HTTP155 Initial commit for open source Flink

[davidradl]

Description

First commit for Flink, includes:
changing the pom in line with the standard Flink connector including

• Flink checkstyle
• Flink spotless
• standardize the name of the Maven artifact
• amend the package to follow Flink standards
• added Apache license to pom files and java files
• changed the unit tests away from jacoco to use surefire
• removed jetbrains annotations
• removed the use of a google annotation it was complaining about
• removed use of ImmutableList which it did not like.
• separate the pom into a parent , code and SQL shading pom in line with Kafka.
• licence checker passes
• notices - added the top level NOTICE and added a NOTICE in the SQL project meta-inf to declare the dependancies - this is the same approach the Flink Kafka connector
  Resolves HTTP155

PR Checklist

• Tests added
• [n/a] Changelog updated

I am looking to get this code into Flink with the commit history as-is.

There are still things to do:

• migrate tests away from Mockito to avoid the banned
• test datastream
• do the docs
• consider upgrade the WireMock which is very back level.
• amend the config option names to be flink with gid aliiases
• amend the prefix to not include gid and move to Flink
• the parent connector version is 1.0.0 which is same as the Kafka connector. 1.1.0 did not work but we should look into upgrading to use this.

I think it makes sense to get this initial commit in to push the commit history over. So all subsequent contributions will count as Flink contributions.

[davidradl]

fyi @grzegorz8

[davidradl]

@grzegorz8 I have split the poms out so there is now a parent , the code and sql (with the shading).

Accummulating a list of tasks I need to do before offering to Flink :

• The build error on java 21 appears to be a known spotless issue.
• assess commented out pom sections

[davidradl]

I did not see this build breaks locally. Changing to draft while I address them.

[davidradl]

@grzegorz8 thanks for your valuable feedback - I have addressed what you have raised - so I can see the build issues again (that I do not see locally) so I can address before taking the PR out of draft. I am aware that Flink is looking to support java 17 and 21 as well as 11. 21 was failing prior to the pom split.

[davidradl]

I have fixed the convergence locally.
I am seeing that if I include flink-table-test-utils in the parent and http poms then the junit tests work locally. But I get an unused dependancy - which I can ignore in the http pom, but the ci yml does not allow me to ignore this dependancy.
If I remove this dependancy the unit tests fail as it cannot find an ExecutorFactory. I assume that the presence of the plugin means that its content is in the classpath for the test, but this is a runtime dependancy not a compile time one.

without the plugin I see errors like:
[ERROR] org.apache.flink.connector.http.internal.sink.HttpSinkConnectionTest.testServerErrorConnection Time elapsed: 0.014 s <<< ERROR!
java.lang.IllegalStateException: No ExecutorFactory found to execute the application.
at org.apache.flink.core.execution.DefaultExecutorServiceLoader.getExecutorFactory(DefaultExecutorServiceLoader.java:88)
at org.apache.flink.streaming.api.environment.StreamExecutionEnvironment.getPipelineExecutor(StreamExecutionEnvironment.java:2995)
at org.apache.flink.streaming.api.environment.StreamExecutionEnvironment.executeAsync(StreamExecutionEnvironment.java:2469)
at org.apache.flink.streaming.api.environment.StreamExecutionEnvironment.execute(StreamExecutionEnvironment.java:2351)
at org.apache.flink.streaming.api.environment.LocalStreamEnvironment.execute(LocalStreamEnvironment.java:68)
at org.apache.flink.streaming.api.environment.StreamExecutionEnvironment.execute(StreamExecutionEnvironment.java:2325)
at org.apache.flink.connector.http.internal.sink.HttpSinkConnectionTest.testServerErrorConnection(HttpSinkConnectionTest.java:268)
at java.base/java.lang.reflect.Method.invoke(Method.java:572)
at java.base/java.util.ArrayList.forEach(ArrayList.java:1541)
at java.base/java.util.ArrayList.forEach(ArrayList.java:1541)

Still investigating - any thoughts ?

[davidradl]

The 2nd commit fixes:

• parent pom was not specifying the versions correctly- the change allows the flink.version to be used in the child
• copied the pattern used in the Kafka connector around bringing in the Flink clients and planner.
• the planner comes in with an associated scala level.
• added exclusions for convergence reasons.
• CI clean locally but fails with an unused import error in build machine

[davidradl]

The unused import error occurs at java 21 and not at 8. I have not tested 17. It seems to related to reflection. There appears to be a circumvention - investigating.

[davidradl]

java 21 compiles on build machine - the license checker is now failing.

[davidradl]

@grzegorz8 I have the Flink CI running cleanly now. Could you review please?

I am planning to do the following prior to donating to Flink

• cleanup commented out content
• address any more points you raise
• directories are being created at the top level of the models - I need to resolve this
• raise Jiras for remaining work
• more local testing

I hope to be be able to donate today or early next week.

[davidradl]

Though the tests are green the sql jar file does not contain all the expected files. Moving to draft while I sort that out. This was a local issue - not to do with the code.