read csv files pandas slice arrays

Firstly DataFrame.from_csv whilst still supported, it's better to use the top level read_csv instead as this supports more functionality.

So this:

a = g.iloc[2:4,1]

is wrong syntax, you want:

a = g.iloc[2:4]['A']

Secondly, by default DataFrame.from_csv uses the first column as the index which is why column 'A' is your index, if you passed index_col=None then you get the desired result:

In [6]:
pd.DataFrame.from_csv(file_path)

Out[6]:
    B
A    
1  56
2  76
3  23
4  45
5  54
6  65
7  22
In [7]:    
pd.DataFrame.from_csv(file_path, index_col=None)

Out[7]:
   A   B
0  1  56
1  2  76
2  3  23
3  4  45
4  5  54
5  6  65
6  7  22

Correct syntax:

In [9]:   
df.iloc[2:4]['A']

Out[9]:
2    3
3    4
Name: A, dtype: int64

Additionally read_csv the default for index_col is None so your problem with the alignment would not have happened if you had used read_csv.

Please check the docs on indexing and selecting.

EDIT

As @Jeff suggested and I always agree with Jeff, for this kind of selection ix is the typical selection method but it's behaviour differs from iloc in that it does include the end row selection unlike iloc:

In [10]:    
df.ix[2:4,'A']

Out[10]:
2    3
3    4
4    5
Name: A, dtype: int64

So I don't know what you wanted row selection-wise but be aware of the different semantics.

Update

Note that .ix will be deprecated in the future, you can achieve the same result using .loc:

In [202]:
df.loc[2:4,'A']

Out[202]:
2    3
3    4
4    5
Name: A, dtype: int64