Function to select only certain key/value pairs from hash?

ruby-talk
1

Whenever Im coding I usually come across having to create a new hash
from the key/value pairs of another hash.

For example,

hash = {:a => 1, :b => 2, :c => 3}

I want to do something like this

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

Is there something like this already? Perhaps in the facets library? I
looked and couldn't find anything.

If anybody thinks that I shouldnt even be getting into this situation in
the first place, please let me know that as well.

···

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

2

Hmm... I thought there was a built in method for that. Here's one
way:

hash = {:a => 1, :b => 2, :c => 3}
Hash[*hash.select {|k,v| [:a,:b].include?(k)}.flatten] #=>
{:b=>2, :a=>1}

HTH,
Chris

···

On Nov 17, 3:50 pm, Aryk Grosz <tennisbum2...@hotmail.com> wrote:

Whenever Im coding I usually come across having to create a new hash
from the key/value pairs of another hash.

For example,

hash = {:a => 1, :b => 2, :c => 3}

I want to do something like this

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

Is there something like this already? Perhaps in the facets library? I
looked and couldn't find anything.

If anybody thinks that I shouldnt even be getting into this situation in
the first place, please let me know that as well.
--
Posted viahttp://www.ruby-forum.com/.

3

I tried to resist making this a one-liner and went for the clean and explicit way :slight_smile:

class Hash
     def from_keys(*keys)
         selected_values = self.values_at(*keys)
         selected_key_values = keys.zip(selected_values)
         Hash[*selected_key_values.flatten]
     end
end

hash = {:a => 1, :b => 2, :c => 3}
p hash.from_keys(:a, :b)

=> {:a=>1, :b=>2}

···

On 17 nov. 08, at 23:50, Aryk Grosz wrote:

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

--
Luc Heinrich - luc@honk-honk.com

4

Aryk Grosz wrote:

I want to do something like this

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

In ruby1.9, Hash#select returns another Hash. But you'd still be
iterating the 'wrong way' (that is, iterating through the hash and doing
a linear search through the keys)

Personally I'd go with:

class Hash
  def from_keys(*keys)
    keys.inject({}) { |h,k| h[k] = self[k] if has_key?(k); h }
  end
end

hash = {:a => 1, :b => 2, :c => 3}
p hash.from_keys(:a, :b)

With ruby19 you can do:

    keys.each_with_object({}) { |k,h| h[k] = self[k] if has_key?(k) }

which is more keystrokes but maybe the teeniest bit more efficient. But
I hate each_with_object on the principle that its arguments are the
opposite way round to inject :frowning:

···

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

5

Yeah, or the simpler: hash.reject {|k,v| ![:a,:b].include?(k)}

···

On Nov 17, 4:46 pm, Chris Shea <cms...@gmail.com> wrote:

On Nov 17, 3:50 pm, Aryk Grosz <tennisbum2...@hotmail.com> wrote:

> Whenever Im coding I usually come across having to create a new hash
> from the key/value pairs of another hash.

> For example,

> hash = {:a => 1, :b => 2, :c => 3}

> I want to do something like this

> hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

> Is there something like this already? Perhaps in the facets library? I
> looked and couldn't find anything.

> If anybody thinks that I shouldnt even be getting into this situation in
> the first place, please let me know that as well.
> --
> Posted viahttp://www.ruby-forum.com/.

Hmm... I thought there was a built in method for that. Here's one
way:

hash = {:a => 1, :b => 2, :c => 3}
Hash[*hash.select {|k,v| [:a,:b].include?(k)}.flatten] #=>
{:b=>2, :a=>1}

HTH,
Chris

6

That is ridiculously inefficient and weird :slight_smile:

hash = {:a => 1, :b => 2, :c => 3}

selected = [:a, :b].map{|key|{key, hash[key]}}

p selected

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

Einar Magnús Boson
+354-661 1649
einarmagnus@tistron.se
einar.boson@gmail.com

···

On 17.11.2008, at 23:46 , Chris Shea wrote:

On Nov 17, 3:50 pm, Aryk Grosz <tennisbum2...@hotmail.com> wrote:

Whenever Im coding I usually come across having to create a new hash
from the key/value pairs of another hash.

For example,

hash = {:a => 1, :b => 2, :c => 3}

I want to do something like this

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

Is there something like this already? Perhaps in the facets library? I
looked and couldn't find anything.

If anybody thinks that I shouldnt even be getting into this situation in
the first place, please let me know that as well.
--
Posted viahttp://www.ruby-forum.com/.

