concat over MultiIndex does not hierarchify further

[twiecki]

df2
Out[62]: 
          A   B      C       D       
level1a 0 foo one   -0.3364 -1.212780
        1 bar one    1.5044  0.832076
level1b 2 foo two    0.1210 -0.547690
        3 bar three -0.6163  0.948797
level1c 4 foo two    0.4783 -1.295089
        5 bar two   -0.6558 -1.515122
level1d 6 foo one   -1.1923 -0.884387
        7 foo three -0.6004  0.003952

pd.concat([df2[0:4], df2[4:8]], keys=['level0a', 'level0b'])
Out[66]: 
                  A   B      C       D       
level0a level1a 0 foo one   -0.3364 -1.212780
                1 bar one    1.5044  0.832076
level0a level1b 2 foo two    0.1210 -0.547690
                3 bar three -0.6163  0.948797
level0b level1c 4 foo two    0.4783 -1.295089
                5 bar two   -0.6558 -1.515122
level0b level1d 6 foo one   -1.1923 -0.884387
                7 foo three -0.6004  0.003952

While I would expect that there are only one of level0a and level0b at the first level.

[wesm]

The output is exactly what I would expect:

In [8]: concat([df2[0:4], df2[4:8]], keys=['level0a', 'level0b']).index
Out[8]: 
MultiIndex([('level0a', 'level1a', 0), ('level0a', 'level1a', 1),
       ('level0a', 'level1b', 2), ('level0a', 'level1b', 3),
       ('level0b', 'level1c', 4), ('level0b', 'level1c', 5),
       ('level0b', 'level1d', 6), ('level0b', 'level1d', 7)], dtype=object)

Could you explain more what you mean / what is the issue?

[twiecki]

I would expect the following output:

                  A   B      C       D       
level0a level1a 0 foo one   -0.3364 -1.212780
                1 bar one    1.5044  0.832076
        level1b 2 foo two    0.1210 -0.547690
                3 bar three -0.6163  0.948797
level0b level1c 4 foo two    0.4783 -1.295089
                5 bar two   -0.6558 -1.515122
        level1d 6 foo one   -1.1923 -0.884387
                7 foo three -0.6004  0.003952

The indexes are correct, I just wonder why there is redundancy when displaying the upmost level.

[wesm]

Ah. That is just how it's outputted. Initially I thought that at every "tick" outside the lowest level that displaying all the levels would aid readability. But maybe not. I don't have a strong opinion on the matter honestly

[twiecki]

For what it's worth, I'd personally prefer it the other way around. I
think it clutters up the display, you additionally have to parse where
the next item starts by comparing them. And, I think it would be more
consistent.

On Thu, Jan 26, 2012 at 5:06 PM, Wes McKinney
reply@reply.github.com
wrote:

Ah. That is just how it's outputted. Initially I thought that at every "tick" outside the lowest level that displaying all the levels would aid readability. But maybe not. I don't have a strong opinion on the matter honestly

---

Reply to this email directly or view it on GitHub:
https://github.com/wesm/pandas/issues/689#issuecomment-3677748

[twiecki]

Thanks for making the change, looks much better now I think.