allow accessing dependant attributes which have a reserve keyword as name

[GustavoCaso]

Fixes #16

@solnic I was talking with @flash-gordon and we agree that this could be a good solution for this issue.

if you have a reserved keyword as one of your attributes end, do, when etc... you can append an underscore to the block argument and access inside the block.

Example

factories.define(:announcement) do |f|
  f.when { ROM::SQL::Postgres::Values::Range.new(3, 9, :'[]') }
  f.begin { |when_| when_.lower }
  f.end { |when_| when_.upper }
end

[flash-gordon]

this is super minor but I wouldn't reveal the details in the name of the method, that is something like dependency_name suits better. Next time you change the rules you'll need to change only the body of the method

[GustavoCaso]

Totally agree with it @flash-gordon 👍. done 317cf70

[solnic]

Maybe we should switch to kwargs for this feature?

[flash-gordon]

@solnic I thought about it, it's not gonna work, you still need to access local variable with reserved words

proc { |when:| "how to access >when< here?" }

[timriley]

I guess binding.local_variable_get is a bit too much to ask of end users.

[timriley]

Another option if using kwargs would be to splat them and pull it off the hash:

proc { |**args| "we can get it as #{args.fetch(:when)}" }

[flash-gordon]

@timriley ** not gonna work either, rom-factory uses deps for figuring out the order in which callable values have to be called (via tsort).

[timriley]

@flash-gordon Ah, of course. Well in this case a trailing underscore sounds like the least worse solution 😅

[GustavoCaso]

@solnic @flash-gordon @timriley I know is not ideal but shall we push it?

[solnic]

OK I need to think about this. Maybe this should be based on lazy evaluation, ie pass an object and whenever you call a method on it, it will generate a value based on factory definition. ie

f.email { fake(:internet, :email) }
f.login { |s| s.email }

[flash-gordon]

@solnic this can be slower because may require a wrapper but I'm not sure

[solnic]

@flash-gordon yeah but it won't require tsorting

[flash-gordon]

@solnic sorting is required just once whereas wrapping is required on every instantiation, no?

[solnic]

@flash-gordon no, we need to create a wrapper for each factory object, so once per factory definition, then we just pass this one to each block when required

[GustavoCaso]

@solnic @flash-gordon I can have a look and try to implement it, or is anyone interested in do it?

[flash-gordon]

@GustavoCaso depends on whether @solnic's thinking process complete :)

[solnic]

Yeah I'd at least try to do it using a wrapper with injected factory, so that it can delegate to it in order to return a value from a factory. If we make it coercible to a hash, then as a bonus something like f.foo { |bar:| } would make bar available via wrapper object, which I think would be cool.

So, Gustavo, if you're interested in experimenting with this, I'd say go ahead :)

[flash-gordon]

|bar:| won't work, only |bar:, **| will

[solnic]

@flash-gordon gahhh right ;( but it still makes sense to experiment with this

[solnic]

Closing due to lack of activity. The related issue remains open, so we should revisit this before 1.0.0.