Lots of file exchange services exist, for example SendBigFiles.com etc. However, all of these services are hosted on someone else's hardware, and most of them share files by transferring URLs -- usually via email -- without good access control or authorization schemes.

We built Convection because a client needed to transfer files with other companies, but they needed to host the system themselves because the contracts they hold with their own clients require them not to store data on services that they don't control. The specifications Convection was built around were:

1. Hosted on our own server.
2. Downloads require a login, and files cannot be shared by email.
3. Users must log in to download files or see available files.
4. User accounts can be grouped, groups can be managed.
5. Files can be shared with an entire group.
6. Files uploaded by users default to minimal permission - visible only to the uploader and to admins.
7. All communications over SSL. (we made this optional)

Installing and hosting Convection

To run Convection, you will need a webserver capable of running a Ruby on Rails application, and a database. Setting such a thing up is beyond the scope of this post. If you have a Dreamhost account, you can set up a Rails-capable domain with a couple of clicks in their web panel. In addition to the server, you will need to set up a database (we have only tested MySQL, but Convection should work with any SQL database for which Rails/ActiveRecord has a supported adapter, including PostgreSQL and Oracle), and initialize the database with these two commands:

  > rake db:migrate RAILS_ENV=production
  > rake db:seed RAILS_ENV=production

This will generate the tables necessary for Convection to run, and create a pair of initial demo users "admin" and "user", both with password "foobar".

If you are setting up a server yourself, there are plenty of guides to deploying Rails on the web. Much of our own guide to deploying CruiseControl.rb can be used to set up any Rails application on Slicehost or any other Ubuntu Linux hosting provider.

Let me know if you're trying to deploy Convection and having trouble: if we know people are using it we may put effort into making it easier to deploy and install, and write a more thorough guide.

A few other links that may help you with deploying a Rails application, depending on your environment:

1. Using Phusion Passenger to Deploy a Rails Application on Apache
2. Deploying Rails Applications (book)

If you Google around you may find plenty of other links relevant to your particular environment.

Configuring Convection

If you log into your running Convection application as an administrator (initial user "admin", password "foobar"), an Admin Tools utility will appear in the right hand column. From here, you can access tools for creating users, and groups, and the general site configuration.

In general site config, you can set your site name and logo, set whether or not the site requires SSL access (Note: your server must already support SSL!) outgoing email and email notification preferences, add Google analytics, and an assortment of other site configuration operations that are mostly self-explanatory.

Upload progress bar: experimental feature.

If your site hosts large uploads that take a while to transfer, you can try our experimental tools to provide an upload progress bar to the user. This tool will only work if your site is served by Apache, and requires installing and configuring an optional module for Apache.

To enable this tool, follow the instructions in the README file and associated links, and turn on the progress bar setting in site preferences.

Helping us improve Convection

Convection is currently in version 1.1.4 and has been in production in two places (that we know of) for about five months as of June 8, 2010.

Please let us know if you are using Convection and enjoy it (or don't). Feel free to request features or alterations, but Convection is open source, so also please consider contributing if you have ideas!

We're using the use_db plugin, and all it takes to make testing transactions happen around multiple DBs is:

In: spec/spec_helper.rb

require 'override_test_callbacks'

My concern comes from the fact that this is a direct and unfiltered monkeypatch on ActiveRecord::TestFixtures. So it relies on use_transactional_fixtures (which could certainly be used without using actual fixures, granted), and if the test transaction code moves within Rails, that's another integration to worry about. Or if we add a spec that doesn't wind up making ActiveRecord::TestFixtures load... Or if we decide to use something other than use_db...

So instead I'm using:

Spec::Runner.configure do |config|
  config.prepend_before do
    (UseDbPlugin.all_use_dbs - [ActiveRecord::Base]).each do |db|
      db.connection.increment_open_transactions
      db.connection.transaction_joinable = false
      db.connection.begin_db_transaction
    end
  end
 
  config.append_after do
    (UseDbPlugin.all_use_dbs - [ActiveRecord::Base]).reverse.each do |db|
      db.connection.rollback_db_transaction
      db.connection.decrement_open_transactions
    end
  end
end

If we weren't already using transactional fixtures, I might pull out the - [ActiveRecord::Base]. And if we were to change off of use_db, there's one place to change the transaction code. Finally, there's much less dependence on the innards of ActiveRecord - only it's published API.

