Scala REPL

Flink comes with an integrated interactive Scala Shell. It can be used in a local setup as well as in a cluster setup.

To use the shell with an integrated Flink cluster just execute:

  1. bin/start-scala-shell.sh local

in the root directory of your binary Flink directory. To run the Shell on a cluster, please see the Setup section below.

Usage

The shell supports DataSet, DataStream, Table API and SQL. Four different Environments are automatically prebound after startup. Use “benv” and “senv” to access the Batch and Streaming ExecutionEnvironment respectively. Use “btenv” and “stenv” to access BatchTableEnvironment and StreamTableEnvironment respectively.

DataSet API

The following example will execute the wordcount program in the Scala shell:

  1. Scala-Flink> val text = benv.fromElements(
  2.   "To be, or not to be,--that is the question:--",
  3.   "Whether 'tis nobler in the mind to suffer",
  4.   "The slings and arrows of outrageous fortune",
  5.   "Or to take arms against a sea of troubles,")
  6. Scala-Flink> val counts = text
  7.     .flatMap { _.toLowerCase.split("\\W+") }
  8.     .map { (_, 1) }.groupBy(0).sum(1)
  9. Scala-Flink> counts.print()

The print() command will automatically send the specified tasks to the JobManager for execution and will show the result of the computation in the terminal.

It is possible to write results to a file. However, in this case you need to call execute, to run your program:

  1. Scala-Flink> benv.execute("MyProgram")

DataStream API

Similar to the batch program above, we can execute a streaming program through the DataStream API:

  1. Scala-Flink> val textStreaming = senv.fromElements(
  2.   "To be, or not to be,--that is the question:--",
  3.   "Whether 'tis nobler in the mind to suffer",
  4.   "The slings and arrows of outrageous fortune",
  5.   "Or to take arms against a sea of troubles,")
  6. Scala-Flink> val countsStreaming = textStreaming
  7.     .flatMap { _.toLowerCase.split("\\W+") }
  8.     .map { (_, 1) }.keyBy(0).sum(1)
  9. Scala-Flink> countsStreaming.print()
  10. Scala-Flink> senv.execute("Streaming Wordcount")

Note, that in the Streaming case, the print operation does not trigger execution directly.

The Flink Shell comes with command history and auto-completion.

Table API

The example below is a wordcount program using Table API:

  1. Scala-Flink> import org.apache.flink.table.functions.TableFunction
  2. Scala-Flink> val textSource = stenv.fromDataStream(
  3.   senv.fromElements(
  4.     "To be, or not to be,--that is the question:--",
  5.     "Whether 'tis nobler in the mind to suffer",
  6.     "The slings and arrows of outrageous fortune",
  7.     "Or to take arms against a sea of troubles,"),
  8.   'text)
  9. Scala-Flink> class $Split extends TableFunction[String] {
  10.     def eval(s: String): Unit = {
  11.       s.toLowerCase.split("\\W+").foreach(collect)
  12.     }
  13.   }
  14. Scala-Flink> val split = new $Split
  15. Scala-Flink> textSource.join(split('text) as 'word).
  16.     groupBy('word).select('word, 'word.count as 'count).
  17.     toRetractStream[(String, Long)].print
  18. Scala-Flink> senv.execute("Table Wordcount")
  1. Scala-Flink> import org.apache.flink.table.functions.TableFunction
  2. Scala-Flink> val textSource = btenv.fromDataSet(
  3.   benv.fromElements(
  4.     "To be, or not to be,--that is the question:--",
  5.     "Whether 'tis nobler in the mind to suffer",
  6.     "The slings and arrows of outrageous fortune",
  7.     "Or to take arms against a sea of troubles,"), 
  8.   'text)
  9. Scala-Flink> class $Split extends TableFunction[String] {
  10.     def eval(s: String): Unit = {
  11.       s.toLowerCase.split("\\W+").foreach(collect)
  12.     }
  13.   }
  14. Scala-Flink> val split = new $Split
  15. Scala-Flink> textSource.join(split('text) as 'word).
  16.     groupBy('word).select('word, 'word.count as 'count).
  17.     toDataSet[(String, Long)].print

Note, that using $ as a prefix for the class name of TableFunction is a workaround of the issue that scala incorrectly generated inner class name.

