Prev: The 'DataFrame' object | Next: Modifying Columns in DataFrame
Data access in DataFrame¶
- Accessing data in a DataFrame object can be done in many ways. Below we present the most important of them, that enable accessing data:
- 1) By column
- 2) By row (.ix, .iloc, .loc)
- 3) By scalar values (.at, .iat)
First make sure that you run the code below so that df is properly constructed for the following examples to run smoothly
In [1]:
import pandas as pd
df = pd.read_csv("../../data/sampledata.csv",sep=',',skiprows=3, header=0, index_col=0)
df
Out[1]:
1. Accessing data by column¶
- Selecting columns can be done in two ways:
- a) Using the column names as indexes
- b) Using the position (integer index) of the columns
a) Using column names¶
- Just provide the column name as index, the way you would do it in a dictionary
In [2]:
c = df['Country']
print(type(c))
c
Out[2]:
- You may provide the column label as a list, in which case you get a new DataFrame
In [3]:
c = df[['Country']]
print(type(c))
c
Out[3]:
- In case a column name is a single non-digit string, using the dot notation will work the same
In [4]:
df.Country
# but not this: df.2007
Out[4]:
- For more columns privide a list of column names (note the double brackets)
In [5]:
df[['Country', '2007', '2015']]
Out[5]:
b) Using column positions¶
- Provide a list of column position. Note the list (not simply an indeger index)
In [6]:
df[[1]]
Out[6]:
- For more columns privide a list of column positions (note the double brackets)
In [7]:
df[[1, 3, 5]]
Out[7]:
- List comprehensions can also be used to define a list of integer indices
In [9]:
df[[i for i in range(3)]]
Out[9]:
2. Accessing data by row¶
- Selecting rows can be done in three ways:
- a) Using the .loc indexer for selection by label (see pandas docs on .loc)
- b) Using the .iloc indexer for selection by position (integer-based indexer) (see pandas docs on .iloc)
- c) Using the .ix indexer that is a handy combination of the above (see pandas docs on .ix)
a) Using the .loc indexer for selection by label¶
- Provide a list with row labels
In [10]:
# Note that a row is now constructed as a new DataFrame
gr = df.loc[['GRC']]
print(type(gr))
gr
Out[10]:
- You can provide a single label as an single index (not in a list). In this case the row is now constructed as a new Series object
In [11]:
gr = df.loc['GRC']
print(type(gr))
gr
Out[11]:
- More rows: provide the list of labels
In [12]:
df.loc[['GRC','ESP','PRT']]
Out[12]:
- Slicing rows is also possible (note that single brackets are used)
In [13]:
df.loc['ESP':'GRC']
# You also get a new DataFrame
Out[13]:
b) Using the .iloc indexer for selection by position¶
- Provide the list of row positions
In [14]:
gr = df.iloc[[2]]
print(type(gr))
gr
Out[14]:
- You can provide a single label as an single index (not in a list). In this case the row is now constructed as a new Series object
In [15]:
esp = df.iloc[0]
print(type(esp))
esp
Out[15]:
- More rows: provide the list of integer position indexes
In [16]:
df.iloc[[2,4,6]]
Out[16]:
- Slicing rows is also possible (note that single brackets are used)
In [17]:
df.iloc[2:5]
Out[17]:
In [18]:
# .ix with row label
df.ix[['PRT']]
Out[18]:
In [19]:
# .ix with list of row positions
df.ix[[0,2,4]]
Out[19]:
In [20]:
# Slicing with labels
df.ix['ESP':'GRC']
Out[20]:
In [21]:
# Slicing with positions
df.ix[3:5]
Out[21]:
Accessing scalars (single data items)¶
- Selecting scalars can be done in two major ways:
- a) Using the [ ] notation to provide column & row indexes
- b) Using the .at and .iat indexers
Using the [ ] notation¶
In [22]:
ita = df['Country']['ITA']
ita
Out[22]:
In [24]:
# Also works without brackets for non-digit string indices
esp = df.Country.ESP
esp
Out[24]:
Using the .at indexer¶
- Provide first the row label and then the column name (remember: .at is a non-integer indexer)
In [25]:
d = df.at['FRA','Country']
d
Out[25]:
In [26]:
d = df.at['FRA','2015']
d
Out[26]:
Using the .iat indexer¶
In [27]:
d = df.iat[0,0]
d
Out[27]:
In [28]:
d = df.iat[7,9]
d
Out[28]:
In [29]:
x = 0
for i in range(len(df)):
for n in range(len(df.columns)):
if i==n:
print(df.iat[i,n])
- Obviously you can use .ix (also: .loc & .iloc) indexers along with column indexing to select single data
In [30]:
m = df.ix['MLT']['Country']
m
Out[30]:
In [31]:
m = df.ix['MLT'].Country
m
Out[31]:
In [32]:
m = df.ix['MLT']['2010']
m
Out[32]:
Copyright¶
. Free learning material
. See full copyright and disclaimer notice