MyModel.new(:field_1 => 'foo', :field_2 => 'bar')
MyModel.new('field_1' => 'foo', 'field_2' => 'bar')

It's convenient, often, to not have to worry about whether your keys are symbols are strings since they get converted around a bit when you pass parameters. The downside of this, however, is that it will happily accept BOTH without complaining, and will quietly default to the symbol key regardless of the order you specify them in:

>> model = MyModel.new(:field_1 => 'foo', 'field_1' => 'bar'); nil; 
>> mymodel.field_1
=> 'foo'
>> model = MyModel.new('field_1' => 'foo', :field_1 => 'bar'); nil;
>> mymodel.field_1
=> 'bar'

Okay, so that's kinda sloppy. Bad ActiveRecord! No Biscuit!

This can cause serious confusion for the unwary. When ActionController hands us a params hash, it always has String keys, like this:

>> eval params
=>  { 'article' => { 'title' => 'Awesome blog post', 'body' => 'I will make you smart' } }

But most of us, canonically, specify params and default AR values with symbols, like this:

   post :article => {:title => 'Awesome blog post', :body => 'I will make you smart'}

So we get used to thinking about them as symbols.

This means we can make mistakes like this one I made recently. Consider this block of code for a shopping cart model that pre-fills some fields for an associated Payment by pulling the address from the user's profile, to save the user re-typing their address:

class ShoppingCart < ActiveRecord::Base
  has_one :payment
 
  def build_default_payment(options = {}) 
    #prepopulate the billing address from the profile and merge
    #with params passed into options
    build_payment(prepopulated_fields.merge!(options)    
  end
 
  def prepopulated_fields
    if (addr = self.person.address)
      {
        :billing_address_1 => addr.line_1,
        :billing_address_2 => addr.line_2,
        :city => addr.city,
        :state => addr.state,
        :zip => addr.zipcode
      }
    else
      {}
    end
  end
end

Looks great, right? And if the user's address has a nil field (like no city, or no line_1), it will get overwritten by the hash merge.

Except not. I specified symbol keys in prepopulated_fields, but the hash getting passed to build_default_payment's 'options' argument has string keys, because it's coming from params. So the merge doesn't overwrite the value for :line_1, it simply adds a new key 'line_1'. So, if a user has a profile address but hadn't entered a line_1 (just city and state), and then manually entered line_1 in the payment form to submit, the Payment build during the create action was getting this hash:

build_payment({
   :line_1 => nil,
   :city => 'Pasadena',
   :state =>'CA',
   :zipcode => '91106'
   'line_1' => '100 Main St.'.
})

ActiveRecord was respecting the :line_1 => nil from the profile, and not the 'line_1' => '100 Main St.' from params. This meant that the user couldn't make payment! The payment had validates_inclusion_of line_1, and even though it was typed into the form it was getting ignored because of the nil from his profile address. Very frustrating for a user to manually type in a billing address and get back "Address Line 1 can't be blank." on every submit!

Nasty ... this one took a while to figure out. Beware of this little foible of ActiveRecord!

Logical Reality did the logo and layout design work for Rails Tutorial.

Specifically, I needed to put annotations in the description of a work order object that said, for example "this work order was escalated from Problem Report 293. This was done in a create action that redirected at the end and never rendered a view, so I really did need to generate that link in the controller. And for consistency with the rest of the application, I wanted to generate the link with link_to(@task).

Now, ActionView::Helpers::UrlHelper is not loaded in a Rails controller, even if you've put helper :all in application.rb (application_controller.rb in newer versions). So, when I tried to use link_to in the controller, I got an error:

NoMethodError: undefined method `link_to' for #
/Users/evan/Development/Ruby/eclipticdb/app/helpers/tasks_helper.rb:64:in `task_link'
/Users/evan/Development/Ruby/eclipticdb/app/controllers/tasks_controller.rb:103:in `escalate'
... etc ...


The first fix - but with a problem

A year ago, I fixed this just by adding include ActionView::Helpers::UrlHelper at the top of that controller. This worked great ... for a while.