SQL

The following example is a wordcount program written in SQL:

  1. Scala-Flink> import org.apache.flink.table.functions.TableFunction
  2. Scala-Flink> val textSource = stenv.fromDataStream(
  3.   senv.fromElements(
  4.     "To be, or not to be,--that is the question:--",
  5.     "Whether 'tis nobler in the mind to suffer",
  6.     "The slings and arrows of outrageous fortune",
  7.     "Or to take arms against a sea of troubles,"), 
  8.   'text)
  9. Scala-Flink> stenv.createTemporaryView("text_source", textSource)
  10. Scala-Flink> class $Split extends TableFunction[String] {
  11.     def eval(s: String): Unit = {
  12.       s.toLowerCase.split("\\W+").foreach(collect)
  13.     }
  14.   }
  15. Scala-Flink> stenv.registerFunction("split", new $Split)
  16. Scala-Flink> val result = stenv.sqlQuery("""SELECT T.word, count(T.word) AS `count` 
  17.     FROM text_source 
  18.     JOIN LATERAL table(split(text)) AS T(word) 
  19.     ON TRUE 
  20.     GROUP BY T.word""")
  21. Scala-Flink> result.toRetractStream[(String, Long)].print
  22. Scala-Flink> senv.execute("SQL Wordcount")
  1. Scala-Flink> import org.apache.flink.table.functions.TableFunction
  2. Scala-Flink> val textSource = btenv.fromDataSet(
  3.   benv.fromElements(
  4.     "To be, or not to be,--that is the question:--",
  5.     "Whether 'tis nobler in the mind to suffer",
  6.     "The slings and arrows of outrageous fortune",
  7.     "Or to take arms against a sea of troubles,"), 
  8.   'text)
  9. Scala-Flink> btenv.createTemporaryView("text_source", textSource)
  10. Scala-Flink> class $Split extends TableFunction[String] {
  11.     def eval(s: String): Unit = {
  12.       s.toLowerCase.split("\\W+").foreach(collect)
  13.     }
  14.   }
  15. Scala-Flink> btenv.registerFunction("split", new $Split)
  16. Scala-Flink> val result = btenv.sqlQuery("""SELECT T.word, count(T.word) AS `count` 
  17.     FROM text_source 
  18.     JOIN LATERAL table(split(text)) AS T(word) 
  19.     ON TRUE 
  20.     GROUP BY T.word""")
  21. Scala-Flink> result.toDataSet[(String, Long)].print

Adding external dependencies

It is possible to add external classpaths to the Scala-shell. These will be sent to the Jobmanager automatically alongside your shell program, when calling execute.

Use the parameter -a <path/to/jar.jar> or --addclasspath <path/to/jar.jar> to load additional classes.

  1. bin/start-scala-shell.sh [local | remote <host> <port> | yarn] --addclasspath <path/to/jar.jar>

Setup

To get an overview of what options the Scala Shell provides, please use

  1. bin/start-scala-shell.sh --help

Local

To use the shell with an integrated Flink cluster just execute:

  1. bin/start-scala-shell.sh local

Remote

To use it with a running cluster start the scala shell with the keyword remote and supply the host and port of the JobManager with:

  1. bin/start-scala-shell.sh remote <hostname> <portnumber>

Yarn Scala Shell cluster

The shell can deploy a Flink cluster to YARN, which is used exclusively by the shell. The shell deploys a new Flink cluster on YARN and connects the cluster. You can also specify options for YARN cluster such as memory for JobManager, name of YARN application, etc.

For example, to start a Yarn cluster for the Scala Shell with two TaskManagers use the following:

  1. bin/start-scala-shell.sh yarn -n 2

For all other options, see the full reference at the bottom.

Yarn Session

If you have previously deployed a Flink cluster using the Flink Yarn Session, the Scala shell can connect with it using the following command:

  1. bin/start-scala-shell.sh yarn

Full Reference

  1. Flink Scala Shell
  2. Usage: start-scala-shell.sh [local|remote|yarn] [options] <args>...
  4. Command: local [options]
  5. Starts Flink scala shell with a local Flink cluster
  6.   -a <path/to/jar> | --addclasspath <path/to/jar>
  7.         Specifies additional jars to be used in Flink
  8. Command: remote [options] <host> <port>
  9. Starts Flink scala shell connecting to a remote cluster
  10.   <host>
  11.         Remote host name as string
  12.   <port>
  13.         Remote port as integer
  15.   -a <path/to/jar> | --addclasspath <path/to/jar>
  16.         Specifies additional jars to be used in Flink
  17. Command: yarn [options]
  18. Starts Flink scala shell connecting to a yarn cluster
  19.   -jm arg | --jobManagerMemory arg
  20.         Memory for JobManager container with optional unit (default: MB)
  21.   -nm <value> | --name <value>
  22.         Set a custom name for the application on YARN
  23.   -qu <arg> | --queue <arg>
  24.         Specifies YARN queue
  25.   -s <arg> | --slots <arg>
  26.         Number of slots per TaskManager
  27.   -tm <arg> | --taskManagerMemory <arg>
  28.         Memory per TaskManager container with optional unit (default: MB)
  29.   -a <path/to/jar> | --addclasspath <path/to/jar>
  30.         Specifies additional jars to be used in Flink
  31.   --configDir <value>
  32.         The configuration directory.
  33.   -h | --help
  34.         Prints this usage text