import pandas as pd
import numpy as np

# 創(chuàng)建示例DataFrame
data = {
    'Name': ['Alice', 'Bob', 'Charlie'],
    'Age': [25, 30, 35],
    'Salary': [50000.0, 60000.5, 70000.0],
    'Join_Date': pd.to_datetime(['2020-01-15', '2019-05-20', '2021-11-10']),
    'Is_Manager': [True, False, True]
}
df = pd.DataFrame(data)

# 查看整個DataFrame的數(shù)據(jù)類型
print(df.dtypes)
"""
Name           object
Age             int64
Salary        float64
Join_Date      datetime64[ns]
Is_Manager       bool
dtype: object
"""

# 查看單列的數(shù)據(jù)類型
print(df['Age'].dtype)  # 輸出: int64

# 查看各列內(nèi)存使用量
print(df.memory_usage())
"""
Index        128
Name          24
Age           24
Salary        24
Join_Date     24
Is_Manager     3
dtype: int64
"""

# 查看詳細內(nèi)存使用情況(deep=True會計算object類型的真實內(nèi)存使用)
print(df.memory_usage(deep=True))
"""
Index        128
Name         174
Age           24
Salary        24
Join_Date     24
Is_Manager     3
dtype: int64
"""

# 將Age從int64轉(zhuǎn)換為float64
df['Age'] = df['Age'].astype('float64')
print(df['Age'].dtype)  # 輸出: float64

# 將Salary從float64轉(zhuǎn)換為int64(會截斷小數(shù)部分)
df['Salary'] = df['Salary'].astype('int64')
print(df['Salary'])
"""
0    50000
1    60000
2    70000
Name: Salary, dtype: int64
"""

# 將Is_Manager從bool轉(zhuǎn)換為str
df['Is_Manager'] = df['Is_Manager'].astype('str')
print(df['Is_Manager'].dtype)  # 輸出: object

# 將Name列轉(zhuǎn)換為分類類型
df['Name'] = df['Name'].astype('category')
print(df['Name'].dtype)  # 輸出: category

# 查看分類的類別
print(df['Name'].cat.categories)
"""
Index(['Alice', 'Bob', 'Charlie'], dtype='object')
"""

# 分類類型的內(nèi)存節(jié)省效果
print(f"原始內(nèi)存: {df['Name'].memory_usage(deep=True)}")
df['Name'] = df['Name'].astype('object')
print(f"轉(zhuǎn)換后內(nèi)存: {df['Name'].memory_usage(deep=True)}")
"""
原始內(nèi)存: 174
轉(zhuǎn)換后內(nèi)存: 180
"""

# 字符串轉(zhuǎn)換為日期時間
date_str = pd.Series(['2023-01-01', '2023-02-15', '2023-03-20'])
dates = pd.to_datetime(date_str)
print(dates.dtype)  # 輸出: datetime64[ns]

# 處理多種日期格式
mixed_dates = pd.Series(['01-01-2023', '2023/02/15', '15-March-2023'])
dates = pd.to_datetime(mixed_dates)
print(dates)
"""
0   2023-01-01
1   2023-02-15
2   2023-03-15
dtype: datetime64[ns]
"""

# 提取日期組成部分
df['Year'] = df['Join_Date'].dt.year
df['Month'] = df['Join_Date'].dt.month
print(df[['Join_Date', 'Year', 'Month']])
"""
   Join_Date  Year  Month
0 2020-01-15  2020      1
1 2019-05-20  2019      5
2 2021-11-10  2021     11
"""

# 創(chuàng)建包含數(shù)字字符串和缺失值的Series
mixed_data = pd.Series(['1', '2.5', '3.0', 'four', None])

# 安全轉(zhuǎn)換為數(shù)值(無法轉(zhuǎn)換的設(shè)為NaN)
numeric_data = pd.to_numeric(mixed_data, errors='coerce')
print(numeric_data)
"""
0    1.0
1    2.5
2    3.0
3    NaN
4    NaN
dtype: float64
"""

# 向下轉(zhuǎn)換為最小可能類型
numeric_data = pd.to_numeric(mixed_data, errors='coerce', downcast='integer')
print(numeric_data)
"""
0    1.0
1    2.5
2    3.0
3    NaN
4    NaN
dtype: float32
"""

# 創(chuàng)建類型不明確的數(shù)據(jù)
df_mixed = pd.DataFrame({
    'A': [1, 2, 3],
    'B': ['4', '5', '6'],
    'C': [True, False, True]
})

# 自動推斷更合適的類型
df_inferred = df_mixed.infer_objects()
print(df_inferred.dtypes)
"""
A      int64
B     object
C       bool
dtype: object
"""

# 轉(zhuǎn)換為支持pd.NA的最佳類型
df_best = df.convert_dtypes()
print(df_best.dtypes)
"""
Name           string
Age             Int64
Salary          Int64
Join_Date    datetime64[ns]
Is_Manager      boolean
Year             Int64
Month            Int64
dtype: object
"""

# 使用apply進行自定義轉(zhuǎn)換
def convert_salary(salary):
    if salary > 60000:
        return 'High'
    elif salary > 50000:
        return 'Medium'
    else:
        return 'Low'

df['Salary_Level'] = df['Salary'].apply(convert_salary)
print(df[['Salary', 'Salary_Level']])
"""
   Salary Salary_Level
0   50000          Low
1   60000       Medium
2   70000         High
"""

# 讀取CSV時指定數(shù)據(jù)類型
dtype_spec = {
    'user_id': 'int32',
    'product_id': 'category',
    'rating': 'float32'
}
# pd.read_csv('large_file.csv', dtype=dtype_spec)