Sample bundle smoke test

Proves the report machinery, not a risk number. A green run says the definition was read, the instance walked its states, the workflow dispatched its steps in the configured order, the grid executed the job, and the output came back to storage. Nothing downstream of that is checked here, because the value of the report is that a failure of any of those can be attributed without a domain calculation in the way.

The decision it feeds is whether to look at any other report yet. While this report is red, every other report's failure is ambiguous.

Cycle intraday (whenever an engineer asks for it), frequency on-demand. It is not scheduled and belongs to no batch. It is the report the smoke test of a freshly provisioned environment triggers by hand.

• A prepared input archive, named by the definition's prepared_input_key. The run selects pre-processing substitute, so it gathers no trades and no market data and assembles no bundle. The archive the repository ships is external/ore/packages/TA002_IR_Swap.tar.gz, the Input/ directory of the Academy's TA002_IR_Swap example: one 20Y EUR fixed-versus-float interest rate swap.
• The engine package the grid runs. The repository ships ORE 1.8.17.0 for the x64-linux triplet; the app version and its platform row are seeded, and the bytes are uploaded to storage per the compute runbook.
• A compute app version visible to the tenant the definition runs in.
• The input archive must already be in storage when the run starts. With pre-processing substituted, nothing in the run uploads it.

The engine's Output/ directory, archived per result in storage at the result's output URI, holding npv.csv, flows.csv and curves.csv. The repository also holds the Academy example's reference output at external/ore/examples/Academy/TA002_IR_Swap/ExpectedOutput/, from which the engine packages were verified.

The report rows are the report instance's own record: its state walk and the output location. Nothing parses the CSVs, because the report asks no question of them.

None. This report computes no measure and carries no Greek set; it exists to exercise the pipeline. A run of it that produces the three CSV artefacts and a completed instance has succeeded.

The run document is the one inside the prepared archive, submitted verbatim. Nothing in the report authors or rewrites it, which is the point of the substitute: the run's input is exactly what a person would hand the engine.

<Analytics>
  <Analytic type="npv">
    <Parameter name="active">Y</Parameter>
    <Parameter name="baseCurrency">EUR</Parameter>
    <Parameter name="outputFileName">npv.csv</Parameter>
  </Analytic>
  <Analytic type="cashflow">
    <Parameter name="active">Y</Parameter>
    <Parameter name="outputFileName">flows.csv</Parameter>
  </Analytic>
  <Analytic type="curves">
    <Parameter name="active">Y</Parameter>
    <Parameter name="configuration">default</Parameter>
    <Parameter name="grid">240,1M</Parameter>
    <Parameter name="outputFileName">curves.csv</Parameter>
  </Analytic>
</Analytics>

The definition states the rows below and nothing else. Every other parameter a production report would carry is irrelevant, because the run pre-processes nothing and post-processes nothing.

This report is the test coverage. It substitutes pre-processing and ignores post-processing, so the chain it runs is four steps rather than seven: resolve_prepared_input, submit_compute, ignore_compute_results, finalise. What it deliberately does not cover is the two substituted phases and the parsing of engine results.

The story that owns it is Run an example bundle through the compute grid from a report definition; the analysis behind it is What a no-preprocessing report run requires.

• Reports — the inventory this page is listed in.
• Headline Position — the report a reader moves to once this one is green.
• Report execution — the workflow that runs it.
• Report execution substitution — what a substituted run proves and what it leaves untested.
• Report compute handoff — the submit contract this report exercises.
• Exercise the workflow engine end to end
• Run a full report: scheduling, reporting and compute

• Open Source Risk Engine — the engine that runs the prepared bundle.