解决pyspark环境下GraphFrames报错问题

背景

Spark图计算实战：在pyspark环境下使用GraphFrames库

环境

1. mac os
2. conda→python=3.8
3. jupyter notebook
4. pyspark=3.3.0
5. graphframes=0.6

代码

from pyspark import SparkConf, SparkContext
from pyspark.sql import SparkSession
from graphframes import GraphFrame

sc = SparkContext()
spark = SparkSession(sc)

# Vertics DataFrame
vertics = spark.createDataFrame([
  ("a", "Alice", 34),
  ("b", "Bob", 36),
  ("c", "Charlie", 37),
  ("d", "David", 29),
  ("e", "Esther", 32),
  ("f", "Fanny", 38),
  ("g", "Gabby", 60)
], ["id", "name", "age"])
vertics.show()

# Edges DataFrame
edges = spark.createDataFrame([
  ("a", "b", "friend"),
  ("b", "c", "follow"),
  ("c", "b", "follow"),
  ("f", "c", "follow"),
  ("e", "f", "follow"),
  ("e", "d", "friend"),
  ("d", "a", "friend"),
  ("a", "e", "friend"),
  ("g", "e", "follow")
], ["src", "dst", "relationship"])
edges.show()

# Create a GraphFrame
graph = GraphFrame(vertics, edges)

报错信息

在执行graph初始化语句时，报错信息如下

pyspark: Py4JJavaError: An error occurred while calling o138.loadClass.: java.lang.ClassNotFoundException: org.graphframes.GraphFramePythonAPI

排查错误

参考

graphframes环境报错：

1. 参考1
2. 参考2

在终端执行

pyspark --packages graphframes:graphframes:0.8.0-spark3.0-s_2.12 --jars graphframes-0.8.0-spark3.0-s_2.12.jar

发现报错原因是缺少graphframes-0.8.0-spark3.0-s_2.12.jar，因此需要到官网下载
官网链接：graphframes
下载之后，由于终端启动pyspark路径定位在user路径下，如果要在终端启动pyspark，需要将jar包放在图片路径上
对于jupyter notebook，只需要声明SparkContext环境后，添加jar包所在路径即可，对jar包位置没有要求

修改后的代码

from pyspark import SparkConf, SparkContext
from pyspark.sql import SparkSession
from graphframes import GraphFrame

sc = SparkContext()
spark = SparkSession(sc)
# 添加jar包路径，修复bug
sc.addPyFile("../envs/project/lib/python3.8/site-packages/pyspark/jars/graphframes-0.8.0-spark3.0-s_2.12.jar")

# Vertics DataFrame
vertics = spark.createDataFrame([
  ("a", "Alice", 34),
  ("b", "Bob", 36),
  ("c", "Charlie", 37),
  ("d", "David", 29),
  ("e", "Esther", 32),
  ("f", "Fanny", 38),
  ("g", "Gabby", 60)
], ["id", "name", "age"])
vertics.show()

# Edges DataFrame
edges = spark.createDataFrame([
  ("a", "b", "friend"),
  ("b", "c", "follow"),
  ("c", "b", "follow"),
  ("f", "c", "follow"),
  ("e", "f", "follow"),
  ("e", "d", "friend"),
  ("d", "a", "friend"),
  ("a", "e", "friend"),
  ("g", "e", "follow")
], ["src", "dst", "relationship"])
edges.show()

# Create a GraphFrame
graph = GraphFrame(vertics, edges)

正常运行

一些关于Spark Graph图计算模块的参考

Spark raphFrames图计算API：https://blog.csdn.net/weixin_45839604/article/details/117751806

基于pyspark图计算的算法实例：

https://blog.csdn.net/weixin_39198406/article/details/104940179

基于SparkGraph的社交关系图谱实战

https://blog.51cto.com/u_11200224/5275069

graphframes环境报错：

https://blog.csdn.net/m0_37754282/article/details/110086095

https://blog.csdn.net/qq_42166929/article/details/105983616