engine='python' df.to_csv() duplicate columns doesn't work

[y-p]

example from 0.10.1, and carried on to to_csv(engine='python') in 0.11

note the first columns duplicates the second columns, which shouldn't be.

Corrolary #3454

from pandas.util.testing import makeCustomDataframe as mkdf

N=10
df= mkdf(N, 3)
df.columns = ['a','a','b']
path = "/tmp/k.csv"
df.to_csv("/tmp/k.csv")
!cat "/tmp/k.csv"
R0,a,a,b
R_l0_g0,R0C1,R0C1,R0C2
R_l0_g1,R1C1,R1C1,R1C2
R_l0_g10,R2C1,R2C1,R2C2
R_l0_g2,R3C1,R3C1,R3C2
R_l0_g3,R4C1,R4C1,R4C2
R_l0_g4,R5C1,R5C1,R5C2
R_l0_g5,R6C1,R6C1,R6C2
R_l0_g6,R7C1,R7C1,R7C2
R_l0_g7,R8C1,R8C1,R8C2
R_l0_g8,R9C1,R9C1,R9C2

[jreback]

added #3459, should close this one and #3455

In [5]: df.to_csv('/tmp/dups')

In [6]: !cat "/tmp/dups"
R0,a,a,b
R_l0_g0,R0C0,R0C1,R0C2
R_l0_g1,R1C0,R1C1,R1C2
R_l0_g2,R2C0,R2C1,R2C2
R_l0_g3,R3C0,R3C1,R3C2
R_l0_g4,R4C0,R4C1,R4C2
R_l0_g5,R5C0,R5C1,R5C2
R_l0_g6,R6C0,R6C1,R6C2
R_l0_g7,R7C0,R7C1,R7C2
R_l0_g8,R8C0,R8C1,R8C2
R_l0_g9,R9C0,R9C1,R9C2

In [8]: pd.read_csv('/tmp/dups',index_col=0)
Out[8]: 
            a   a.1     b
R0                       
R_l0_g0  R0C0  R0C1  R0C2
R_l0_g1  R1C0  R1C1  R1C2
R_l0_g2  R2C0  R2C1  R2C2
R_l0_g3  R3C0  R3C1  R3C2
R_l0_g4  R4C0  R4C1  R4C2
R_l0_g5  R5C0  R5C1  R5C2
R_l0_g6  R6C0  R6C1  R6C2
R_l0_g7  R7C0  R7C1  R7C2
R_l0_g8  R8C0  R8C1  R8C2
R_l0_g9  R9C0  R9C1  R9C2

[jreback]

fyi, I am not sure if there is a way to make read_csv not create a new named column 'a.1', if you specify names=[], it doesn't work....is this a bug?

[y-p]

It may not be clean, but I think it's by design, predates dupe column support I should think,
a quick way to allow data with dupe columns from an external source to be imported into pandas.

not a big deal IMO.

[jreback]

closed by #3459

[y-p]

I believe not so. forgot the engine='python' bit, leave it raising a warning exception?

[jreback]

ahh...I see about engine=python, why don't just raise on that, seeing as we fixed the regular engine

[y-p]

Fine by me. #3458 does that