Feature #3131: add Kernel#Hash() method like Kernel#Array()

I forgot to show that case 2 supports empty arrays:

>> Hash([])
=> {}

Just like Kernel#Array() supports empty hashes:

>> Array({})
=> []

Thanks for your consideration.  And sorry for the noise.

Hi,

In message "Re: [ruby-core:29462] [Feature #3131] add Kernel#Hash() method like Kernel#Array()"
    on Mon, 12 Apr 2010 08:32:52 +0900, Suraj Kurapati <[e-mail]> writes:

|There is an imbalance of power in the Ruby core API (when it comes 
|to arrays and hashes) because it is easier to convert nil values 
|into empty arrays, thanks to Kernel#Array(), than it is to convert 
|nil values into empty hashes, due to lack of Kernel#Hash().

Having Hash() might be a good idea.  But since conversion from
arrays only meaningful for specific case (array of 2-element arrays),
I am not sure whether Hash() should support conversion from Array in
general or not.

							matz.

Hi,

matz wrote:
> since conversion from arrays only meaningful for specific case
> (array of 2-element arrays), I am not sure whether Hash() should
> support conversion from Array in general or not.

Hash::[] does not support conversion from odd-length Array (because
it is uncertain what Ruby must do with the 2n+1'th element) and will
raise an error in such case.  So, in my view, Kernel#Hash() need not
support conversion from odd-length Array either.

If you wish to avoid even-length Array conversion in Kernel#Hash()
because odd-length Array conversion is not supported, then I request
that Kernel#Hash() must still support conversion from empty Array
because Kernel#Array() already supports conversion from empty Hash:

>> Array({})
=> []

In summary, I request that Kernel#Hash() supports conversion from:

* nil
* Hash
* empty Array

It would be nice if Kernel#Hash() also supports conversion from:

* even-length Array

Thanks for your consideration.

Hi,

Please allow me to clarify.

Suraj Kurapati wrote:
> Hash::[] does not support conversion from odd-length Array (because
> it is uncertain what Ruby must do with the 2n+1'th element) and will
> raise an error in such case.

I was referring to this particular behavior:

>> Hash[1,2,3,4,5,6]
=> {1=>2, 3=>4, 5=>6}

>> Hash[1,2,3,4,5,6,7]
ArgumentError: odd number of arguments for Hash
	from (irb):3:in `[]'
	from (irb):3
	from /usr/bin/irb:12:in `<main>'

I forgot that Hash::[] can convert Array of 2-element Array into Hash:

>> Hash[1,2,3,4,[5,6],[7]]            # non 2-element Array ignored
=> {1=>2, 3=>4, [5, 6]=>[7]}

>> Hash[[1,2,3,4,[5,6],[7]]]          # implicit pair value is nil
=> {5=>6, 7=>nil}

>> Hash[[1,2,3,4,[5,6],[7,8],[7]]]    # previous pair is overwritten
=> {5=>6, 7=>nil}

Thanks for your consideration.

Hi Matz,

To avoid delaying this request forever due to unnecessary features,
I have narrowed the requirements for Kernel#Hash() to the following:

Hash() must convert (1) nil, (2) Hash, and (3) empty Array into Hash.

    module Kernel
      def Hash(value)
        if value.respond_to? :to_hash
          value.to_hash
        elsif value.nil? or Array(value).empty?
          {}
        else
          raise ArgumentError, "invalid value for Hash: #{value}"
        end
      end
    end

Thanks for your consideration.

Hi,

In message "Re: [ruby-core:29644] [Feature #3131] add Kernel#Hash() method like Kernel#Array()"
    on Tue, 20 Apr 2010 13:36:00 +0900, Suraj Kurapati <[e-mail]> writes:

|To avoid delaying this request forever due to unnecessary features,
|I have narrowed the requirements for Kernel#Hash() to the following:
|
|Hash() must convert (1) nil, (2) Hash, and (3) empty Array into Hash.

Sounds reasonable.  But the new feature window for 1.9.2 was closed.
So, I will consider (positively) adding this feature after 1.9.2 release.

							matz.

• Status changed from Open to Assigned
• Assignee set to matz (Yukihiro Matsumoto)

Hi,

Matz wrote:
> I will consider (positively) adding this feature after 1.9.2
> release.

Thanks Matz!  In the mean time, I have created a gem with this
functionality so that people can start using it, if they wish:

  http://rubygems.org/gems/kernel_hash

The source code (and unit tests) is available here:

  http://github.com/sunaku/kernel_hash

Cheers.

Hi,

Please set the "Done %" for this issue to 80%.

The remaining 20% is for (possibly) rewriting
my proposed Kernel#Hash() Ruby code in C.

Thanks for your consideration.

• File kernel-hash.patch added

I believe the attached patch reflects the consensus between matz and Suraj. The only material difference is that a TypeError is raised instead of an ArgumentError for consistency with Integer(), Float(), etc.

Integer() and Float() in Ruby 1.9.2 raise TypeError and ArgumentError:

$ irb
## ruby 1.9.2p180 (2011-02-18 revision 30909) [x86_64-linux]
>> Float(nil)
TypeError: can't convert nil into Float
	from (irb):6:in `Float'
	from (irb):6
	from /usr/bin/irb:12:in `<main>'
>> Float('meow')
ArgumentError: invalid value for Float(): "meow"
	from (irb):8:in `Float'
	from (irb):8
	from /usr/bin/irb:12:in `<main>'

I don't know which error type we should raise in Kernel#Hash().

Hi,

In message "Re: [ruby-core:35857] [Ruby 1.9 - Feature #3131] add Kernel#Hash() method like Kernel#Array()"
    on Sat, 23 Apr 2011 08:52:12 +0900, Suraj Kurapati <[e-mail]> writes:

|Integer() and Float() in Ruby 1.9.2 raise TypeError and ArgumentError:

I think they should raise TypeError, if incompatibly here is not an
issue, I will fix.

							matz.

Any chance of this getting into Ruby 1.9.3? Thanks.

On Mon, Sep 26, 2011 at 1:52 AM, Suraj Kurapati <[e-mail]> wrote: > > Issue #3131 has been updated by Suraj Kurapati. > > Any chance of this getting into Ruby 1.9.3?  Thanks. Doesn't look like it. See ruby-core 39693. -- Anurag Priyam

Thanks for the pointer. I'll just wait for Ruby 1.9.4 then.

I accept this patch for trunk, except that Hash(nil) should return an empty hash, not nil. Matz.

Who should work on it?

• Status changed from Assigned to Closed

This has already done, only missed to close.