Struct gets coerced into **kwargs (when *args is also present) about dry-struct HOT 7 CLOSED

Thanks for clarification

from dry-struct.

Monkey-patch fix:

class Dry::Struct
  undef_method :to_hash

  def to_h
    self.class.schema.keys.each_with_object({}) do |key, result|
      result[key] = Hashify[self[key]]
    end
  end
end

But I am not 100% of the implications of it

from dry-struct.

Yeah, I know that, this is how ruby works, they really messed with it IMO. Structs are supposed to work as hashes so this is expected. I wouldn't go with a monkey patch though, you can subclass Dry::Struct and use it as a base where you need

class MyStruct < Dry::Struct
  undef_method :to_hash
end

^also, you don't need to override to_h, undefining of to_hash doesn't affect it.

from dry-struct.

You say "Structs are supposed to work as hashes", but I wonder where is this requirement coming from.

In Ruby neither Struct nor OpenStruct implements the implicit conversion method to_hash. Both implement the explicit conversion method #to_h
More on the differences between implicit and explicit conversion methods here.

irb(main):001:0> A = Struct.new(:a, :b)
=> A
irb(main):002:0> def iii(**kwargs)
irb(main):003:1> puts kwargs
irb(main):004:1> end
=> :iii
irb(main):005:0> iii A.new(1,2)
Traceback (most recent call last):
        3: from /Users/sevos/.rbenv/versions/2.5.0/bin/irb:11:in `<main>'
        2: from (irb):5
        1: from (irb):2:in `iii'
ArgumentError (wrong number of arguments (given 1, expected 0))
irb(main):006:0> iii A.new(1,2).to_h
{:a=>1, :b=>2}
=> nil
irb(main):009:0> iii OpenStruct.new(a: 1, b: 2)
Traceback (most recent call last):
        3: from /Users/sevos/.rbenv/versions/2.5.0/bin/irb:11:in `<main>'
        2: from (irb):9
        1: from (irb):2:in `iii'
irb(main):010:0> iii OpenStruct.new(a: 1, b: 2).to_h
{:a=>1, :b=>2}
=> nil
irb(main):007:0>

I am afraid that there may be more cases when the struct may be deconstructed into a hash prematurely or accidentally causing us to lose the information about the type of the data it carries. Please tell me what are the cases you find this behaviour useful? Maybe there is really more good than harm...
Personally I use Dry::Struct as a base class for business domain entities and it's more of a better-Struct than a better-Hash, so I would prefer not to deconstruct these objects easily. Of course there should be a way to convert it to a hash, when needed.

Thanks for the idea for the base class, but now I worry more about dry-struct compatibility with other libraries, which may deconstruct it accidentally.

from dry-struct.

This is not a requirement, this is a design choice. One quick example is rom's interface for writing data, it relies on coercions made by Hash[value], anything hash-like can be passed in, including structs.

from dry-struct.

For the record, this also just bit me. I was definitely not expecting my entity object to deconstruct into a hash because I had a keyword arg in a composite class initializer.

from dry-struct.

FYI: This does not even require **kwargs, regular keyword arguments are enough.

def display(*items, color: true)
  puts "items: %p color: %p" % [items, color]
end

Calling this with a Dry::Struct causes: ArgumentError : unknown keywords: item_attribute, item_attribute2, item_attribute_x

There's an issue (that originated from using Dry::Validation :)) at https://bugs.ruby-lang.org/issues/14909 and they suggest "just wait for ruby 3"

from dry-struct.