Cleanup from self-review: redundant abspath, missing DeprecationWarning, logger inconsistency, eager __eq__

mhctools/base_predictor.py

@@ -11,6 +11,7 @@
 # limitations under the License.
 
 import logging
+import warnings
 from collections import defaultdict
 
 from typechecks import require_iterable_of
@@ -207,6 +208,10 @@ def predict_peptides_dataframe(self, peptides):
         ``predictor_version``, ``kind``, ``value`` and used
         ``prediction_method_name`` instead of ``predictor_name`` — see #193.
         """
+        warnings.warn(
+            "predict_peptides_dataframe is deprecated; use predict_dataframe()",
+            DeprecationWarning,
+            stacklevel=2)
         return self.predict_dataframe(peptides)
 
     def _check_peptide_lengths(self, peptide_lengths=None):

mhctools/cli/args.py

@@ -96,9 +96,15 @@ def __getattr__(self, item):
         return getattr(self._resolve(), item)
 
     def __eq__(self, other):
+        if other is self:
+            return True
         if isinstance(other, _LazyPredictor):
             return (self._module_name, self._class_name) == (
                 other._module_name, other._class_name)
+        # Only classes can meaningfully equal a lazy-loaded class reference.
+        # Bail out without triggering the import for anything else.
+        if not isinstance(other, type):
+            return NotImplemented
         return self._resolve() is other
 
     def __hash__(self):

mhctools/mhcflurry.py

@@ -38,7 +38,7 @@ def _normalize_models_path(models_path):
     """
     if models_path is None:
         return None
-    return os.path.realpath(os.path.abspath(os.path.expanduser(models_path)))
+    return os.path.realpath(os.path.expanduser(models_path))
 
 
 class MHCflurry(BasePredictor):
@@ -84,7 +84,7 @@ def __init__(
             cache_key = ("presentation", _normalize_models_path(models_path))
             if cache_key not in _model_cache:
                 if models_path:
-                    logging.info(
+                    logger.info(
                         "Loading MHCflurry models from %s", models_path)
                     _model_cache[cache_key] = \
                         Class1PresentationPredictor.load(models_path)
@@ -260,7 +260,7 @@ def __init__(
             cache_key = ("affinity", _normalize_models_path(models_path))
             if cache_key not in _model_cache:
                 if models_path:
-                    logging.info(
+                    logger.info(
                         "Loading MHCflurry models from %s", models_path)
                     _model_cache[cache_key] = \
                         Class1AffinityPredictor.load(models_path)

tests/test_dataframe_schema_parity.py

@@ -11,15 +11,27 @@
 treat rows from either path uniformly.
 """
 
+import warnings
+
+import pytest
+
 from mhctools import RandomBindingPredictor
 from mhctools.pred import COLUMNS
 
 
+def _peptides_df(p, peptides):
+    # predict_peptides_dataframe emits DeprecationWarning; silence here so
+    # the fixture doesn't pollute every assertion.
+    with warnings.catch_warnings():
+        warnings.simplefilter("ignore", DeprecationWarning)
+        return p.predict_peptides_dataframe(peptides)
+
+
 def test_peptides_and_proteins_dataframes_share_schema():
     p = RandomBindingPredictor(
         alleles=["HLA-A*02:01"], default_peptide_lengths=[9],
     )
-    df_peptides = p.predict_peptides_dataframe(["SIINFEKLA"])
+    df_peptides = _peptides_df(p, ["SIINFEKLA"])
     df_proteins = p.predict_proteins_dataframe({"src": "MASIINFEKLA"})
 
     assert list(df_peptides.columns) == list(COLUMNS), (
@@ -35,7 +47,7 @@ def test_peptides_dataframe_has_predictor_identity_columns():
     """Downstream (e.g. topiary CachedPredictor) needs a stable
     (predictor_name, predictor_version) identity on every row."""
     p = RandomBindingPredictor(alleles=["HLA-A*02:01"], default_peptide_lengths=[9])
-    df = p.predict_peptides_dataframe(["SIINFEKLA"])
+    df = _peptides_df(p, ["SIINFEKLA"])
     assert "predictor_name" in df.columns
     assert "predictor_version" in df.columns
     assert "kind" in df.columns
@@ -44,3 +56,11 @@ def test_peptides_dataframe_has_predictor_identity_columns():
     assert "prediction_method_name" not in df.columns
     # canonical name is populated (RandomBindingPredictor sets it)
     assert df["predictor_name"].iloc[0] != ""
+
+
+def test_peptides_dataframe_emits_deprecation_warning():
+    """Callers on the legacy path should get a runtime signal to migrate
+    to predict_dataframe()."""
+    p = RandomBindingPredictor(alleles=["HLA-A*02:01"], default_peptide_lengths=[9])
+    with pytest.warns(DeprecationWarning, match="predict_peptides_dataframe"):
+        p.predict_peptides_dataframe(["SIINFEKLA"])