In Pandas, does .iloc method give a copy or view?

Question

I find the result is a little bit random. Sometimes it's a copy sometimes it's a view. For example:

df = pd.DataFrame([{'name':'Marry', 'age':21},{'name':'John','age':24}],index=['student1','student2'])

df
              age   name
   student1   21  Marry
   student2   24   John

Now, Let me try to modify it a little bit.

df2= df.loc['student1']
df2 [0] = 23
df
              age   name
   student1   21  Marry
   student2   24   John

As you can see, nothing changed. df2 is a copy. However, if I add another student into the dataframe...

df.loc['student3'] = ['old','Tom']
df
               age   name
    student1   21  Marry
    student2   24   John
    student3  old    Tom

Try to change the age again..

df3=df.loc['student1']
df3[0]=33
df
               age   name
    student1   33  Marry
    student2   24   John
    student3  old    Tom

Now df3 suddenly became a view. What is going on? I guess the value 'old' is the key?

score 29 · Answer 1 · answered Dec 25 '17 at 23:32

You are starting with a DataFrame that has two columns with two different dtypes:

df.dtypes
Out: 
age      int64
name    object
dtype: object

Since different dtypes are stored in different numpy arrays under the hood, you have two different blocks for them:

df.blocks

Out: 
{'int64':           age
 student1   21
 student2   24, 'object':            name
 student1  Marry
 student2   John}

If you attempt to slice the first row of this DataFrame, it has to get one value from each different block which makes it necessary to create a copy.

df2.is_copy
Out[40]: <weakref at 0x7fc4487a9228; to 'DataFrame' at 0x7fc4488f9dd8>

In the second attempt, you are changing the dtypes. Since 'old' cannot be stored in an integer array, it casts the Series as an object Series.

df.loc['student3'] = ['old','Tom']

df.dtypes
Out: 
age     object
name    object
dtype: object

Now all data for this DataFrame is stored in a single block (and in a single numpy array):

df.blocks

Out: 
{'object':           age   name
 student1   21  Marry
 student2   24   John
 student3  old    Tom}

At this step, slicing the first row can be done on the numpy array without creating a copy, so it returns a view.

df3._is_view
Out: True

`_is_view` is a really good functionality, i didn't know about this. Thanks — Puneet Tripathi, Dec 26 '17 at 07:07

juanpa.arrivillaga · Accepted Answer · 2017-12-25T23:42:25.517

In general, you can get a view if the data-frame has a single dtype, which is not the case with your original data-frame:

In [4]: df
Out[4]:
          age   name
student1   21  Marry
student2   24   John

In [5]: df.dtypes
Out[5]:
age      int64
name    object
dtype: object

However, when you do:

In [6]: df.loc['student3'] = ['old','Tom']
   ...:

The first column get's coerced to object, since columns cannot have mixed dtypes:

In [7]: df.dtypes
Out[7]:
age     object
name    object
dtype: object

In this case, the underlying .values will always return an array with the same underlying buffer, and changes to that array will be reflected in the data-frame:

In [11]: vals = df.values

In [12]: vals
Out[12]:
array([[21, 'Marry'],
       [24, 'John'],
       ['old', 'Tom']], dtype=object)

In [13]: vals[0,0] = 'foo'

In [14]: vals
Out[14]:
array([['foo', 'Marry'],
       [24, 'John'],
       ['old', 'Tom']], dtype=object)

In [15]: df
Out[15]:
          age   name
student1  foo  Marry
student2   24   John
student3  old    Tom

On the other hand, with mixed types like with your original data-frame:

In [26]: df = pd.DataFrame([{'name':'Marry', 'age':21},{'name':'John','age':24}]
    ...: ,index=['student1','student2'])
    ...:

In [27]: vals = df.values

In [28]: vals
Out[28]:
array([[21, 'Marry'],
       [24, 'John']], dtype=object)

In [29]: vals[0,0] = 'foo'

In [30]: vals
Out[30]:
array([['foo', 'Marry'],
       [24, 'John']], dtype=object)

In [31]: df
Out[31]:
          age   name
student1   21  Marry
student2   24   John

Note, however, that a view will only be returned if it is possible to be a view, i.e. if it is a proper slice, otherwise, a copy will be made regardless of the dtypes:

In [39]: df.loc['student3'] = ['old','Tom']


In [40]: df2
Out[40]:
          name
student3   Tom
student2  John

In [41]: df2.loc[:] = 'foo'

In [42]: df2
Out[42]:
         name
student3  foo
student2  foo

In [43]: df
Out[43]:
          age   name
student1   21  Marry
student2   24   John
student3  old    Tom

So does that mean I will always get a copy if the data frame has multiple dtypes? — Qiyu, Dec 26 '17 at 01:28

In Pandas, does .iloc method give a copy or view?

2 Answers2

Linked