompson and Heroku for quite some time now. ?If you don?t know what it is, it?s a drop in app bundle for the PostgreSQL database. ?There are many ways that work, this just happens to be really simple.

I use PostgreSQL with my Ruby on Rails projects and combine that with the pg ruby gem. ?

I ran into a situation where the pg gem would not install because it could not find pg_config in a known location on my Mac. ?The error occurred on Rails 3.2 but 4.0 may show the same behavior. ?

The Error

The error can come up when running a bundle install or just a straight gem install pg from the command line. The resulting error may look something like this:

Installing pg (0.17.0) with native extensions 
Gem::Installer::ExtensionBuildError: ERROR: Failed to build gem native extension
.
.
.
.
An error occurred while installing pg (0.17.0), and Bundler cannot continue.
Make sure that `gem install pg -v ?0.17.0'` succeeds before bundling.

The Solution

I already mentioned the problem is the gem install not finding pg_config during installation. ?So let?s find it.

1. First, find where pg_config is located. ?Run this command from a terminal window:
which pg_config

Should display something like this:

/Applications/Postgres.app/Contents/MacOS/bin/pg_config

2. You can tell RubyGems where your pg_config file is located:

gem install pg -- --with-pg-config='PATH_TO_YOUR_PG_CONFIG'

For example, pg_config is here on my system:

/Applications/Postgres.app/Contents/MacOS/bin/pg_config

So I would install the gem this way:

gem install pg -- --with-pg-config='/Applications/Postgres.app/Contents/MacOS/bin/pg_config'

The pg gem should now install. I hope this helps.

UPDATE: Scott Watermasysk points out another good solution: