Preserve insert order in a Hash


#1

Hi,
I would like to know if it’s possible to insert values in a Hash and
then extract all of them in the same insertion order.

I noticed that if I then try to print the content of an Hash using
“each” the order is not the same as the insertion.

Thanks


#2

From: Me Me [mailto:removed_email_address@domain.invalid]

I would like to know if it’s possible to insert values in a Hash and

then extract all of them in the same insertion order.

only in ruby 1.9

C:\ruby1.9\bin>irb.bat

h={}
=> {}

h[1]=1
=> 1

h[2]=1
=> 1

h[3]=1
=> 1

h
=> {1=>1, 2=>1, 3=>1}

h[1.5]=1
=> 1

h
=> {1=>1, 2=>1, 3=>1, 1.5=>1}

h[0]=1
=> 1

h
=> {1=>1, 2=>1, 3=>1, 1.5=>1, 0=>1}

RUBY_VERSION
=> “1.9.0”


#3

Peña, Botp wrote:

From: Me Me [mailto:removed_email_address@domain.invalid]

I would like to know if it’s possible to insert values in a Hash and

then extract all of them in the same insertion order.

only in ruby 1.9

Could anybody explain why this feature was added? Isn’t it going to slow
down the operations on the Hash? I think it is useless to mix Array with
Hash.

TPR.


#4

Thomas B. wrote:

Peña, Botp wrote:

From: Me Me [mailto:removed_email_address@domain.invalid]

I would like to know if it’s possible to insert values in a Hash and

then extract all of them in the same insertion order.

only in ruby 1.9

Could anybody explain why this feature was added? Isn’t it going to slow
down the operations on the Hash? I think it is useless to mix Array with
Hash.

TPR.

well, basically I just need a Hash to use String as indexes like

sourceInfo = Hash.new
sourceInfo[“var1”]=123
sourceInfo[“var2”]=2
sourceInfo[“var3”]=3
sourceInfo[“var4”]=23

and then print them in the exact order of insertion


#5

Dick D. wrote:

On Thu, Sep 18, 2008 at 11:13 AM, Me Me removed_email_address@domain.invalid wrote:

Hash.

and then print them in the exact order of insertion

Can’t you just sort them on the way out?
i.e.

sourceinfo.keys.sort.each { |k|
sourceinfo[k].do_stuff
}

thanks
well in this case they will be sorted alphabetically, right?
the example is simplyfied, in my case the keys have differernt names.


#6

On Thu, Sep 18, 2008 at 11:13 AM, Me Me removed_email_address@domain.invalid wrote:

Hash.

and then print them in the exact order of insertion

Can’t you just sort them on the way out?
i.e.

sourceinfo.keys.sort.each { |k|
sourceinfo[k].do_stuff
}


#7

unknown wrote:

a = []
a.push 1
a.push 2
a.push 3
a.each {|e|puts e}

I think this is not my case, I need a Hash as I explained


#8

a = []
a.push 1
a.push 2
a.push 3
a.each {|e|puts e}


#9

Hi,

In message “Re: Preserve insert order in a Hash”
on Thu, 18 Sep 2008 18:54:20 +0900, “Thomas B.” removed_email_address@domain.invalid
writes:

|Could anybody explain why this feature was added?

Useful for some cases, especially for keyword arguments.

|Isn’t it going to slow down the operations on the Hash?

No. hash reference operation does not touch order information, only
for iteration. Memory consumption increased a bit.

          matz.

#10

well, basically I just need a Hash to use String as indexes like

sourceInfo = Hash.new
sourceInfo[“var1”]=123
sourceInfo[“var2”]=2
sourceInfo[“var3”]=3
sourceInfo[“var4”]=23

What do you want to happen if sourceInfo[“var2”] is assigned a second
time? Do you want to replace it where it originally was in the sequence,
or do you want to delete it and add the new value to the end? Or do you
want both elements to appear at the same time? Or doesn’t it matter?

I am just wondering because perhaps all you need is

sourceInfo = []
sourceInfo << [“var1”,123]
sourceInfo << [“var2”,2]
sourceInfo << [“var3”,3]
sourceInfo << [“var4”,4]
sourceInfo.each { |k,v| puts “#{k}=>#{v}” }

At least, that’s fine if all you want to do is iterate over the
collection and add new elements. Finding or deleting elements by key
requires a linear search:

dummy,value = sourceInfo.find { |k,v| k == “var2” }

However you can optimise this by building a hash as you go which points
to the same elements; or you could build an array containing just the
keys.

class HashAndArray
def initialize
@h, @a = {}, []
end
def
@h[k]
end
def []=(k,v)
@a << k unless @h.has_key?(k)
@h[k] = v
end
def each
@a.each { |k| yield k,@h[k] }
end
end

sourceInfo = HashAndArray.new
sourceInfo[“var1”]=123
sourceInfo[“var2”]=2
sourceInfo[“var3”]=3
sourceInfo[“var4”]=23
sourceInfo.each { |k,v| puts “#{k}=>#{v}” }
puts sourceInfo[“var2”]

Adding a delete() method is left as an exercise.


