Sort array by two attributes? (like sql "order by A, B")

ruby-talk
1

IN sql we can pass two arguments to the 'order by' component, and it
will order the results by A, and then B in the cases where A is the
same.

Can anyone think of a way to do the same thing with a ruby array? For
example, if the array holds objects that have attributes/methods
"lastname" & "firstname", to order the objects in a similar way to the
sql query?

thanks
max

···

--
Posted via http://www.ruby-forum.com/.

2

Can anyone think of a way to do the same thing with a ruby array? For
example, if the array holds objects that have attributes/methods
"lastname" & "firstname", to order the objects in a similar way to the
sql query?

Just use Enumerable#sort:
http://ruby-doc.org/core/classes/Enumerable.html#M003150

objects.sort do |a,b|
  comp = (a.lastname <=> b.lastname)
  comp.zero? ? (a.firstname <=> b.firstname) : comp
end

James
http://blog.jcoglan.com

3

Here's one way: the trick is to create an array with the fields you want
to order by in the sort_by block:

irb(main):001:0> class A
irb(main):002:1> attr_accessor :a,:b
irb(main):003:1> def initialize a,b
irb(main):004:2> @a = a
irb(main):005:2> @b = b
irb(main):006:2> end
irb(main):007:1> end
=> nil
irb(main):008:0> ary = [A.new(1,2), A.new(1,3), A.new(1,1),
A.new(2,3), A.new(2,1)]
=> [#<A:0xb7b68890 @b=2, @a=1>, #<A:0xb7b6887c @b=3, @a=1>,
#<A:0xb7b68868 @b=1, @a=1>, #<A:0xb7b68854 @b=3, @a=2>, #<A:0xb7b68840
@b=1, @a=2>]
irb(main):009:0> ary.sort_by {|x| [x.a,x.b]}
=> [#<A:0xb7b68868 @b=1, @a=1>, #<A:0xb7b68890 @b=2, @a=1>,
#<A:0xb7b6887c @b=3, @a=1>, #<A:0xb7b68840 @b=1, @a=2>, #<A:0xb7b68854
@b=3, @a=2>]

Hope this helps,

Jesus.

···

On Mon, Aug 11, 2008 at 5:16 PM, Max Williams <toastkid.williams@gmail.com> wrote:

IN sql we can pass two arguments to the 'order by' component, and it
will order the results by A, and then B in the cases where A is the
same.

Can anyone think of a way to do the same thing with a ruby array? For
example, if the array holds objects that have attributes/methods
"lastname" & "firstname", to order the objects in a similar way to the
sql query?

4

I would just create an array containing the parts you want to compare:

objects.sort { |a,b| [a.lastname, a.firstname] <=> [b.lastname,
b.firstname] }

···

--
Posted via http://www.ruby-forum.com/.

5

Jesús Gabriel y Galán wrote:

Here's one way: the trick is to create an array with the fields you want
to order by in the sort_by block:
Hope this helps,

Jesus.

That's a neat trick, thanks! I actually like the first one better
though as you can swap a & b around in each test to get (eg) sorted by
name ascending and then a date field descending.

Thanks a lot guys!
max

···

--
Posted via http://www.ruby-forum.com/\.

6

For this usecase there's also Numeric#nonzero?

  objects.sort do |a, b|
    (a.lastname <=> b.lastname).nonzero? ||
    (a.firstname <=> b.firstname)
  end

Regards,
Pit

···

2008/8/11 James Coglan <jcoglan@googlemail.com>:

objects.sort do |a,b|
comp = (a.lastname <=> b.lastname)
comp.zero? ? (a.firstname <=> b.firstname) : comp
end

7

In theory this is less efficient since there are two Array instances
created per comparison. If you go for the Array solution, using
#sort_by is probably better:

objects.sort_by {|a| [a.lastname, a.firstname]}

Kind regards

robert

···

On Mon, Feb 18, 2013 at 6:42 PM, Jack V. <lists@ruby-forum.com> wrote:

I would just create an array containing the parts you want to compare:

objects.sort { |a,b| [a.lastname, a.firstname] <=> [b.lastname,
b.firstname] }

--
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/

8

That's a neat trick, thanks! I actually like the first one better
though as you can swap a & b around in each test to get (eg) sorted by
name ascending and then a date field descending.

Just use -x.b:

a.sort_by{|x| [x.name, -x.date]}

Sort_by is much faster than sort with a block.

gegroet,
Erik V.

9

Pit Capitain wrote:

  objects.sort do |a, b|
    (a.lastname <=> b.lastname).nonzero? ||
    (a.firstname <=> b.firstname)
  end

Wow, this is all great stuff. thanks folks.

···

--
Posted via http://www.ruby-forum.com/\.

10

Sort_by is much faster than sort with a block.

This is often true as sort_by only calls methods on each object once, but it
is not universally true -- see
module Enumerable - RDoc Documentation for more
information, and run some benchmarks for your use case if it's a big issue.

11

Doesn't always work. I keep this handy:

class RevCmp
    attr_reader :this

    def initialize(obj)
      @this = obj
    end

    def <=>(other)
      other.this <=> @this
    end

    # not delegating anything else because this is explicitly a throwaway
    # object used only inside a sort_by block
end

and then you have

  a.sort_by {|x| [x.name, RevCmp.new(x.date)]}

you could even use a top-level method so you can say

  a.sort_by {|x| [x.name, descending(x.date)]}

where descending(x) returns RevCmp.new(x)

You could even mix it into object to get

  a.sort_by {|x| [x.name, x.date._descending_]}

where I use the underscores as a cosmetic way of making it stand out
inside the sort block

martin

···

On Mon, Aug 11, 2008 at 8:48 AM, Erik Veenstra <erikveen@gmail.com> wrote:

That's a neat trick, thanks! I actually like the first one better
though as you can swap a & b around in each test to get (eg) sorted by
name ascending and then a date field descending.

Just use -x.b:

a.sort_by{|x| [x.name, -x.date]}