通过套接字发送以下词： Spark Hadoop Spark is good Hadoop is good 并用Structured Streaming读取套接字流实现词频统计，分别设置输出模式为Append模式

首先，你需要编写一个Python脚本来通过套接字发送指定的字符串。这里是一个简单的示例：

import socket
import time

def send_words(words):
    host = 'localhost'
    port = 9999  # 确保你的服务器监听在这个端口
    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
    sock.connect((host, port))

    for word in words:
        message = f"{word}\n"
        sock.sendall(message.encode())
        time.sleep(1)  # 每发送完一个单词间隔一段时间，模拟实时数据

send_words(['Spark', 'Hadoop', 'Spark is good', 'Hadoop is good'])
sock.close()

接下来，使用Apache Spark的Structured Streaming来接收套接字流并计算词频。创建一个名为`process_socket_stream.py`的文件，添加以下内容：

from pyspark.sql import SparkSession, functions as F
from pyspark.sql.types import StructType, StringType

spark = SparkSession.builder.appName('SocketWordCount').getOrCreate()

# 定义数据源结构
schema = StructType([StructField('line', StringType(), True)])

# 创建一个输入流，连接到之前定义的套接字地址
socket_stream = spark.readStream.format("socket").option("host", "localhost").option("port", 9999).load()

# 解析每行数据
parsed_data = socket_stream.selectExpr("CAST(line AS STRING)")

# 分词
words = parsed_data.select(F.explode(F.split(F.col("line"), " ")).alias("word"))

# 使用mapPartitionsWithState函数来保存状态（词频）
state = words.groupBy().count().withColumnRenamed("count", "total_count")
windowed_state = state.withWatermark("timestamp", "10 seconds")

# 计算词频，输出模式为Append，意味着会累计所有新到来的单词
word_counts = windowed_state \
    .join(state, (F.col('word') == F.col('word')) & ((F.lit(None).cast('timestamp') <= F.col('window.end')) & (F.col('window.start') <= F.lit(None).cast('timestamp'))), "outer") \
    .when(F.col('window.end') > F.lit(None).cast('timestamp'), F.when(F.col('total_count') >= F.col('old.total_count'), F.col('total_count')).otherwise(F.col('old.total_count'))) \
    .otherwise(F.col('new.total_count'))

query = word_counts.select("word", "total_count").writeStream.outputMode("append").format("console").start()

query.awaitTermination()

运行此脚本后，它会持续监听套接字流中的新数据，并在控制台显示实时的词频统计，按Append模式累加计数。注意，实际部署时可能需要调整服务器端的相应设置以匹配上述代码。