[lnkForumImage]
TotalShareware - Download Free Software
Confronta i prezzi di migliaia di prodotti.
Asp Forum
 Home | Login | Register | Search 


 

Forums >

comp.lang.ruby

stable sort_by?

Patrick Gundlach

12/13/2005 10:47:00 AM

Hi,

I have to sort a list using a number of criterions (a,b,c,d) where d
is the least important criterion and a is the most important one. All
comparisons should be 'higher number: better position'. This is what I
have tried:

--------------------------------------------------
require 'pp'

h= {"Team A"=>{:a=>3, :b=>2, :c=>6, :d=>114 },
"Team B"=>{:a=>0, :b=>-2, :c=>4, :d=>112 },
"Team C"=>{:a=>3, :b=>4, :c=>4, :d=>110 },
"Team D"=>{:a=>3, :b=>4, :c=>4, :d=>108 },
}

tmp=h.to_a

[:d,:c,:b,:a].each do |criterion|
tmp=tmp.sort_by { |a| a[1][criterion] }.reverse
end

pp tmp

--------------------------------------------------

[["Team D", {:d=>108, :a=>3, :b=>4, :c=>4}],
["Team C", {:d=>110, :a=>3, :b=>4, :c=>4}],
["Team A", {:d=>114, :a=>3, :b=>2, :c=>6}],
["Team B", {:d=>112, :a=>0, :b=>-2, :c=>4}]]


but as far as I can see 'Team C' should be better than 'Team D',
because of criterion d. Is it possible that sort_by is not stable? Or
is there something I did wrong?

Patrick
27 Answers

Patrick Gundlach

12/13/2005 11:18:00 AM

0

A follow-up:

> --------------------------------------------------
> require 'pp'
>
> h= {"Team A"=>{:a=>3, :b=>2, :c=>6, :d=>114 },
> "Team B"=>{:a=>0, :b=>-2, :c=>4, :d=>112 },
> "Team C"=>{:a=>3, :b=>4, :c=>4, :d=>110 },
> "Team D"=>{:a=>3, :b=>4, :c=>4, :d=>108 },
> }

this seems to be stable:

pp h.sort{ |a,b|
res = a[1][:a] <=> b[1][:a]
if res == 0
res = a[1][:b] <=> b[1][:b]
end
if res == 0
res = a[1][:c] <=> b[1][:c]
end
if res == 0
res = a[1][:d] <=> b[1][:d]
end
res
}.reverse

But I still would like to know if sort_by is unstable (or if am I
doing something wrong).

Patrick

Frederick Ros

12/13/2005 11:37:00 AM

0

Patrick Gundlach wrote:
> Hi,
>
> I have to sort a list using a number of criterions (a,b,c,d) where d
> is the least important criterion and a is the most important one. All
> comparisons should be 'higher number: better position'. This is what I
> have tried:
>
> --------------------------------------------------
> require 'pp'
>
> h= {"Team A"=>{:a=>3, :b=>2, :c=>6, :d=>114 },
> "Team B"=>{:a=>0, :b=>-2, :c=>4, :d=>112 },
> "Team C"=>{:a=>3, :b=>4, :c=>4, :d=>110 },
> "Team D"=>{:a=>3, :b=>4, :c=>4, :d=>108 },
> }
>
> tmp=h.to_a
>
> [:d,:c,:b,:a].each do |criterion|
> tmp=tmp.sort_by { |a| a[1][criterion] }.reverse
> end
>
> pp tmp
>
> --------------------------------------------------
>
> [["Team D", {:d=>108, :a=>3, :b=>4, :c=>4}],
> ["Team C", {:d=>110, :a=>3, :b=>4, :c=>4}],
> ["Team A", {:d=>114, :a=>3, :b=>2, :c=>6}],
> ["Team B", {:d=>112, :a=>0, :b=>-2, :c=>4}]]
>
>
> but as far as I can see 'Team C' should be better than 'Team D',
> because of criterion d. Is it possible that sort_by is not stable? Or
> is there something I did wrong?