Lately, I've been rewriting this application into a RESTful style - it had previously been a controller/action style application. In the process, I started linking things with resource paths and polymorphic paths ... a lot of link_to @task and edit_polymorphic_path(@task) sorts of bits. And these started breaking. I began seeing this mysterious error:

Error:

You have a nil object when you didn't expect it!
The error occurred while evaluating nil.url_for

vendor/rails/actionpack/lib/action_view/helpers/url_helper.rb:71:in `send'
vendor/rails/actionpack/lib/action_view/helpers/url_helper.rb:71:in `url_for'

This one was a real bitch to debug, I have to say. The line in question that was failing in url_helper.rb said this: url = @controller.send(:url_for, options). Clearly, @controller was nil ... which was very bizarre, because I never interact with that instance variable anywhere.

I thrashed around trying to find the cause of this error for quite some time. Eventually I realized that the link_to method was only failing when called from a view in TasksController, and not from any other controller. And then I realized that TasksController was the one where, a year ago, I'd put include ActionView::Helpers::UrlHelper at the top. Somehow, including that helper in the controller was nullifying @controller when those helper method we called from within the view. I removed the include and my polymorphic and resource links all started working again.

Now back to the original problem!

Of course, that then left me back with the problem I'd had a year ago ... needing to use link_to from within the controller and having no way to do it. After a fair bit of googling around I found this post from Neeraj, which had an interesting approach -- but a commenter had suggested a much easier solution:

[sourcecode language='ror']self.class.helpers.link_to[/sourcecode]

I'm not certain where one would find this in the docs, but it does seem to have solved my problem for now. Onward and upward!

So, I ran into this little challenge: all three STI subclasses use the same controller, "tasks", because they all have essentially the same behavior and differ only in minor details. But, when I do a resources map:

map.resources :tasks

Then I get errors in much of my code when I say things like redirect_to @task, because if that task happens to be an ActionItem, it's trying to call action_item_path(@task), which doesn't exist.

I googled around a bit to no result. Striking out on my own, it turns out the answer is as simple as mapping each resource independently, and just overriding the controller in map.resources:

In config/routes.rb

map.resources :tasks
map.resources :action_items, :controller => 'tasks'
map.resources :work_orders, :controller => 'tasks'
map.resources :problem_reports, :controller => 'tasks'


Now, redirect_to @task works just fine regardless of which subclass @task happens to be.

I'm following Chapweske's approach of blocking mass assignment by default in all models, by putting this line in an initializer:

ActiveRecord::Base.send(:attr_accessible, nil)

This had the expected side effect of breaking several zillion tests, because tests frequently use things like Model.build() and Model.create!() to generate on-demand fixtures during testing. Hartl has a great bit of code that creates unsafe_build() and unsafe_create() methods in ActiveRecord. You can use these methods instead of build() and create() to function as expected in your tests.

This works great, except that I also use the mass-assignment method update_attributes! in my tests and specs frequently, particularly when I want to spec the effect a change on one model has on an associated models' methods. So, I expanded on Hartl's helper code a bit, to give myself the necessary methods. In case it helps anyone else:

/lib/initializers/unsafe_build_and_create.rb

class ActiveRecord::Base

def unsafe_attributes=(attrs)
attrs.each do |k, v|
send("#{k}=", v)
end
end
end

NoMethodError in 'LoansController POST 'create' with valid parameters should succeed'
undefined method `env' for

This happens on the line where I call post :create in a controller spec.   Undefined method 'env'?   What's that about? I'm certainly not trying to call a method named "env".

It took me a little bit to figure out what was going on.  See, this series of tests needed access to a particular LoanRequest object I was pulling out of fixtures.   So I'd put above the tests:

before(:each) do
... some other stuff ...
@request = loan_requests(:johns_loan_request) # fetch fixture
end

Well, kids, it just so turns out that it's a bad idea to overwrite the @request instance variable in any rails context. Who knew?

Come to think of it, it would be nice to change the accessibility and/or mutability of Rails' basic instance variables and classes to prevent this kind of accidental overwrite by the programmer. Because when you make that mistake, it's invariably a bit of a pain to figure out because the error it causes is obscure.

Maybe I'll have to dig into the code one of these days to see if anything can be done about it.

I've had dozens of problems, in fact it's fair to say I've spent upwards of 15 hours just debugging ultrasphinx and getting my tests to pass.   There were several problems; here are the main three and how I fixed each one.

This should be useful to anyone upgrading Insoshi to the sphinx version, or to anyone else trying to get ultrasphinx working in their Rails project. I definitely don't recommend starting with this post if you're just starting out with sphinx. Instead, go read this much better introductory tutorial from the guys over at Insoshi. Then if you have problems, come back here and you may find solutions.

