joining on index

[genya0407]

I request some means to join two Daru::DataFrame on their index. Like: pandas's one.

[athityakumar]

If I've understood properly,

Input

2.3.1 :006 > l1 = Daru::DataFrame.new({ a: [1,2,3], b: [4,5,6]}, index: ['x','y','z'])
 => #<Daru::DataFrame(3x2)>
       a   b
   x   1   4
   y   2   5
   z   3   6 
2.3.1 :007 > l2 = Daru::DataFrame.new({ c: [7,8,9], d: [10,11,12]}, index: ['x','y','z'])
 => #<Daru::DataFrame(3x2)>
       c   d
   x   7  10
   z   8  11
   y   9  12 

Then, desired output is something like

 => #<Daru::DataFrame(3x4)>
       a   b   c   d
   x   1   4   7   10
   y   2   5   8   11
   z   3   6   9   12 

Right? I'd like to work on this. :smile:

[athityakumar]

I had a look at the core/merge.rb file, however I'm not able to make out how to check the join function on console (irb). Any help would be appreciated.

[v0dro]

You can start a session with rake pry and then call join on your DataFrame like df.join(....).

[Shekharrajak]

If no one is working on it then I am interested to try to work on it.

[v0dro]

@Shekharrajak yes you can.

[Shekharrajak]

I think join must work on the index as default, right ?

[v0dro]

@gnilrets WDYT?

[gnilrets]

Makes sense to me...

[Shekharrajak]

It took me time to understand the core/merge.rb, since there is no examples/docs added in file.

1. I see that merge is defined in dataframe.rb with no option. But as we know merge + some condition is called different types of join.

2. merge must be defined in core/merge.rb and same kind of(like join), option must be used in merge as well.

3. One special case will be when there is some column(s) is(are) common on df1 and df2 then suffix must be added. (at this time we will first add suffix on both df column then pass into join).

E.g.

irb(main):065:0> left
=> #<Daru::DataFrame(3x2)>
       A   B
   0  A0  B0
   1  A1  B1
   2  A2  B2
irb(main):066:0> right
=> #<Daru::DataFrame(3x2)>
       A   c
  K0  A0  D0
  K2  A1  D2
  K3  A3  D3

irb(main):064:0> left.merge(right)
=> #<Daru::DataFrame(3x4)>
     A_1   B A_2   c
   0  A0  B0  A0  D0
   1  A1  B1  A1  D2
   2  A2  B2  A3  D3

Problem in merge :

• Doesn't use index to merge on.

see this example :

irb(main):002:0* left = Daru::DataFrame.new({
irb(main):003:2* :A => ['A0', 'A1', 'A2'],
irb(main):004:2* :B => ['B0', 'B1', 'B2']},
irb(main):005:1* index: ['K0', 'K1', 'K2'])
=> #<Daru::DataFrame(3x2)>
       A   B
  K0  A0  B0
  K1  A1  B1
  K2  A2  B2
irb(main):006:0>
irb(main):007:0* right = Daru::DataFrame.new({
irb(main):008:2* :C => ['C0', 'C2', 'C3'],
irb(main):009:2* :D => ['D0', 'D2', 'D3']},
irb(main):010:1* index: ['K0', 'K2', 'K3'])
=> #<Daru::DataFrame(3x2)>
       C   D
  K0  C0  D0
  K2  C2  D2
  K3  C3  D3

# index values are different but still merge

irb(main):011:0> left.merge(right)
=> #<Daru::DataFrame(3x4)>
       A   B   C   D
   0  A0  B0  C0  D0
   1  A1  B1  C2  D2
   2  A2  B2  C3  D3

# when no index is passed then its fine.

irb(main):012:0> left = Daru::DataFrame.new({
irb(main):013:2* :A => ['A0', 'A1', 'A2'],
irb(main):014:2* :B => ['B0', 'B1', 'B2']},
irb(main):015:1* )
=> #<Daru::DataFrame(3x2)>
       A   B
   0  A0  B0
   1  A1  B1
   2  A2  B2
irb(main):016:0> right = Daru::DataFrame.new({
irb(main):017:2* :C => ['C0', 'C2', 'C3'],
irb(main):018:2* :D => ['D0', 'D2', 'D3']},
irb(main):019:1* )
=> #<Daru::DataFrame(3x2)>
       C   D
   0  C0  D0
   1  C2  D2
   2  C3  D3
irb(main):020:0> left.merge(right)
=> #<Daru::DataFrame(3x4)>
       A   B   C   D
   0  A0  B0  C0  D0
   1  A1  B1  C2  D2
   2  A2  B2  C3  D3

Reference :

1. http://stackoverflow.com/questions/38549/what-is-the-difference-between-inner-join-and-outer-join

2. http://www.shanelynn.ie/merge-join-dataframes-python-pandas-index-1/#mergetypes

[v0dro]

If I understand you correctly, you are basically proposing that parts of join should be included into merge right? Why is that? You can very well do the same thing by specifying options to join.

The purpose of this issue is to come up with suitable APIs to join two dataframes on index.

Also, how is your first example (in the section below Problem with merge) any different from the first? We all know that joining on index is not yet supported. Your example is perfectly demonstrating that.

And why do you want to perform a merge in the first place when the problem is with join?

[Shekharrajak]

Actually my above comment is not much related to this issue, but I think that if 'joining on index' is fixed then using inner join (default join for merge) merge can be done, default on: index. So we don't need to solve separate merge issue .

[v0dro]

Why should the default be changed from :inner to :index? Is there any compelling reason?

So we don't need to solve separate merge issue .

What exactly do you mean by this?