WIP: Conditional exception validations

[rewinfrey]

Taking inspiration from #10 I wanted to see how it felt maintaining two sets of validation methods and invoke them at the ActionLogic::ActionCore level.

I also took inspiration from ActiveModel::Errors so if the non-bang validation method is used and errors are found, ActionLogic provides a context.errors.messages interface:

{
  :integer_attribute=>"Value expected to be of type Fixnum but is NilClass", 
  :float_attribute=>"Value expected to be of type Float but is NilClass"
}

I've updated the three types of validations (Attribute, Presence and Type) to support this difference, and am still working through updating the tests and fixtures.

My goal isn't to replace #10 but to compare approaches and see if some mix and match would be ideal, or if we like one direction or the other.

/cc @davesims

[codecov-io]

Codecov Report

Merging #11 into master will decrease coverage by 4.5%.
The diff coverage is 78.1%.

@@            Coverage Diff             @@
##           master      #11      +/-   ##
==========================================
- Coverage    99.1%   94.59%   -4.51%     
==========================================
  Files          22       23       +1     
  Lines         668      740      +72     
==========================================
+ Hits          662      700      +38     
- Misses          6       40      +34

Impacted Files	Coverage Δ
lib/action_logic/action_validation/errors.rb	100% <100%> (ø)
lib/action_logic/action_context.rb	100% <100%> (ø)	⬆️
lib/action_logic/action_core.rb	100% <100%> (ø)	⬆️
.../action_logic/action_validation/type_validation.rb	70% <14.28%> (-30%)	⬇️
...on_logic/action_validation/attribute_validation.rb	68.75% <16.66%> (-31.25%)	⬇️
...ion_logic/action_validation/presence_validation.rb	62.5% <42.85%> (-37.5%)	⬇️
spec/action_logic/active_use_case_spec.rb	92.9% <83.87%> (-7.1%)	⬇️
lib/action_logic/action_validation.rb	98.24% <96.87%> (-1.76%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d4599fc...1b3e54c. Read the comment docs.

[davesims]

👍 This was actually the approach I took on my first stab at it. This way feels less hacky to me and a lot cleaner. I am moving forward on unit tests on #10, I think they'll be pretty interchangeable with either approach.

[rewinfrey]

Closing out as stale. Happy to revisit this if this functionality would be useful.

cc/ @davesims