loader
======

.. py:module:: loader

.. autoapi-nested-parse::

   Load RCA artifact bundles from a full-result JSON file or a fixtures directory.

   Run the Streamlit app from this directory so ``import loader`` resolves::

       cd DACKAR/src/dackar/RCA/viz && streamlit run app.py



Attributes
----------

.. autoapisummary::

   loader.JsonDict
   loader._FIXTURE_FILE_MAP


Functions
---------

.. autoapisummary::

   loader._allowed_path
   loader._read_json
   loader.detect_input_mode
   loader.load_from_full_result
   loader.load_from_fixtures_dir
   loader.merge_bundles
   loader.load_artifacts
   loader.load_pre_refine_causality
   loader.list_bundle_keys


Module Contents
---------------

.. py:data:: JsonDict

.. py:data:: _FIXTURE_FILE_MAP
   :type:  Dict[str, str]

.. py:function:: _allowed_path(path)

.. py:function:: _read_json(path)

.. py:function:: detect_input_mode(path)

   Classify a path as a full-result JSON file or a fixtures/run directory.

   @ In, path, str, filesystem path to a ``.json`` file or a directory
   @ Out, mode, str, ``"fixtures_dir"`` for a directory, ``"full_result"`` for
       a ``.json`` file


.. py:function:: load_from_full_result(path)

   Load a single JSON object (e.g. v32_full_result.json). Keys pass through unchanged.


.. py:function:: load_from_fixtures_dir(directory)

   Merge every ``*.json`` in a directory into one bundle, keyed by filename.

   The directory may be a hand-built fixtures folder or a real orchestrator run
   folder (``ArtifactStore`` writes one ``<artifact_name>.json`` per artifact),
   so scanning the directory surfaces whatever artifacts are actually present
   rather than a fixed allow-list. ``_FIXTURE_FILE_MAP`` is consulted as an
   alias layer: a basename it names maps to that key, otherwise the filename
   stem is the key. Missing files are simply absent; never raises for them.


.. py:function:: merge_bundles(primary, supplemental)

   Return *primary* with any keys it lacks filled in from *supplemental*.

   Primary wins on every shared key, so a complete run bundle is never
   overwritten; the supplement only contributes artifacts the primary is
   missing (e.g. the raw ``event`` / ``telemetry_summary`` inputs that a
   ``v32_full_result.json`` does not carry, or the generated artifacts a
   hand-built fixtures folder lacks). This lets one view combine the data
   provided as input with the data produced by every pipeline stage.


.. py:function:: load_artifacts(path)

   Auto-detect file vs directory and load.


.. py:function:: load_pre_refine_causality(path)

   Load optional pre-refine causality artifact.

   Accepts either a bare ``causality_candidates`` object (has ``candidates``)
   or a full bundle (uses ``causality_candidates`` key if present).


.. py:function:: list_bundle_keys(bundle)

   Return the bundle's artifact keys, sorted, excluding load-error markers.

   @ In, bundle, dict, a loaded bundle (may contain ``<key>__load_error`` markers)
   @ Out, keys, list, sorted artifact keys with the error markers removed