#11

From: “Me Me” removed_email_address@domain.invalid

I think this is not my case, I need a Hash as I explained

Here’s an implementation for ruby 1.8 … I apologize for the lack of
tests.

class InsertOrderPreservingHash
include Enumerable

def initialize(*args, &block)
@h = Hash.new(*args, &block)
@ordered_keys = []
end

def []=(key, val)
@ordered_keys << key unless @h.has_key? key
@h[key] = val
end

def each
@ordered_keys.each {|k| yield(k, @h[k])}
end
alias :each_pair :each

def each_value
@ordered_keys.each {|k| yield(@h[k])}
end

def each_key
@ordered_keys.each {|k| yield k}
end

def keys
@ordered_keys
end

def values
@ordered_keys.map {|k| @h[k]}
end

def clear
@ordered_keys.clear
@h.clear
end

def delete(k, &block)
@ordered_keys.delete k
@h.delete(k, &block)
end

def reject!
del = []
each_pair {|k,v| del << k if yield k,v}
del.each {|k| delete k}
del.empty? ? nil : self
end

def delete_if(&block)
reject!(&block)
self
end

%w(merge!).each do |name|
define_method(name) do |*args|
raise NotImplementedError, “#{name} not implemented”
end
end

def method_missing(*args)
@h.send(*args)
end
end

example:

h = InsertOrderPreservingHash.new

h[:aaa] = 0
h[:foo] = 123
h[:bar] = 456
h[:baz] = 789
h.delete :aaa
h[:aaa] = 1

h.each_pair {|k,v| p [k,v]}

produces:

[:foo, 123]
[:bar, 456]
[:baz, 789]
[:aaa, 1]

Regards,

Bill


#12

On Thu, Sep 18, 2008 at 6:28 AM, Yukihiro M.
removed_email_address@domain.invalidwrote:

|Isn’t it going to slow down the operations on the Hash?

No. hash reference operation does not touch order information, only
for iteration. Memory consumption increased a bit.

If I remember correctly, it’s been a while since I looked at the code,
1.9
implements this by using a singly linked list which introduces a small
overhead only when elements are either added or deleted.


Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/


#13

Hi,

Yukihiro M. wrote in [ruby-talk:315237]:
At Thu, 18 Sep 2008 19:28:28 +0900,

|Could anybody explain why this feature was added?

Useful for some cases, especially for keyword arguments.

And the performance of iterations improved a little, but
significantly.


#14

Hi,

In message “Re: Preserve insert order in a Hash”
on Thu, 18 Sep 2008 21:44:23 +0900, “Rick DeNatale”
removed_email_address@domain.invalid writes:

|> No. hash reference operation does not touch order information, only
|> for iteration. Memory consumption increased a bit.
|>
|
|If I remember correctly, it’s been a while since I looked at the code, 1.9
|implements this by using a singly linked list which introduces a small
|overhead only when elements are either added or deleted.

You’re right. Thank you for pointing out.

          matz.

#15

Bill K. wrote:

From: “Me Me” removed_email_address@domain.invalid

I think this is not my case, I need a Hash as I explained

Here’s an implementation for ruby 1.8 … I apologize for the lack of
tests.

Thanks a lot! That worked perfectly
Bye


#16

On Thu, Sep 18, 2008 at 8:42 AM, Brian C. removed_email_address@domain.invalid
wrote:

or do you want to delete it and add the new value to the end? Or do you
want both elements to appear at the same time? Or doesn’t it matter?

FWIW, Ruby 1.9 seems to keep the original insertion order, reassiging a
value to an existing key leaves the order unchanged:

$ irb1.9
irb(main):001:0> a = {a: 1, b: 2, c: 3}
=> {:a=>1, :b=>2, :c=>3}
irb(main):002:0> a.keys
=> [:a, :b, :c]
irb(main):003:0> a[:b] = 4
=> 4
irb(main):004:0> a.keys
=> [:a, :b, :c]
irb(main):005:0> a
=> {:a=>1, :b=>4, :c=>3}
irb(main):006:0>

At least my somewhat out of date 1.9 does

$ ruby1.9 -v
ruby 1.9.0 (2008-03-21 revision 0) [i686-darwin9.2.2]


Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/


#17

On Sep 18, 2008, at 3:42 AM, Me Me wrote:

gem install orderedhash

a @ http://codeforpeople.com/


#18

If scale is needed this has been implemented by Takuma Ozawa as a very
efficient RB tree which I used on a Mac:

sudo gem install rbtree
irb
require ‘rbtree’
a=RBTree.new
a[‘x’]=1
a[‘z’]=2
a[‘y’]=3
a
============> #RBTree:{‘x’=>1,‘y’=>3,‘z’=>2}


#19

On Sep 18, 2008, at 10:01 AM, jim finucane wrote:

a
cfp:~ > cat a.rb
require ‘rubygems’
require ‘rbtree’

rbtree is sorted by the keys ‘natural’ sort order

not the insertion order

rb = RBTree.new

rb[‘z’] = 3
rb[‘y’] = 2
rb[‘x’] = 1

rb.each do |key, value|
puts “#{ key } : #{ value }”
end

cfp:~ > ruby a.rb
x : 1
y : 2
z : 3

a @ http://codeforpeople.com/


#20

Peña, Botp wrote:

From: Me Me [mailto:removed_email_address@domain.invalid]

I would like to know if it’s possible to insert values in a Hash and

then extract all of them in the same insertion order.

only in ruby 1.9

Well, that is not entirely true.

h={}
p h
h[1]=1
p h
h[2]=1
p h
h[3]=1
p h
h
p h
h[1.5]=1
p h
h
p h
h[0]=1
p h
h
p h

yields:

{}
{1=>1}
{1=>1, 2=>1}
{1=>1, 2=>1, 3=>1}
{1=>1, 2=>1, 3=>1}
{1=>1, 2=>1, 3=>1, 1.5=>1}
{1=>1, 2=>1, 3=>1, 1.5=>1}
{1=>1, 2=>1, 3=>1, 1.5=>1, 0=>1}
{1=>1, 2=>1, 3=>1, 1.5=>1, 0=>1}

in JRUBY 1.1 which I am running on windows under netbeans.