Getting search tests (or specs) to pass with sphinx

This one is pretty simple, in retrospect, but it can be frustrating and opaque if you are used to ferret.  Unlike ferret, sphinx (at least via ultrasphinx) runs only via a daemon.   Where acts_as_ferret uses a daemon only for the production environment and just accesses the index files directly in test or development, ultrasphinx can only get to the indexes through the daemon.

So, to run your tests, you just build up the indexes for test and run them.  In this case, I'm running the specs for Insoshi's searches controller:

From the command line in $RAILS_ROOT:

rake db:test:prepare
rake ultrasphinx:configure RAILS_ENV=test
rake ultrasphinx:index RAILS_ENV=test
rake ultrasphinx:daemon:start RAILS_ENV=test
script/spec spec/controllers/searches_controller_spec.rb

The problem, of course, is that it doesn't work!   The reason is that db:test:prepare creates the structure of your database, but doesn't load any of your fixtures as data: the test db is empty..  So when you run the index command, an empty index is built.   You can see this from the output of that first index command, which will look something like this:

collected 0 docs, 0.0 MB
total 0 docs, 0 bytes
total 0.078 sec, 0.00 bytes/sec, 0.00 docs/sec

Ultrasphinx has built an empty index.

The solution

The solution, believe it or not, is to run the tests, let them fail, re-index, and run the tests again (Many thanks to Long Nguyen at Insoshi for helping me figure this one out):

rake db:test:prepare
rake ultrasphinx:configure RAILS_ENV=test
rake ultrasphinx:index RAILS_ENV=test
rake ultrasphinx:daemon:start RAILS_ENV=test
script/spec spec/controllers/searches_controller_spec.rb #FAIL!!
rake ultrasphinx:index RAILS_ENV=test
script/spec spec/controllers/searches_controller_spec.rb #PASS!!


The first attempt to run the specs loads the fixtures, and leaves them in the database, thus letting the subsequent index command build an actual index.

Running sphinx for both test and development environments at the same time

The next big challenge was enabling behavior-driven development. I like to work with autotest and growl running constantly in the background. But this was tough to do with sphinx, because the daemon needed to be stopped and re-started, and the index re-created for each environment, alternately running all of the above commands either with or without RAILS_ENV=test.

The solution is to set up your ultrasphinx base configuration to completely separate both the test and development indexes and to let the daemons for the two environments listen on different ports. I had tried something like this and come close, but not quite, when Long at Insoshi again bailed me out. You need to change the port (in two places), and the paths of the logs, pidfile, and index directories so that test and development daemons are using entirely separate resources. Here's a diff of my test.conf and default.conf:

33c33
< port = 3312
---
> port = 3322
35c35
< log = log/searchd.log
---
> log = log/searchd_test.log
39c39
< pid_file = log/searchd.pid
---
> pid_file = log/searchd_test.pid
50c50
< server_port = 3312
---
> server_port = 3322
57c57
< sql_range_step = 5000
---
> sql_range_step = 999999999
64c64
< path = sphinx
---
> path = sphinx_test

The sql_range_step is related to the next issue, which is that sphinx does not play well with foxy fixtures. Anyway, make the above changes and you should be able to run test and development sphinx daemons at the same time:

rake db:test:prepare
rake ultrasphinx:configure
rake ultrasphinx:configure RAILS_ENV=test
rake ultrasphinx:index
rake ultrasphinx:index RAILS_ENV=test
rake ultrasphinx:daemon:start
rake ultrasphinx:daemon:start RAILS_ENV=test


If it worked, you should see separate indexes in $RAILS_ROOT/sphinx and $RAILS_ROOT/sphinx_test, and two daemons running, which you can confirm with ps waux | grep searchd:
evan 1339 0.0 0.0 78100 292 s000 S 5:37PM 0:00.52 searchd --config /config/ultrasphinx/test.conf
evan 1326 0.0 0.0 78100 292 s000 S 5:36PM 0:00.68 searchd --config /config/ultrasphinx/development.conf