Hmm... I thought there was a built in method for that. Here's one
way:

hash = {:a => 1, :b => 2, :c => 3}
Hash[*hash.select {|k,v| [:a,:b].include?(k)}.flatten] #=>
{:b=>2, :a=>1}

HTH,
Chris

7

require 'facets/hash/slice'

T.

···

On Nov 18, 3:31 am, Luc Heinrich <l...@honk-honk.com> wrote:

On 17 nov. 08, at 23:50, Aryk Grosz wrote:

> hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

I tried to resist making this a one-liner and went for the clean and
explicit way :slight_smile:

class Hash
def from_keys(*keys)
selected_values = self.values_at(*keys)
selected_key_values = keys.zip(selected_values)
Hash[*selected_key_values.flatten]
end
end

hash = {:a => 1, :b => 2, :c => 3}
p hash.from_keys(:a, :b)

=> {:a=>1, :b=>2}

8

Chris Shea wrote:

> Posted viahttp://www.ruby-forum.com/.

Hmm... I thought there was a built in method for that. �Here's one
way:

hash = {:a => 1, :b => 2, :c => 3}
Hash[*hash.select {|k,v| [:a,:b].include?(k)}.flatten] �#=>
{:b=>2, :a=>1}

HTH,
Chris

Yeah, or the simpler: hash.reject {|k,v| ![:a,:b].include?(k)}

And indeed, it's in facets:

require 'facets'
p hash.slice(:a,:b)

# However, this breaks on

hash.slice(:a,:d)

# and the hash.reject does not

hth,

Siep

Siep

···

On Nov 17, 4:46�pm, Chris Shea <cms...@gmail.com> wrote:

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

9

and now I see that I was stupd.
heh. nevermind <.<

Einar Magnús Boson
+354-661 1649
einarmagnus@tistron.se
einar.boson@gmail.com

···

On 18.11.2008, at 01:34 , Einar Magnús Boson wrote:

On 17.11.2008, at 23:46 , Chris Shea wrote:

On Nov 17, 3:50 pm, Aryk Grosz <tennisbum2...@hotmail.com> wrote:

Whenever Im coding I usually come across having to create a new hash
from the key/value pairs of another hash.

For example,

hash = {:a => 1, :b => 2, :c => 3}

I want to do something like this

hash.from_keys(:a,:b) => {:a=> 1, :b=> 2}

Is there something like this already? Perhaps in the facets library? I
looked and couldn't find anything.

If anybody thinks that I shouldnt even be getting into this situation in
the first place, please let me know that as well.
--
Posted viahttp://www.ruby-forum.com/.

Hmm... I thought there was a built in method for that. Here's one
way:

hash = {:a => 1, :b => 2, :c => 3}
Hash[*hash.select {|k,v| [:a,:b].include?(k)}.flatten] #=>
{:b=>2, :a=>1}

HTH,
Chris

That is ridiculously inefficient and weird :slight_smile:

hash = {:a => 1, :b => 2, :c => 3}

selected = [:a, :b].map{|key|{key, hash[key]}}

p selected

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

Einar Magnús Boson
+354-661 1649
einarmagnus@tistron.se
einar.boson@gmail.com

10

That is ridiculously inefficient and weird :slight_smile:

hash = {:a => 1, :b => 2, :c => 3}

selected = [:a, :b].map{|key|{key, hash[key]}}

p selected

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

and now I see that I was stupd.
heh. nevermind <.<

This is what I meant to do, a lot more efficient to look up the values you're lookin for than looping through all keys for every element.

hash = {:a => 1,
      :b => 2,
      :c => 3,
      :d => 4,
      :str => "test"}

selected = [:a, :d, :str].inject({}){|result, key| result[key]= hash[key];result}

p selected

# >> {:str=>"test", :a=>1, :d=>4}

.

11

You will get nils for none existing keys that way.

selected = [:whatever].inject({}){|result, key| = hash[key]; result}
#=> {:whatever => nil}

If that's what you want, great.

