Indexing , Slicing & Subsetting DataFram in Python

1. Introduction

The Python and NumPy indexing operators [ ] and attribute operator ‘.’ (dot) provide quick and easy access to pandas data structures across a wide range of use cases. The index is like an address, that’s how any data point across the data frame or series can be accessed. Rows and columns both have indexes.

The axis labeling information in pandas objects serves many purposes:

· Identifies data (i.e. provides metadata) .

· Enables automatic and explicit data alignment.

· Allows intuitive getting and setting of subsets of the data.

2. Indexing in terms of analytics

· implicit (numeric values)

· explicit ( label type)

3. Indexing in terms of programming

3.1.label-based indexing(.loc)

3.1.1. A single label, e.g. 5 or 'a' (5 is interpreted as a label of the index, not an integer position index).

3.1.2. A list or array of labels ['a', 'b', 'c'].

3.1.3. A slice object with labels 'a':'f' (both the start and the stop are included)

3.1.4. A boolean array (any NA values will be treated as False).

3.1.5. A callable function with one argument).

# a single label
df1=df.loc['a', '5']
# a list / array of labels
df2=df.loc[['a', 'b', 'c'], ]
# a slice object with labels
df3=df.loc['a':'b',]
# a boalean array
df4=df.loc['a']>5
# a callabel function 
df5=df.loc[lambda df: df['A'] > 0, :]

3.2. Integer/position-based indexing (.iloc)

3.2.1. An integer e.g. 5.

3.2.2. A list or array of integers [4, 3, 0].

3.2.3. A slice object with ints 1:7.

3.2.4. A boolean array (any NA values will be treated as False).

3.2.5. A callable function with one argument.

 a single label
df6=df.iloc[5]
# a list / array of labels
df7=df.iloc[[3, 4, 5],]
# a slice object with labels
df8=df.iloc[7:1,]
# a boalean array
df9=df.iloc[5]>5
# a callabel function 
df10=df.iloc[:, lambda df: [0, 1]]