Getting sphinx to play well with foxy fixtures

The next problem I discovered was that on some machines, but not others, running my search specs would result in these weird errors:
1)
ActiveRecord::RecordNotFound in 'SearchesController Person searches should search by name'
Couldn't find Person with ID=328556765
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search/internals.rb:308:in `reify_results'
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search/internals.rb:286:in `each'
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search/internals.rb:286:in `reify_results'
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search.rb:362:in `run'
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search/internals.rb:352:in `perform_action_with_retries'
/var/www/domains/unithrive/vendor/plugins/ultrasphinx/lib/ultrasphinx/search.rb:342:in `run'
/var/www/domains/unithrive/app/controllers/searches_controller.rb:38:in `index'
./spec/controllers/searches_controller_spec.rb:51:
script/spec:4:

When I poked into this "Couldn't find Person with ID=328556765" error, it seemed like sphinx was almost working. The index was set up, and the search was finding someone in the index during the test. Ultrasphinx was passing back the id 328556765, which didn't exist in the database. So why would Sphinx "find" a record in its index but then pass back an ID for a database record that didn't exist?

And furthermore, why would it work on one machine, but not on another?

The brainstorm came when I checked what the actual database IDs were for this particular record, with Person.find_by_name("fixtures' name").id. On machines where it worked, the id was a huge number (is it generally is with foxy fixtures), but on machines where it didn't work, the id was an even huger number.

Sphinx tries to make sure that all items that get indexed have a different index in sphinx, and it does this by multiplying all of your id's by N, where N is the number of models getting indexed, and adding an offset of 0 for the first model, 1 for the second, etc. This guarantees that every record from every table will have a unique id. In the case of this application, all of my Person records were getting indexed by sphinx as (Person#id * 4 + 2).

Danger, Will Robinson: 32-bit int rollover!

The problem is that foxy fixtures generate their own ids from a hash of the fixture label, and those ids can be anywhere in the 32-bit unsigned integer space. But Sphinx also stores ids as 32-bit unsigned integers. This means if you happen to get a large fixture id, and then sphinx multiplies it by 4 (or whatever; it could be higher if you have more indexed models), your id will rollover and come out as (id * N + n) % (2^32). Sphinx will store that result, and then when it finds the record in a search, it will try to recreate the original id by subtracting n and dividing by N ... giving you the wrong id. Your test will fail to find the record.

Incidentally, this problem with foxy fixtures is why your test.base file needs the line sql_range_step = 999999999. Sphinx builds indexes by searching a few ids at a time. But the ids generated by foxy fixtures are so big that if sphinx only collects them in ranges of 5000 at a time, it will take forever to find them all.

After some googling, I found that these issues are discussed in a thread over at RubyForge.

The solution

I'm working on a plugin that monkeypatches foxy fixtures to create sequential, low-numbered IDs. In the meantime, you can just compile sphinx to support 64-bit ids, which should give you plenty of headroom to handle foxy fixture ids multiplied by N in sphinx^*:

In your sphinx source directory:

configure --enable-id64
make
sudo make install

That should do it. Let me know in comments if any of this information helped you.

^*At least until you start approaching 2^32 models in your application, that is.

create_table "log_entries", :force => true do |t|
t.integer "user_id", :limit => 11
t.datetime "created_at"
t.datetime "updated_at"
t.string "item_type"
t.integer "source_id", :limit => 11
t.string "table"
t.string "action"
t.text "changes"
end

Whenever a record is created, updated, or deleted, I create and save an instance of LogEntry, containing for example {:table => 'task', :action => 'update'} and in the 'changes' column I save a serialized hash showing which attributes of the task object changed before and after save. In addition, I save which logged-in user made these changes, and when.

This is convenient, it gives my client a log that's much more user-accessible and allows them to easily back-trace who did what to the database and when, which is important for their process and certain certifications.

When I upgraded the project to Rails 2.1 a couple of weeks ago, most of the code still worked fine, but the hashes showing the object changes were no longer showing up in views of the log. The culprit turned out to be Rails 2.1's new dirty feature. Why? Because it adds the method 'changes' to ActiveRecord::Base. This is a great new feature that lets you know what has changed to an ActiveRecord object since you loaded it from the database, with all kinds of benefits like doing more selective updates, or not writing to the database at all if no changes have occurred, thus reducing system load if you call save! a lot.

Unfortunately, the new method 'changes' was obscuring my attribute 'changes' in some circumstances, and at the very least confusing the heck out of me, the programmer.

The solution, of course, is not to name your fields anything that corresponds to any Ruby core method or any method of ActiveRecord::Base. I fixed my problem with a migration to rename the column 'details':

def self.up
rename_column :log_entries, :changes, :details
end

def self.down
rename_column :log_entries, :details, :changes
end