Codementor Events

Conditional execution with DSL in Ruby

Published Jun 12, 2018Last updated Dec 09, 2018
Conditional execution with DSL in Ruby

I’m a fan of minimalistic and clear code. Recently, I’ve faced some repetitive code. I believe, that everyone should keep the code as DRY (don’t repeat yourself) as possible. The code was like this: call to a third party API, check the response and do something if it’s successful, do something else if it’s not. A basic code, but the implementation is boring.

Problem

To not be verbose and not repeat that I’ve already said, I'll just provide an example of the code:

response = StripeCall.new(number: 'valid').call
if response.success? puts response.body
else puts response.body
end

Basically, there is no any issue with this code. But, I find it’s not good enough, because there are too many details to be aware of:

  • the two classes where their public interface is has to be known (it’s StripeCall and the class of response object);
  • don’t forget to check the response everywhere where it’s used and use an if clause for this;
  • don’t forget to instantiate the object of StripeCall class properly (pass params into new, but not into call).

There are may be other objections, but unfortunately, I can’t identify them for now. All in all, we are humans and everyone has their own feelings.

Solution

From my practice, the bad feelings could be eliminated by introducting some sort of DSL. Start with imagination, but don’t go too far away from Ruby syntax (otherwise you will need a new language to implement, but I don’t want this today, as I’m good with Ruby).

First, the problem with keeping in mind the details, whether pass params into new or call can be gotten rid of by defining the call method on the class level. Then, knowing that .call(params) can be replaced with .(params), the number of typed symbols is reduced.

After this, knowledge from other languages comes into the action: in JavaScript, there is a pretty syntax for processing similar cases like this —.onSuccess(func1).onError(func2). I personally find it useful and handy. So, the final solution could be look something like this:

StripeCall.(number: 'valid') .on_success { |response| puts response.body } .on_error { |response| puts response.body }

Let’s implement it:

# A base class for all classes implement calls to API.
class ApiCall attr_reader :params def self.call(params) new(params).call end def initialize(params) @params = params end def call @res = execute self end def on_success yield @res if @res.success self end def on_error yield @res unless @res.success self end private def execute fail NotImplementedError end
end
# A concrete class implements call to API.
class StripeCall < ApiCall Response = Struct.new(:success, :body) private def execute success = params[:number] == 'valid' body = success ? 'ok response' : 'bad response' Response.new(success, body) end
end

Now, the code is ready to be played with:

StripeCall.(number: 'valid') .on_success { |response| puts response.body } .on_error { |response| puts response.body }
# => ok response StripeCall.(number: 'invalid') .on_success { |response| puts response.body } .on_error { |response| puts response.body }
# => bad response

Actually, the definition of blocks everywhere can be annoying. Therefore, it's simplified as well:

def handle_success(response) puts response.body
end def handle_error(response) puts response.body
end StripeCall.(number: 'valid') .on_success(&method(:handle_success)) .on_error(&method(:handle_error))

Now only one class intercase needs to be memorized — it’s StripeCall. The lines number is reduced from 6 to 3 (the implementation of conditional branches is not taken into account). But the main strength of such a DSL is that the implementation is hidden and there could be raised and caught exceptions along the way. By catching them and processing in the base class, we reduce even more repetitive code.

For example, the call method of the base class could be implemented like this:

class ApiCall ... def call @res = begin execute rescue StripeError => e OpenStruct.new(success: false) end self end ...
end

Conclution

A big project usually has a lot of code (surprise!). Every new line of code increases coupling and introduces complications. It gets harder to maintain and test it, especially when the code doesn’t follow DRY paradigm.

In other words, it’s repetitive. Keep your code clean and don’t hesitate to introduce your DSL to solve YOUR issues. And this way the code will be readable and close to the business domain, what is dreamed of by every developer. Happy coding!

Discover and read more posts from Andrey Koleshko
get started
post commentsBe the first to share your opinion
Dylan Cazaly
3 months ago

You are doing good work keep it up

Show more replies