Todd

···

On Mon, Nov 17, 2008 at 7:44 PM, Einar Magnús Boson <einarmagnus@tistron.se> wrote:

This is what I meant to do, a lot more efficient to look up the values
you're lookin for than looping through all keys for every element.

hash = {:a => 1,
                       :b => 2,
                       :c => 3,
                       :d => 4,
                       :str => "test"}

selected = [:a, :d, :str].inject({}){|result, key| result[key]=
hash[key];result}

p selected

# >> {:str=>"test", :a=>1, :d=>4}

12

This is what I meant to do, a lot more efficient to look up the values
you're lookin for than looping through all keys for every element.

hash = {:a => 1,
                      :b => 2,
                      :c => 3,
                      :d => 4,
                      :str => "test"}

selected = [:a, :d, :str].inject({}){|result, key| result[key]=
hash[key];result}

p selected

# >> {:str=>"test", :a=>1, :d=>4}

You will get nils for none existing keys that way.

selected = [:whatever].inject({}){|result, key| = hash[key]; result}
#=> {:whatever => nil}

If that's what you want, great.

Todd

if that is a problem it's easy to fix

hash = {:a => 1,
         :b => 2,
         :c => 3,
         :d => 4,
         :str => "test"}

find = [:a, :d, :str, :extra]

selected = find.inject({}) {
      >result, key|
      val=hash[key]
      result[key]=val if val
      result }

p selected
# >> {:str=>"test", :a=>1, :d=>4}

given: f, h = find.size, hash.size
This method is O(f) because hash lookup is O(1)
The other way it's O(f*h).
so if the hash is big it should make a difference.

If the elements always are just a few it doesn't really matter.

einarmagnus

···

On 18.11.2008, at 05:00 , Todd Benson wrote:

On Mon, Nov 17, 2008 at 7:44 PM, Einar Magnús Boson > <einarmagnus@tistron.se> wrote:

13

I'm not apologizing for the missing result[key] on the lhs, though I
know it's pretty important to many to have working code in a message.
Anyone with a half a brain knows what I meant, but just to be clean...

h = Hash[:a, 1, :b, 2]
p [:missing_key].inject({}) {|result, key| result[key] = h[key]; result}
=> {:missing_key => nil}

In any case, reducing a hash (being selective) I think is best done
using a reject with a negative in the block (as per Chris Shea's
second example).

Todd

···

On Mon, Nov 17, 2008 at 11:04 PM, Todd Benson <caduceass@gmail.com> wrote:

On Mon, Nov 17, 2008 at 7:44 PM, Einar Magnús Boson > <einarmagnus@tistron.se> wrote:

This is what I meant to do, a lot more efficient to look up the values
you're lookin for than looping through all keys for every element.

hash = {:a => 1,
                       :b => 2,
                       :c => 3,
                       :d => 4,
                       :str => "test"}

selected = [:a, :d, :str].inject({}){|result, key| result[key]=
hash[key];result}

p selected

# >> {:str=>"test", :a=>1, :d=>4}

You will get nils for none existing keys that way.

selected = [:whatever].inject({}){|result, key| = hash[key]; result}
#=> {:whatever => nil}

14

This is what I meant to do, a lot more efficient to look up the values
you're lookin for than looping through all keys for every element.

hash = {:a => 1,
                     :b => 2,
                     :c => 3,
                     :d => 4,
                     :str => "test"}

selected = [:a, :d, :str].inject({}){|result, key| result[key]=
hash[key];result}

p selected

# >> {:str=>"test", :a=>1, :d=>4}

You will get nils for none existing keys that way.

selected = [:whatever].inject({}){|result, key| = hash[key]; result}
#=> {:whatever => nil}

If that's what you want, great.

Todd

if that is a problem it's easy to fix

hash = {:a => 1,
       :b => 2,
       :c => 3,
       :d => 4,
       :str => "test"}

find = [:a, :d, :str, :extra]

selected = find.inject({}) {
                       >result, key|
                       val=hash[key]
                       result[key]=val if val
                       result }

p selected
# >> {:str=>"test", :a=>1, :d=>4}

given: f, h = find.size, hash.size
This method is O(f) because hash lookup is O(1)
The other way it's O(f*h).

Good point. The use of #inject, though, may cloud that performance analysis.

so if the hash is big it should make a difference.

If the elements always are just a few it doesn't really matter.

I'm an #inject sort of guy so I like the way you approach this. But,
there must be a reason why you don't prefer the negative #reject that
seems to work for most people.

I might benchmark this, but I think object creation and destruction
might outweigh the O(f*h).

Todd

···

On Mon, Nov 17, 2008 at 11:23 PM, Einar Magnús Boson <einarmagnus@tistron.se> wrote:

On 18.11.2008, at 05:00 , Todd Benson wrote:

On Mon, Nov 17, 2008 at 7:44 PM, Einar Magnús Boson >> <einarmagnus@tistron.se> wrote: