Python Pandas - read_csv makes duplicates to NaN - python-2.7

Question

I am currently working with pandas on a Dataframe and after reading a csv file and converting a specific column into str, pandas seems to transform all duplicates of this row into NaNs.
bla = pd.read_csv(bla_path, sep=',',converters={'order_id':str})
and it gives me this results:
internal_conversion_id order_id conversion_target_id \
0 85 9222 67
1 20 9224 65
2 20 NaN 65
3 20 NaN 65
4 33 9233 67
5 33 NaN 67
Does anybody know what I'm missing? The original file does contain the duplicates.
EDIT: I just checked - this also happens, when i don't use converters.
EDIT 2: here some lines from the original csv:
internal_conversion_id,order_id,conversion_target_id,product_nr
85,9222,67,1
20,9224,65,1
20,9224,65,2
20,9224,65,3
33,9223,67,1
33,9223,67,2
EDIT3:
ok, i think i found the source.
At some point in the code I wanted to create a second variable with the same content as the first one, but without the duplicates. Pandas deletes all the duplicates in the first variable too. How can i stop Pandas from doing so?
here is the piece of code:
bla2 = bla
bla2['order_id'] = bla2['order_id'].drop_duplicates()
bla2 = bla2[pd.notnull(bla2['order_id'])]

Answer 1

if you wanted to just drop the duplicates you could've done it this way:
bla2 = bla2.drop_duplicates(subset='order_id')
what you did was to overwrite the column with the returned results from drop_duplicates for that column by doing this:
bla2['order_id'] = bla2['order_id'].drop_duplicates()
at that point you introduced NaN where the values were duplicates:
In [3]:
df['order_id'].drop_duplicates()
Out[3]:
0 9222
1 9224
4 9223
Name: order_id, dtype: int64
In [4]:
df['order_id'] = df['order_id'].drop_duplicates()
df
Out[4]:
internal_conversion_id order_id conversion_target_id product_nr
0 85 9222 67 1
1 20 9224 65 1
2 20 NaN 65 2
3 20 NaN 65 3
4 33 9223 67 1
5 33 NaN 67 2
However, your last line of code should've worked:
In [5]:
df = df[pd.notnull(df['order_id'])]
df
Out[5]:
internal_conversion_id order_id conversion_target_id product_nr
0 85 9222 67 1
1 20 9224 65 1
4 33 9223 67 1
So I don't know if you're getting confused
EDIT
if you want to make a copy then just make a copy:
bla2 = bla.copy()
then you can do whatever you want with bla2 and it won't affect bla
or you create bla2 from the result of bla.drop_duplicates:
bla2 = bla.drop_duplicates(subset='order_id')