构建 ETL 与 ELT 流水线

使用 Python 的 ETL 和 ELT

Jake Roach

Data Engineer

从 CSV 文件提取数据

import pandas as pd

# Read in the CSV file to a DataFrame
data_frame = pd.read_csv("raw_data.csv")

# Output the first few rows
data_frame.head()

   name         num_firms   total_income
0  Advertising     58         3892.41
1  Apparel         39         5422.69
...
49 Trucking        35         17324.36

read_csv()

接收文件路径，返回 DataFrame
delimiter、header、engine

.head()

输出 DataFrame 的前 n 行

筛选 DataFrame

   name         num_firms   total_income
0  Advertising     58         3892.41
1  Apparel         39         5422.69
...
49 Trucking        35         17324.36


         name         num_firms
      1  Apparel         39
      37 Apparel         61

# First, by rows
data_frame.loc[data_frame["name"] == "Apparel", :]

# Then, by columns
data_frame.loc[:, ["name", "num_firms"]]

.loc

过滤 DataFrame
: 表示"全部"

将 DataFrame 写入 CSV 文件

# Write a DataFrame to a .csv file
data_frame.to_csv("cleaned_data.csv")

.to_csv()

接收 path，将 DataFrame 写入该路径的文件
可接受其他参数以定制输出

其他选项，如：

.to_json()、.to_excel()、.to_sql()

运行 SQL 查询

data_warehouse.execute(  # Use Python clients or other tools to run SQL queries
    """
    CREATE TABLE total_sales AS 
        SELECT
            ds,
            SUM(sales)
        FROM raw_sales_data
        GROUP BY ds;
    """
)

使用如 .execute() 的工具来运行 SQL 查询

整合示例

# Define extract(), transform(), and load() functions
...

def transform(data_frame, value):
    return data_frame.loc[data_frame["name"] == value, ["name", "num_firms"]]

# First, extract data from a .csv
extracted_data = extract(file_name="raw_data.csv")

# Then, transform the `extracted_data`
transformed_data = transform(data_frame=extracted_data, value="Apparel")

# Finally, load the `transformed_data`
load(data_frame=transformed_data, file_name="cleaned_data.csv")

Passons à la pratique !

使用 Python 的 ETL 和 ELT