Feature #7341



Added by nathan.f77 (Nathan Broadbent) about 10 years ago. Updated almost 5 years ago.

Target version:


Jeremy Kemper proposed Enumerable#associate during the discussion in #7297, with the following details:

Some background:

#4151 proposes an Enumerable#categorize API, but it's complex and hard to understand its behavior at a glance.
#7292 proposes an Enumerable#to_h == Hash[...] API, but I don't think of association/pairing as explicit coercion, so #to_h feels misfit.

Associate is a simple verb with unsurprising results. It doesn't introduce ambiguous "map" naming. You associate an enumerable of keys with yielded values.

Some before/after examples:

Before: Hash[ { |filename| [ filename, download_url(filename) ]}]
After: filenames.associate { |filename| download_url filename }

=> {"foo.jpg"=>"http://...", ...}

Before: alphabet.each_with_index.each_with_object({}) { |(letter, index), hash| hash[letter] = index }
After: alphabet.each_with_index.associate

=> {"a"=>0, "b"=>1, "c"=>2, "d"=>3, "e"=>4, "f"=>5, ...}

Before: keys.each_with_object({}) { |k, hash| hash[k] = self[k] } # a simple Hash#slice
After: keys.associate { |key| self[key] }

It's worth noting that this would compliment ActiveSupport's Enumerable#index_by method:
#index_by produces '{ => el, ...}', while #associate would produce '{el => , ...}'.

For cases where you need to control both keys and values, you could use '[1,2,3].map{|i| [i, i * 2] }.associate', or continue to use 'each_with_object({})'.

Updated by bitsweat (Jeremy Daer) about 10 years ago

Thanks for posting, Nathan. See for the full pitch and a demonstration implementation.

In short: associating a collection of keys with calculated values should be easy to do and the code should reflect the programmer's intent. But it's hard for a programmer to discover which API is appropriate to achieve this. Hash[] and each_with_object({}) seem unrelated. And using these API requires boilerplate code that obscures the programmer's intent.

Must write code to build a Hash[] argument in the format it expects:

an array of [key, value] pairs. The intent is hidden by unrelated code

needed to operate the Hash[] method.

Hash[* { |elem| [elem, calculate(elem)] }]

This is better. Much less boilerplate code. But the programmer is

reimplementing association every time: providing a hash and setting the

value for each key in the collection. This is what an implementation

of association looks like. It shouldn't be repeated in our code.

collection.each_with_object({}) { |elem, hash| hash[elem] = calculate(elem) }

Most Rubyists just use this instead. It uses simple, easy-to-discover API.

But it suffers the same issues: it's an implementation of association

that's now repeated in our code, blurring its intent. And it forces us to

disrupt chains of enumerable methods and write boilerplate code.

hash = {}
collection.each { |element| hash[element] = calculate(element) }

Now the code is stating precisely what the programmer wants to achieve.

Associate is easy to find in docs and uses a verb that "rings a bell" to

programmers who need to associate keys with yielded values.

collection.associate { |element| calculate element }

Marc-André Lafortune proposed a similar Enumerable#associate in #4151. The basic behavior is the same, so I consider that a point in favor of this method name. It associates values with the enumerated keys. He introduces additional collision handling that I consider out of scope. For more complex scenarios, using more verbose, powerful API like #inject, #each_with_object, or #map + #associate feels appropriate.

Updated by marcandre (Marc-Andre Lafortune) about 10 years ago

  • Category changed from lib to core
  • Priority changed from 3 to Normal


bitsweat (Jeremy Kemper) wrote:

In short: associating a collection of keys with calculated values should be easy to do and the code should reflect the programmer's intent.

A strong +1 from me


A good start. I'd make one important change: return an enumerator when no block is given. Here's why:

  1. The form you suggest would be redundant with Enumerable#to_h

  2. It would be more powerful, for example to associate things that need an index...

    rng.each_with_index.associate {|elem, index| ....} # => { [elem, index] => ... }, not what you want

    Easy this form:

    rng.associate.with_index {|elem, index| ... } # => { elem => ... }

  3. Consistency with modern methods dealing with enumerable.

Updated by nathan.f77 (Nathan Broadbent) about 10 years ago

  1. The form you suggest would be redundant with Enumerable#to_h

I agree that 'Enumerable#to_h' would seem more appropriate than the
block-less version of 'associate'. To me, the 'associate' verb implies that
the programmer will provide some logic to determine how the elements will
be associated. So I also feel that invocation without a block should return
an enumerator.

However, if 'to_h' is rejected and 'associate' is all we have to work with,
then it would probably be more useful to make 'associate' 'multi-purpose'
in the way that is currently proposed.

Updated by Anonymous about 10 years ago

Agree with Marc-Andre.

Updated by trans (Thomas Sawyer) about 10 years ago

One problem I have with this is the terminology. The term "associate" already applies to arrays. ((Associative arrays)) are arrays of arrays where the first element of an inner array acts a key for the rest.

[[:a,1],[:b,2]].assoc(:a)  #=> [:a,1]

For this reason I would expect an #associate method to take a flat array and group the elements together.

[:a,1,:b,2].associate  #=> [[:a,1],[:b,2]]

An argument could determine the number elements in each group, the default being 2.

Since Hash#to_a returns an associative array, to me it makes sense that Array#to_h would reverse the process.

{:a=>1,:b=>2}.to_a    #=> [[:a,1],[:b,2]]
[[:a,1],[:b,2]].to_h  #=> {:a=>1,:b=>2}

Putting the two together, your version of associate is easy enough to achieve:


As it turns out, with the exception of the default argument, #associate is same as #each_slice. But I think it would be nice to have #associate around for it's default and the fact that it reads better in these cases.


Updated by Anonymous about 10 years ago

@Tom: Associative arrays are nice, but they are just arrays. No need to pamper them too much in the core.

Updated by Ajedi32 (Andrew M) over 8 years ago

This is related to #6669

Actions #8

Updated by naruse (Yui NARUSE) almost 5 years ago

  • Target version deleted (2.6)

Also available in: Atom PDF