Wouldn't this be simpler ? :

h.sort_by {|k,v| [v[:a],v[:b],v[:c],v[:d]]}.reverse

Best Regards,
Fred

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

Christer Nilsson

12/13/2005 12:10:00 PM

0

Or using Array instead of Hash:

I don't think sort is stable. You have to specify the sorting criteria,
and sort only once.

require 'test/unit'
class TestSortBy < Test::Unit::TestCase
def test_all
a= [["A", 3, 2, 6, 114],
["B", 0,-2, 4, 112],
["C", 3, 4, 4, 110],
["D", 3, 4, 4, 108]]

assert_equal [["A", 3, 2, 6, 114],
["C", 3, 4, 4, 110],
["D", 3, 4, 4, 108],
["B", 0, -2, 4, 112]],
a.sort_by {|name,x,y,z,w| [x, w] }.reverse
end
end

Christer

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

Patrick Gundlach

12/13/2005 12:24:00 PM

0


[...]

> Wouldn't this be simpler ? :
>
> h.sort_by {|k,v| [v[:a],v[:b],v[:c],v[:d]]}.reverse

Oh, yes, that's so much nicer. Thanks.

Patrick

dblack

12/13/2005 1:25:00 PM

0

Frederick Ros

12/13/2005 1:39:00 PM

0

unknown wrote:

>> h.sort_by {|k,v| [v[:a],v[:b],v[:c],v[:d]]}.reverse
>
> Or:
>
> h.sort_by {|k,v| v.values_at(:a,:b,:c,:d) }.reverse
>
> (Possibly a little clearer visually.)


Yeah ! This one rocks !

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

Patrick Gundlach

12/13/2005 1:55:00 PM

0


>> h.sort_by {|k,v| v.values_at(:a,:b,:c,:d) }.reverse
>>
>> (Possibly a little clearer visually.)
>
>
> Yeah ! This one rocks !

Right, and it passes all my unittests :). So beautiful!

Patrick

Kevin Olbrich

12/13/2005 2:50:00 PM

0

Does the principle of 'least surprise' suggest that the default sort_by
algorithm should be modified to be stable?

It seems to me that a 'stable' algorithm is the expected behavior.

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

Mike Fletcher

12/13/2005 2:57:00 PM

0

Frederick Ros wrote:
> h.sort_by {|k,v| [v[:a],v[:b],v[:c],v[:d]]}.reverse

OK, this or something like it should be in the rdoc for sort_by. I was
doing something similar and wound up writing a similar sort with
multiple if tests like the OP because I didn't remember that Array
implements a sane <=>.

Also, would there be a similar cool way to do this in one step where you
want a mix of ascending and descending sorts on different parts? Say I
had a hash (in YAML):

Bobby:
age: 11
lastname: Smith
Suzy:
age: 13
lastname: Jones
Ted:
age 12
lastname: Smith

And I wanted to sort
- alphabetically by lastname
- then age oldest to youngest

So in Perl I'd do something like:

my @sorted_keys = sort { $a->{lastname} cmp $b->{lastname}
|| $b->{age} <=> $a->{age} }
keys %data;

In this case I could use

sorted_keys = data.sort_by { |k,v| [ v["lastname"], -1*v["age"] ] }

But what if I wanted reverse alphabetically by lastname (Z-A)? In perl
I'd swap $a and $b in the first comparison, but I can't think of a
spiffy way to do the same using sort_by.

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

Christian Neukirchen

12/14/2005 5:20:00 PM

0

Patrick Gundlach <clr10.10.randomuser@spamgourmet.com> writes:

> pp h.sort{ |a,b|
> res = a[1][:a] <=> b[1][:a]
> if res == 0
> res = a[1][:b] <=> b[1][:b]
> end
> if res == 0
> res = a[1][:c] <=> b[1][:c]
> end
> if res == 0
> res = a[1][:d] <=> b[1][:d]
> end
> res
> }.reverse

Checkout Numeric#nonzero?.

> Patrick
--
Christian Neukirchen <chneukirchen@gmail.com> http://chneuk...