Proper error handling

[elifoster]

Right now, if there's an error, we simply return the error code as a String. This makes it difficult or weird to check for errors. For example, we want to be able to do this, as is idiomatic in Ruby:

begin
  hash = butt.some_function
  return hash['important_value']
rescue SomeError
  p 'Oh no we errored'
end

but right now, we have to do this:

hash = butt.some_function
if hash.is_a?(String)
  p 'Oh no we errored'
else
  return hash['important_value']
end

or this, for methods that return a String either way

string = butt.other_function
if /Error:/ =~ string
  p 'Oh no we errored'
else
  return string.reverse!
end

This is awful.

[elifoster]

(See #42).

We have a couple options I've thought of:

Categorized errors with a description given in the constructor:

# We have determined that the edit has errored:
fail EditError.new(error_message)

Specifically defined errors

# We have determined that the edit has errored
case error_message
when 'badtoken'
  fail BadToken.new
# ...
end

The benefits to the first:

• No chance for unidentified errors, for example, when an extension adds its own errors. There's no way we could add a specific error class for every possible error in every extension.

The benefits to the second:

• Performing different things based on different kinds of errors is much easier. Instead of rescuing a single error and checking the error message, the various errors could be rescued individually, thus making the code more maintainable.

@APerson241 @xbony2

[enterprisey]

(For the record, I strongly support the first one, for the reason you mentioned.)

[enterprisey]

@elifoster and @xbony2, what are your opinions on option 1 vs option 2?

[xbony2]

Whatever work.

[elifoster]

I like option 1

[enterprisey]

I've implemented option 1 (for just edit()) in #42.

[elifoster]

I have finally finished this up.