Ruby 2.7 — Enumerable#tally
Christmas has come and passed, 2.6 has been released, and now it’s time to mercilessly hawk the releases page for 2.7 features so we can start our fun little annual tradition of blogging about upcoming features.
Typically this means another December release, but there have been cases of methods making it in earlier if they’re merged to trunk this early in the year.
This round? We have the new method Enumerable#tally
!
The Short Version
tally
counts things:
[1, 1, 2].tally
# => { 1 => 2, 2 => 1 }[1, 1, 2].map(&:even?).tally
# => { false => 2, true => 1 }
Examples
The example used in Ruby’s official test code is:
[1, 2, 2, 3].tally
# => { 1 => 1, 2 => 2, 3 => 1 }
Without a block, tally
works by counting the occurrences of each element in an Enumerable type. If we apply that to a list of another type it might be a bit clearer:
%w(foo foo bar foo baz foo).tally
=> {"foo"=>4, "bar"=>1, "baz"=>1}
Currently tally_by has not been accepted into core, so to tally by a function you would instead use map first:
%w(foo foo bar foo baz foo).map { |s| s[0] }.tally
=> {“f” => 4, “b” => 2}
There’s discussion happening at the moment on accepting this feature, which would make the above syntax:
%w(foo foo bar foo baz foo).tally_by { |s| s[0] }
=> {“f” => 4, “b” => 2}
Why Use It?
If you’ve been using Ruby, chances are you’ve used code something like one of these lines to do the same thing tally
is doing above:
list.group_by { |v| v.something }.transform_values(&:size)list.group_by { |v| v.something }.map { |k, vs| [k, vs.size] }.to_hlist.group_by { |v| v.something }.to_h { |k, vs| [k, vs.size] }list.each_with_object(Hash.new(0)) { |v, h| h[v.something] += 1 }
There are likely several more variants of this, but those are a few of the more common ones you might see around. This is a nicety method to abbreviate a very common idiom in the Ruby language, and a very welcome one.
Vanilla Ruby Equivalent
What does this method do? Well if we were to implement it in plain Ruby it might look a bit like this:
module Enumerable
def tally_by(&function)
function ||= -> v { v }
each_with_object(Hash.new(0)) do |value, hash|
hash[function.call(value)] += 1
end
end
def tally
tally_by(&:itself)
end
end
In the case of no provided function, it would effectively be tallying by itself
, or rather an identity function.
An identity function is a function that returns what it was given. If you give it
1
, it returns1
. If you give ittrue
, it returnstrue
. Ruby also uses this concept in a method calleditself
.
This article will not go into great depth on what the above code does. Part Five of “Reducing Enumerable” covers this code in more detail:
The Source Code
Nobu recently committed a patch to Ruby core to add this method:
It was accepted by the Ruby core team under the name tally
:
Tally?
Let’s start with what the word means:
A tally is a record of amounts or numbers which you keep changing and adding to as the activity which affects it progresses.
- https://www.collinsdictionary.com/us/dictionary/english/tally
Now where did this name come from? Originally the name count_by
was proposed, but the name was rejected as it differed from count
which has a different return type and behavior.
On a car ride back from the Tahoe area and RailsCamp West we (myself, David, Stephanie, and Shannon) were discussing potentially alternate names to try and propose to see if the feature could get in under a different name.
David had proposed tally
, and formally suggested it. It looks like the name stuck, and the code’s been merged into trunk.
Now I’d presented a talk at a few conferences, and decided to mention tally_by
instead of count_by
in my RubyConf talk in one section. The written version is over here:
Just a bit of interesting backstory.
Wrapping Up
2.7 is on its way, let’s see what it’ll bring! I’m looking forward to seeing where Ruby goes from here.