Fix task creation

[ladinesa]

Purpose

Fix reported issue with recursive sections in workflow tasks.

Scope

Included

Out of scope

Context & Links

"outputs": [
      {
        "name": "MD results",
        "section": "#/workflow2/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Geometry optimization results",
        "section": "#/workflow2/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/tasks/0/results"
      },
      {
        "name": "Outputs",
        "section": "#/data/outputs/10"

Reviewer Notes

Anything reviewers should pay special attention to? Trade-offs, known limitations, or areas of uncertainty.

Status

• Draft / In progress
• Ready for review
• Blocked (explain):

Breaking Changes

• None
• (explain):

Dependencies / Blockers

• None
• Open design questions (explain):
• External dependencies (explain):

Testing / Validation

How was this change validated?

• None (explain):
• Unit tests
• Integration tests
• Manual testing (explain):
  Run gromacs tests data.

[Bernadette-Mohr]

I have tested the fix, and went through all the auxillary parsers and the gromacs parser (which is using the mapping parser).

Unfortunately, the situation has not changed at all.

Here is the archive I have generated: https://drive.google.com/file/d/1cZsXPK7SvIF0YxT61GU7x-ndcclk987N/view?usp=sharing
with the gromacs parser at it's current state: FAIRmat-NFDI/nomad-parser-plugins-simulation#156
The test data is still tests/data/gromacs/water/reference_s.log.

I HAVE AN IDEA: it might be an artefact from the Gromacs schema. I will look into it.

[Bernadette-Mohr]

I HAVE AN IDEA: it might be an artefact from the Gromacs schema. I will look into it.

Let me describe what I am seeing:

• My test data is of a production run (integrator md), so the workflow should be recognized as MolecularDynamics, I should not be seeing GeometryOptimization in the archive at all.
• workflow2 is correctly MolecularDynamics, but tasks/0 inside it is GeometryOptimization. This means map_tasks is adding a GeometryOptimization instance as a task to MolecularDynamics.
• In the Gromacs schema (https://github.com/FAIRmat-NFDI/nomad-parser-plugins-simulation/tree/gromacs-parser_results), these two annotations have LOG_KEY annotation '@':

397 add_mapping_annotation(geometry_optimization.GeometryOptimization.m_def, LOG_KEY, '@')
399 add_mapping_annotation(molecular_dynamics.MolecularDynamics.m_def, LOG_KEY, '@')

• When processing MolecularDynamics, the third lookup level in build_section_mapper finds GeometryOptimization as a valid Task dependency with LOG_KEY and adds it to tasks. Same recursion also happens inside only MolecularDynamics or GeometryOptimization — they find themselves as a Task dependency and nest again.

My current fix follows the pattern established in the FHI-aims parser (comment in fhiaims.py, l. 198). Annotating SimulationWorkflow.tasks with a Level 1 annotation (.tasks path) for both LOG_KEY and EDR_KEY prevents Level 3 from running for the tasks SubSection.