Tabulify - Pipeline

Pipeline

A pipeline defines a serie of step where:

the first step called the supplier step:
- selects data resources
- and send them to the next step
and the following steps called intermediate steps:
- accept the output (target) of the previous step as source
- and produce target (output) resources.

Syntax

A pipeline is a manifest file that follows the following syntax:

kind: pipeline
spec:
  # A comment
  comment: "A comment"
  # Pipeline Arguments
  arguments:
    # strictness
    strict-execution: true
    xxx: xxx
  # Pipeline steps
  steps:
      # first operation of a pipeline is a supplier step
      - name: supplierStep1
        operation: xxx
        arguments:
          xxx: xxx
          .....
      # next operation are called intermediate
      # They works on the resources returned from the upstream (a supplier or intermediate step)
      - name: intermediateStep2
        operation: xxx
        arguments:
          xxx: xxx
          .....

Arguments

Arguments are tabulify pipeline Parameters attribute

Duration Control

Duration Control Parameters control the duration of a pipeline.

Name	Default	Description
max-cycle-count	Illimited	The maximum number of cycle (ie the count of data path send in the pipeline).
timeout	Illimited	A timeout duration
timeout-type	Error	A timeout type (duration or error) The duration value will not throw an error while error will.

Stream

See Stream pipeline Arguments

Error Control

See Error Handling

Strictness

strict-execution permits to set the execution strictness

Derived Attributes

attributes retrieved or computed.

Name	Description
logical-name	The name of the pipeline (ie the file without extension)
processing-type	The processing type as determined by the type of the supplier (ie batch or stream)
start-time	The start time of the pipeline execution

Steps

The steps are a series of step (ie operation and arguments)

Execution

You can execute a pipeline with the flow execute command

Attributes

Pipeline supports 2 kind of attributes:

arguments
derived_attributes

Metrics

Pipeline Metrics

Pipelines shows the following metrics:

Name	Description
Total Elapsed Time	The total elapsed time of the pipeline execution and completion
Execution Elapsed Time	The elapsed time of the pipeline execution to the timeout without the completion (ie last cycle)
For Stream Pipeline
Total Poll Wait Time	The total time that the pipeline was waiting to poll due to poll-interval
Total Push Wait Time	The total time that the pipeline was waiting to push due to push-interval

Note on:

Total Elapsed time and Timeout: The timeout is the maximum duration of the main execution. Because the pipeline needs to close and complete the pending intermediate steps operation, the total elapsed time is always a little bit greater than the timeout.
Total Poll Wait Time and Timeout: the Total Poll Wait Time is always greater than the Timeout because the pipeline may wait in the completion (ie last cycle)

Pipeline Step Metrics

Input Counter: the number of data resource received
Output Counter : the number of data resource supplied
Execution Counter: the number of step execution. In a stream pipeline, the execution is:
- for the supplier, the number of time, the poll function was called
- for a batch intermediate step, the number of time, the window interval was exhausted and the step was executed
- for a stream intermediate step, the step execution (ie one data resource, one execution)
Error Counter: the number of errors