# Import packages
import numpy as np
import pandas as pd

# Create a Series from a ndarray
s = pd.Series(np.arange(1,20,5))
# Indexes are created by default
s

# Create a Series with indexes
s = pd.Series([1, 2, 3, 4, 5], index=['a', 'b', 'c', 'd', 'e'])

# Query the indexes of a Series
s.index

# Create a Series from a dict
d = {'k1': 'hello', 'k2': 10, 'k3': 19.8}
pd.Series(d)

# Create a Series from a dict
d = {'k1': 1.0, 'k2': 111, 'k3': 19.8}
pd.Series(d)

# Create a Series from a scalar value
s = pd.Series(10., index=['a', 'b', 'c'])
s

# Query the data type
s.dtype

# Series to ndarray-like
s.values

# Access values by index values
s = pd.Series(np.random.randn(5), index=['a', 'b', 'c', 'd', 'e'])
s['e'] 

# Access and Set values by index values
s['a'] = 0
s

# Query if index exists
'c' in s

# Sort values
s.sort_values(ascending=False)

# Create a DataFrame from a dict of Series
d = {'first': pd.Series([10, 20, 30], index=['a', 'b', 'c']),
     'second': pd.Series([1, 2, 3, 4], index=['a', 'b', 'c', 'd'])}
df = pd.DataFrame(d)
df

# Query the indexes
df.index

# Query the columns
df.columns

# Query the datatypes
df.dtypes

# Create a dataframe picking some indexes
pd.DataFrame(d, index=['c', 'b', 'd'])

# Create a dataframe picking some columns too
pd.DataFrame(d, index=['c', 'b', 'd'], columns=['second', 'third'])

# Create a dataframe from a dict of ndarrays/list
d = {'first': [10, 20, 30, 40], 'second': [1.1, 1.2, 1.3, 1.4]}
df = pd.DataFrame(d)
df

# Create dataframe defining indexes
pd.DataFrame(d, index = ['a', 'b', 'c', 'd'])

# Create a dataframe from a NumPy array
df = pd.DataFrame(np.random.randn(10, 4), index=list('abcdefghij'),
             columns=list('ABCD'))
df

# View first rows
df.head()

# View last rows
df.tail(3)

# Return a numpy representation
A = df.values
A

# Get dataframe column as NumPy presentation
import numpy as np
X = np.c_[df['A']]
X

# Shape of A are rows and columns of the dataframe
A.shape

# Show a stat summary of the data
df.describe()

# Show a summary of our dataframe
df.info()

# Transpose data
df.T

# Sort by values
df.sort_values(by='C', ascending=False)

# Initial Dataframe
df = pd.DataFrame(np.random.randn(10, 4), index=list('abcdefghij'),
             columns=list('ABCD'))
df

# Select a single column. Returns a Series
df['A']

# Select a single column. Returns a Series
df.B

# Select a slice of rows. Recall indexes start at 0
df[2:4]

# Select on multi-axis
df.loc[:, ['A', 'D']]

# Select a row on multi-axis. Returns a Series
df.loc['c', ['A', 'D']]

# Select on multi-axis
df.loc[['a', 'c'], ['A', 'D']]

# Get a scalar value
df.loc['c', 'C']

# Get a scalar value using 'at'
df.at['c', 'C']

# Set values on a dataframe
indices = list(['a', 'd', 'f'])
df.loc[indices, 'D'] = 100
df

# Select by position
df.iloc[2]

# Select like slicing
df.iloc[2:4,1:3]

# Select specific locations
df.iloc[[1, 3],[0, 2]]

# Slice rows
df.iloc[4:7, :]

# Slice columns
df.iloc[:, 2:4]

# Get a scalar value at specific position
df.iloc[2,3]

# Get a scalar value at specific position using 'iat'
df.iat[2,3]

# Boolean indexing

# Load dataset into a dataframe
import os
datapath = os.path.join("datasets", "islr", "")
auto = pd.read_csv(datapath + "dataset_filename.csv", 
                   delim_whitespace=True, na_values='?')

# Query NaN values
auto[auto.isnull().any(axis=1)]

# Count values for a particular column
auto['year'].value_counts()

# Generate a correlation matrix
auto.corr()

# Copy a dataframe
auto_copy = auto.copy()

# Clean data with NaN Values for a column
auto_copy = auto_copy.dropna(subset=['year'])

# Clean data with NaN Values for all dataframe
auto_copy = auto_copy.dropna()

# Drop a column
auto_copy = auto_copy.drop('name', axis=1)
auto_copy.head()

# Make a list from dataframe column names
list(auto_copy)