src.dackar.RCA.pm_compliance.vocabulary_loader
==============================================

.. py:module:: src.dackar.RCA.pm_compliance.vocabulary_loader

.. autoapi-nested-parse::

   PMVocabularyLoader — load health-status vocabulary from the DACKAR data directory.

   Drives ``analyze_degradation`` and ``compute_pm_found_defect_rate`` in
   ``effectiveness_analyzer``.  Falls back to hardcoded stems when *data_dir* is
   ``None`` or the CSV files are absent, so callers without the data path are
   unaffected.



Attributes
----------

.. autoapisummary::

   src.dackar.RCA.pm_compliance.vocabulary_loader._COLS
   src.dackar.RCA.pm_compliance.vocabulary_loader._NEG_FILE
   src.dackar.RCA.pm_compliance.vocabulary_loader._POS_FILE
   src.dackar.RCA.pm_compliance.vocabulary_loader._NEU_FILE
   src.dackar.RCA.pm_compliance.vocabulary_loader._FALLBACK


Classes
-------

.. autoapisummary::

   src.dackar.RCA.pm_compliance.vocabulary_loader.PMVocabularyLoader


Functions
---------

.. autoapisummary::

   src.dackar.RCA.pm_compliance.vocabulary_loader._read_terms
   src.dackar.RCA.pm_compliance.vocabulary_loader._term_matches
   src.dackar.RCA.pm_compliance.vocabulary_loader.matches_any


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

.. py:data:: _COLS
   :type:  Sequence[str]
   :value: ('Nouns', 'Verbs', 'Adjectives')


.. py:data:: _NEG_FILE
   :value: 'health_status_keywords_negative.csv'


.. py:data:: _POS_FILE
   :value: 'health_status_keywords_positive.csv'


.. py:data:: _NEU_FILE
   :value: 'health_status_keywords_neutral.csv'


.. py:data:: _FALLBACK
   :type:  Dict[str, FrozenSet[str]]

.. py:function:: _read_terms(path, cols)

   Return lowercased, stripped terms from *cols* of a keyword CSV.


.. py:function:: _term_matches(term, blob)

   Return True if *term* appears in *blob* at a leading word boundary.

   Multi-word phrases (containing spaces) are matched as substrings.
   Single-word terms use a left-side word boundary so that the vocabulary
   term is matched as a prefix of a word — "crack" matches "crack", "cracks",
   "cracked", "cracking" and "leak" matches "leakage", but neither matches in
   the middle of an unrelated word (e.g., "crack" does not match "firecracker"
   because 'c' in "cracker" has no leading word boundary).


.. py:function:: matches_any(blob, terms)

   Return True if any term in *terms* matches within *blob*.


.. py:class:: PMVocabularyLoader

   Load and cache health-status vocabulary keyed by *data_dir* path.

   Usage::

       vocab = PMVocabularyLoader.load(cfg.data_dir)
       is_degrading = matches_any(blob, vocab["degrading"])


   .. py:attribute:: _cache
      :type:  Dict[str, Dict[str, FrozenSet[str]]]


   .. py:method:: load(data_dir)
      :classmethod:


      Return ``{"degrading": ..., "improving": ...}`` vocabulary sets.

      When *data_dir* is ``None``, returns the hardcoded fallback stems so
      existing behaviour is fully preserved.



   .. py:method:: clear_cache()
      :classmethod:


      Invalidate the vocabulary cache (useful in tests).