Many seems to be broken

[Bastes]

First of all, thanks for your talk about Ruby monads, it blew my mind ^^

I've been trying to imitate your logic as a TDD exercise, and was comparing what I was writing to your code and found a difference puzzling me in the "Many" monad.

When I'm trying to use yours, I find that:

Blaher = Struct.new(:blahs)
Bloher = Struct.new(:blohs)
blahers = [Blaher.new([Bloher.new([1, 2, 3, 4])])]
Monads::Many.new(blahers).blahs.blohs.values

Ends up in a:

NoMethodError: undefined method `bloh' for [#<struct bloh=[1, 2, 3, 4]>]:Array

I thought it had something to do with the Many.from_value method and tried a quick-and-dirty fix with:

def from_value(value)
  Many.new(value)
end

It makes the method_missing do its job and return the subelements expected, though I find one test (spec/monads/many_spec.rb:49 the one that verifies that Many.from_value wraps the value in an array before wrapping it in the monad) fails.

I'm a bit unsure on what to do to fix this, so I thought I'd let you know ^^°

Again, thanks for the awesome talk and ideas.

[Bastes]

I also tried another approach, changing the Monad.within method:

def within(&block)
  and_then do |value|
    self.class.new(block.call(value))
  end
end

Which, in turn, breaks the eventually specs ^^°:

spec/monads/eventually_spec.rb:73
spec/monads/eventually_spec.rb:83
spec/monads/eventually_spec.rb:102
spec/monads/eventually_spec.rb:107

Because it tries to initialize the monad giving it a block, not a proc as a parameter. I'm beginning to think part of the factoring went wrong somewhere, but I'm not yet sure what and where. I'll be trying to find that when I'm rested.

[sclinede]

Hi. Can you describe what exactly you are expecting from your code? Maybe you have more meaningful example. I don't understand what do you think is broken...

[Bastes]

Hi @sclinede, it's been a while so I may be fuzzy on the details, but I submitted a pull request to fix the thing I think was broken.

What was broken?

# an author writes books...
Author = Struct.new(:books)
# ...a book contains chapters...
Book = Struct.new(:chapters)

# ...here are some books, with or without chapters...
books = [
  Book.new([1, 2, 3, 4]),
  Book.new([5, 6, 7]),
  Book.new([]),
  Book.new([8, 9])
]

# ...here are authors with or without books...
authors = [
  Author.new(books),
  Author.new([])
]

# ...I'm trying to use the Many monad to extract the chapters from all the books from all the authors
Monads::Many.new(authors).books.chapters.values

# here is what I expected to obtain:
=> [1, 2, 3, 4, 5, 6, 7, 8, 9]

# but what I actually got is the following error:
# NoMethodError: undefined method `chapters' for #<Array:0x0055f546fb4e88>
# from lib/monads/monad.rb:11:in `public_send'
# from lib/monads/monad.rb:11:in `block in method_missing'
# from lib/monads/monad.rb:5:in `call'
# from lib/monads/monad.rb:5:in `block in within'
# from lib/monads/monad.rb:21:in `call'
# from lib/monads/monad.rb:21:in `block in ensure_monadic_result'
# from lib/monads/many.rb:10:in `map'
# from lib/monads/many.rb:10:in `and_then'
# from lib/monads/monad.rb:4:in `within'
# from lib/monads/monad.rb:10:in `method_missing'

Now here's why I think it went wrong, and how I fixed it:

def and_then(&block)
  block = ensure_monadic_result(&block)

  # this line looks wrong, because flat_map doesn't do what was expected of it
  Many.new(values.map(&block).flat_map(&:values))
end

# when I do this:
authors.map { |a| Monads::Many.from_value(a.books) }.flat_map(&:values)
# I get:
=> [[#<struct Book chapters=[1, 2, 3, 4]>, #<struct Book chapters=[5, 6, 7]>, #<struct Book chapters=[]>, #<struct Book chapters=[8, 9]>], []]

# But, wait, I hear you ask, this is not flat, not at all!!!
# No, it isn't. flat_map didn't do (map.flatten), it did (flatten.map)!

# Well then, instead of relying on flat_map, like this:
Many.new(values.map(&block).flat_map(&:values))
# We should do a map, then flatten it, likewise:
Many.new(values.map(&block).map(&:values).flatten)

[sclinede]

OK, now I see. I think map + flatten is right solution. @tomstuart, what about you?