Type Members

1.  class ArtifactManager extends AnyRef

   The Artifact Manager is responsible for handling and transferring artifacts from the local client to the server (local/remote).

2.  trait ClassFinder extends AnyRef
3.  class ExecutePlanResponseReattachableIterator extends WrappedCloseableIterator[ExecutePlanResponse] with Logging

   Retryable iterator of ExecutePlanResponses to an ExecutePlan call.

   Retryable iterator of ExecutePlanResponses to an ExecutePlan call.

   It can handle situations when:

   • the ExecutePlanResponse stream was broken by retryable network error (governed by retryPolicy)
   • the ExecutePlanResponse was gracefully ended by the server without a ResultComplete message; this tells the client that there is more, and it should reattach to continue.

   Initial iterator is the result of an ExecutePlan on the request, but it can be reattached with ReattachExecute request. ReattachExecute request is provided the responseId of last returned ExecutePlanResponse on the iterator to return a new iterator from server that continues after that. If the initial ExecutePlan did not even reach the server, and hence reattach fails with INVALID_HANDLE.OPERATION_NOT_FOUND or INVALID_HANDLE.SESSION_NOT_FOUND, we attempt to retry ExecutePlan.

   In reattachable execute the server does buffer some responses in case the client needs to backtrack. To let server release this buffer sooner, this iterator asynchronously sends ReleaseExecute RPCs that instruct the server to release responses that it already processed.

4.  case class PlanCompressionOptions(thresholdBytes: Int, algorithm: String) extends Product with Serializable
5.  class REPLClassDirMonitor extends ClassFinder

   A generic ClassFinder implementation that traverses a specific REPL output directory.

6.  class ResponseValidator extends Logging
7.  case class RetryPolicy(maxRetries: Option[Int] = None, initialBackoff: FiniteDuration = FiniteDuration(1000, "ms"), maxBackoff: Option[FiniteDuration] = None, backoffMultiplier: Double = 1.0, jitter: FiniteDuration = FiniteDuration(0, "s"), minJitterThreshold: FiniteDuration = FiniteDuration(0, "s"), canRetry: (Throwable) => Boolean, name: String, recognizeServerRetryDelay: Boolean = false, maxServerRetryDelay: Option[FiniteDuration] = None) extends Product with Serializable

   RetryPolicy configure the retry mechanism in GrpcRetryHandler

   RetryPolicy configure the retry mechanism in GrpcRetryHandler

   maxRetries

   Maximum number of retries.

   initialBackoff

   Start value of the exponential backoff (ms).

   maxBackoff

   Maximal value of the exponential backoff (ms).

   backoffMultiplier

   Multiplicative base of the exponential backoff.

   jitter

   Sample a random value uniformly from the range [0, jitter] and add it to the backoff.

   minJitterThreshold

   Minimal value of the backoff to add random jitter.

   canRetry

   Function that determines whether a retry is to be performed in the event of an error.

   name

   Name of the policy.

   recognizeServerRetryDelay

   Per gRPC standard, the server can send error messages that contain RetryInfo message with retry_delay field indicating that the client should wait for at least retry_delay amount of time before retrying again, see: https://github.com/googleapis/googleapis/blob/master/google/rpc/error_details.proto#L91 If this flag is set to true, RetryPolicy will use RetryInfo.retry_delay field in the backoff computation. Server's retry_delay can override client's maxBackoff. This flag does not change which errors are retried, only how the backoff is computed. DefaultPolicy additionally has a rule for retrying any error that contains RetryInfo.

   maxServerRetryDelay

   Limit for the server-provided retry_delay.

8.  class SparkConnectStubState extends Logging
9.  case class ToStub(value: Long) extends Product with Serializable

   Class used to test stubbing.

   Class used to test stubbing. This needs to be in the main source tree, because this is not synced with the connect server during tests.