在《0基础学习PyFlink——使用PyFlink的Sink将结果输出到Mysql》一文中,我们讲到如何通过定义Souce、Sink和Execute三个SQL,来实现数据读取、清洗、计算和入库。
如下图所示SQL是最高层级的抽象,在它之下是Table API。本文我们会将例子中的SQL翻译成Table API来实现等价的功能。
# """create table source (
# word STRING
# ) with (
# 'connector' = 'filesystem',
# 'format' = 'csv',
# 'path' = '{}'
# )
# """.format(input_path)
下面的SQL分为两部分:
SQL中的Table对应于Table API中的schema。它用于定义表的结构,比如有哪些类型的字段和主键等。
上述整个SQL整体对应于descriptor。即我们可以认为descriptor是表结构+连接器。
我们可以让不同的表和不同的连接器结合,形成不同的descriptor。这是一个组合关系,我们将在下面看到它们的组合方式。
# define the source schema
source_schema = Schema.new_builder() \
.column("word", DataTypes.STRING()) \
.build()
new_builder()会返回一个Schema.Builder对象;
column(self, column_name: str, data_type: Union[str, DataType])方法用于声明该表存在哪些类型、哪些名字的字段,同时返回之前的Builder对象;
最后的build(self)方法返回Schema.Builder对象构造的Schema对象。
# Create a source descriptor
source_descriptor= TableDescriptor.for_connector("filesystem") \
.schema(source_schema) \
.option('path', input_path) \
.format("csv") \
.build()
for_connector(connector: str)方法返回一个TableDescriptor.Builder对象;
schema(self, schema: Schema)将上面生成的source_schema 对象和descriptor关联;
option(self, key: Union[str, ConfigOption], value)用于指定一些参数,比如本例用于指定输入文件的路径;
format(self, format: Union[str, ‘FormatDescriptor’], format_option: ConfigOption[str] = None)用于指定内容的格式,这将指导怎么解析和入库;
build(self)方法返回TableDescriptor.Builder对象构造的TableDescriptor对象。
# """CREATE TABLE WordsCountTableSink (
# `word` STRING,
# `count` BIGINT,
# PRIMARY KEY (`word`) NOT ENFORCED
# ) WITH (
# 'connector' = 'jdbc',
# 'url' = 'jdbc:mysql://127.0.0.1:3306/words_count_db?useSSL=false',
# 'table-name' = 'WordsCountTable',
# 'driver'='com.mysql.jdbc.Driver',
# 'username'='admin',
# 'password'='pwd123'
# );
# """
sink_schema = Schema.new_builder() \
.column("word", DataTypes.STRING().not_null()) \
.column("count", DataTypes.BIGINT()) \
.primary_key("word") \
.build()
大部分代码在之前已经解释过了。我们主要关注于区别点:
# Create a sink descriptor
sink_descriptor = TableDescriptor.for_connector("jdbc") \
.schema(sink_schema) \
.option("url", "jdbc:mysql://127.0.0.1:3306/words_count_db?useSSL=false") \
.option("table-name", "WordsCountTable") \
.option("driver", "com.mysql.jdbc.Driver") \
.option("username", "admin") \
.option("password", "pwd123") \
.build()
这块代码主要是通过option来设置一些连接器相关的设置。可以看到这是用KV形式设计的,这样就可以让option方法有很大的灵活性以应对不同连接器千奇百怪的设置。
使用下面的代码将表创建出来,以供后续使用。
t_env.create_table("source", source_descriptor)
tab = t_env.from_path('source')
t_env.create_temporary_table("WordsCountTableSink", sink_descriptor)
# execute insert
# """insert into WordsCountTableSink
# select word, count(1) as `count`
# from source
# group by word
# """
tab.group_by(col('word')) \
.select(col('word'), lit(1).count) \
.execute_insert("WordsCountTableSink") \
.wait()
这儿需要介绍的就是lit。它用于生成一个表达式,诸如sum、max、avg和count等。
execute_insert(self, table_path_or_descriptor: Union[str, TableDescriptor], overwrite: bool = False)用于将之前的计算结果插入到Sink表中
import argparse
import logging
import sys
from pyflink.common import Configuration
from pyflink.table import (EnvironmentSettings, TableEnvironment, Schema)
from pyflink.table.types import DataTypes
from pyflink.table.table_descriptor import TableDescriptor
from pyflink.table.expressions import lit, col
def word_count(input_path):
config = Configuration()
# write all the data to one file
config.set_string('parallelism.default', '1')
env_settings = EnvironmentSettings \
.new_instance() \
.in_batch_mode() \
.with_configuration(config) \
.build()
t_env = TableEnvironment.create(env_settings)
# """create table source (
# word STRING
# ) with (
# 'connector' = 'filesystem',
# 'format' = 'csv',
# 'path' = '{}'
# )
# """
# define the source schema
source_schema = Schema.new_builder() \
.column("word", DataTypes.STRING()) \
.build()
# Create a source descriptor
source_descriptor = TableDescriptor.for_connector("filesystem") \
.schema(source_schema) \
.option('path', input_path) \
.format("csv") \
.build()
t_env.create_table("source", source_descriptor)
# """CREATE TABLE WordsCountTableSink (
# `word` STRING,
# `count` BIGINT,
# PRIMARY KEY (`word`) NOT ENFORCED
# ) WITH (
# 'connector' = 'jdbc',
# 'url' = 'jdbc:mysql://127.0.0.1:3306/words_count_db?useSSL=false',
# 'table-name' = 'WordsCountTable',
# 'driver'='com.mysql.jdbc.Driver',
# 'username'='admin',
# 'password'='pwd123'
# );
# """
# define the sink schema
sink_schema = Schema.new_builder() \
.column("word", DataTypes.STRING().not_null()) \
.column("count", DataTypes.BIGINT()) \
.primary_key("word") \
.build()
# Create a sink descriptor
sink_descriptor = TableDescriptor.for_connector("jdbc") \
.schema(sink_schema) \
.option("url", "jdbc:mysql://127.0.0.1:3306/words_count_db?useSSL=false") \
.option("table-name", "WordsCountTable") \
.option("driver", "com.mysql.jdbc.Driver") \
.option("username", "admin") \
.option("password", "pwd123") \
.build()
t_env.create_temporary_table("WordsCountTableSink", sink_descriptor)
# execute insert
# """insert into WordsCountTableSink
# select word, count(1) as `count`
# from source
# group by word
# """
tab = t_env.from_path('source')
tab.group_by(col('word')) \
.select(col('word'), lit(1).count) \
.execute_insert("WordsCountTableSink") \
.wait()
if __name__ == '__main__':
logging.basicConfig(stream=sys.stdout, level=logging.INFO, format="%(message)s")
parser = argparse.ArgumentParser()
parser.add_argument(
'--input',
dest='input',
required=False,
help='Input file to process.')
argv = sys.argv[1:]
known_args, _ = parser.parse_known_args(argv)
word_count(known_args.input)