Get Pandas Snippets and Cheats Collection eBook

Select Data With .loc & .iloc

The common ways of selecting rows and columns in a pandas DataFrame are the functions .loc (label based selection, which means that one needs to specify the name of the rows and columns) and .iloc (index based selection, which means that one needs to specify rows and columns by their integer index).

.loc selection by index/labels and columns

df.loc[<row selection>, <column selection>]

.loc select all rows and some columns by index/labels

df.loc[:, ['column1', 'column2', 'column3']]

.loc select some rows and all columns by index/labels

df.loc[['row1', 'row2'], :]
# or ommiting the colon for columns labels
df.loc[['row1', 'row2']]

.loc select some rows and some columns by index/labels

df.loc[['row1', 'row2'], ['column1', 'column2', 'column3']]

Select rows whose column values equal some value/scalar

df.loc[df['column_x'] == some_value]

Select rows whose column values do not equal some_value/scalar

df.loc[df['column_x'] != some_value]

Select rows whose column values are in a list

df.loc[df['column_x'].isin(['A', 'B', 'C'])]

Select rows whose column values are not in a list

df.loc[~df['column_x'].isin(['A', 'B', 'C'])]

Select rows that satisfy multiple and (&) boolean conditions

# note the parentheses due to python operators precedence!
df.loc[(df['column_x'] >= A) & (df['column_y'] <= B)]

Select rows that satisfy multiple or (|) boolean conditions

# note the parentheses due to python operators precedence!
df.loc[(df['column_x'] >= A) | (df['column_y'] <= B)]

Update column values based on rows condition

df.loc[df['column_x'] >= A, 'column_y'] = new_value

Update multiple columns values based on rows condition

df.loc[df['column_x'] >= A, ['column_y', 'column_z']] = [new_value_y, new_value_z]

.loc selection for single label, all columns / returns a Series

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc['a']

>>> df.loc['a']
Name    Alex
Age       10
Name: a, dtype: object

.loc selection for list of label, all columns / returns a DataFrame

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc[['a', 'b']]

>>> df.loc[['a', 'b']]
   Name  Age
a  Alex   10
b   Bob   12

.loc selection for single label, single column / returns one value

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc['c', 'Age']

>>> df.loc['c', 'Age']
13

.loc selection for all label, single column / returns a Series

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc[:, 'Age']

>>> df.loc[:, 'Age']
a    10
b    12
c    13
Name: Age, dtype: int64

.loc selection for multiple labels, multiple columns / returns a DataFrame

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc[['a', 'c'], ['Age', 'Name']]

>>> df.loc[['a', 'c'], ['Age', 'Name']]
   Age    Name
a   10    Alex
c   13  Clarke

.loc selection for slice of labels for row, all columns / returns a DataFrame

# giving DataFrame df
>>> df
     Name  Age
a    Alex   10
b     Bob   12
c  Clarke   13

df.loc['b':'c']

# or

df.loc['b':'c', ['Name', 'Age']]

>>> df.loc['b':'c']
     Name  Age
b     Bob   12
c  Clarke   13

.loc selection for all rows, slice of columns / returns a DataFrame

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

df.loc[:, 'Age':'Occupation']

>>> df.loc[:, 'Age':'Occupation']
   Age Occupation
a   10    Student
b   12    Teacher
c   13   Engineer

.loc selection for rows and columns with boolean conditions

# length of boolean arrays must match dataframe size
# select rows 1 and 3 and first column
df.loc[[True, False, True], [True]]

Select rows 1 and 3, and second column with boolean conditions

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

df.loc[[True, False, True], [False, True]]

>>> df.loc[[True, False, True], [False, True]]
   Age
a   10
c   13
>>> df

Select rows whose column values equals a scalar

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

df.loc[df['Name'] == 'Alex']

>>> df.loc[df['Name'] == 'Alex']
   Name  Age Occupation
a  Alex   10    Student

# df['Name'] == 'Alex' produces a Pandas Series with a True/False value for every row in the data
# where there are 'True' values for the rows where the Name is 'Alex'.

>>> df['Name'] == 'Alex'
a     True
b    False
c    False
Name: Name, dtype: bool

Select rows that satisfy multiple boolean conditions; keep selected columns

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

df.loc[(df['Name'] == 'Alex') | (df['Age'] == 12), ['Name', 'Occupation']]

>>> df.loc[(df['Name'] == 'Alex') | (df['Age'] == 12), ['Name', 'Occupation']]
   Name Occupation
a  Alex    Student
b   Bob    Teacher

Update column values based on rows condition

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

# update Alex Occupation to Grad Student
df.loc[(df['Name'] == 'Alex'), ['Occupation']] = 'Grad Student'

>>> df
     Name  Age    Occupation
a    Alex   10  Grad Student
b     Bob   12       Teacher
c  Clarke   13      Engineer

Update multiple columns values based on rows condition

# giving DataFrame df
>>> df
     Name  Age Occupation
a    Alex   10   Student
b     Bob   12   Teacher
c  Clarke   13  Engineer

# update Alex Age and Occupation to 20 years and Grad Student
df.loc[(df['Name'] == 'Alex'), ['Age', 'Occupation']] = [20, 'Grad Student']

>>> df
     Name  Age    Occupation
a    Alex   20  Grad Student
b     Bob   12       Teacher
c  Clarke   13      Engineer

.iloc position based selection (from 0 to length-1 of the axis)

df.iloc[<row selection>, <column selection>]

.iloc single row selections, all columns

# return Series object
df.iloc[0]  # first row of dataframe
df.iloc[1]  # second row of dataframe
df.iloc[-1] # last row of data frame

.iloc single column selections, all rows

# return Series object
df.iloc[:, 0]  # first column of dataframe
df.iloc[:, 1]  # second column of dataframe
df.iloc[:, -1] # last column of data frame

.iloc multiple row and column selections

# return DataFrame object
df.iloc[0:3]             # first three rows of dataframe, all columns
df.iloc[:, 0:3]          # first three columns of dataframe, all rows
df.iloc[0:5, 4:7]        # first five rows, column 4th, 5th and 6th of dataframe
df.iloc [[3, 4], [1, 2]] # retrieve rows 4th & 5th, columns 2nd & 3rd with lists

.iloc selection for rows and columns with boolean conditions

# length of boolean arrays must match dataframe size
# select rows 1 and 3 and first column
df.iloc[[True, False, True], [True]]

Select the rows whose index label is an even number

df.iloc[lambda x: x.index % 2 == 0]

Select slice of rows with .iloc

# start_pos is the index of the first row to retrieve
# stop_pos is the index of the row to stop (last row retrieved will have the index of `stop_pos -1`)
df.iloc[start_pos:stop_pos]

Update/set column values based on rows condition with .iloc

# set column 4 values to `new_value` for first two rows of data frame
df.iloc[0:2, 3] = 'new_value'

# set all column values to `new_value` for first two rows of data frame
df.iloc[0:2] = 'new_value'