Address class name changes in R polars 1.0.0. (#21)

The R package polars introduced a few changes in
its class definitions.

The fixes highlighted an issue in rpy2 (conversion
rules for R environment) that will be fixed in rpy2-robjects 3.6.2.
That fix is required.

Author: lgautier

.github/workflows/webpages.yml

@@ -28,7 +28,7 @@ jobs:
         with:
           r-version: ${{ env.R-VERSION }}
       - name: Cache R packages
-        uses: actions/cache@v2
+        uses: actions/cache@v4
         env:
           cache-name: cache-R-packages
         with:

pyproject.toml

@@ -25,7 +25,7 @@ classifiers = [
 ]
 dependencies = [
     "pyarrow",
-    "rpy2 >= 3.5.15"
+    "rpy2-robjects >= 3.6.2"
 ]
 dynamic = ["version"]
 

rpy2_arrow/__init__.py

@@ -1 +1 @@
-__version__ = '0.1.2'
+__version__ = '0.1.3'

rpy2_arrow/polars.py

@@ -19,7 +19,7 @@ def ensure_r_polars() -> types.ModuleType:
         rpack_polars = rpy2.robjects.packages.importr('polars',
                                                       on_conflict='warn')
     rpack_polars_version_info = rpack_polars.__version__.split('.')
-    assert tuple(int(_) for _ in rpack_polars_version_info[:2]) >= (0, 12)
+    assert tuple(int(_) for _ in rpack_polars_version_info[:2]) >= (1, 0)
     return rpack_polars
 
 
@@ -58,7 +58,7 @@ def pypolars_to_rpolars_dataframe(
     rpack_polars = ensure_r_polars()
     # TODO: There appear to be an odd shortcircuiting that requires toggling
     # additional conversion off.
-    with rpy2.robjects.default_converter.context():
+    with rpy2arrow.converter.context():
         return rpack_polars.as_polars_df(r_arrow_table)
 
 
@@ -69,7 +69,19 @@ def rpolar_to_pypolars_dataframe(
     # R polars to R arrow.
     rpack_arrow = ensure_r_arrow()
     ensure_r_polars()
-    r_arrow_table = rpack_arrow.as_arrow_table(dataf)
+    with rpy2.robjects.default_converter.context():
+        r_arrow_table = rpack_arrow.as_arrow_table(dataf)
+    return rarrow_to_pypolars_dataframe(r_arrow_table)
+
+
+def rpolars_env_to_pypolars_dataframe(
+        dataf: rpy2.rinterface.sexp.SexpEnvironment
+) -> polars.DataFrame:
+    # R polars to R arrow.
+    rpack_arrow = ensure_r_arrow()
+    ensure_r_polars()
+    with rpy2.robjects.default_converter.context():
+        r_arrow_table = rpack_arrow.as_arrow_table(dataf)
     return rarrow_to_pypolars_dataframe(r_arrow_table)
 
 
@@ -92,13 +104,15 @@ def rpolar_to_pypolars_dataframe(
 
 converter._rpy2py_nc_map[rpy2.rinterface.SexpEnvironment].update(
     {
-        'Table': rarrow_to_pypolars_dataframe,
+        'polars_data_frame': rpolars_env_to_pypolars_dataframe,
+        'Table': rarrow_to_pypolars_dataframe
     }
 )
 
 converter._rpy2py_nc_map[rpy2.rinterface.SexpExtPtr].update(
     {
-        'RPolarsDataFrame': rpolar_to_pypolars_dataframe,
+        # TODO: is this still needed?
+        'polars_data_frame': rpolar_to_pypolars_dataframe,
     }
 )
 

rpy2_arrow/tests_polars.py

@@ -13,7 +13,7 @@
 R_ASVECTOR = rpy2.rinterface.baseenv['as.vector']
 R_SQBRACKET = rpy2.rinterface.baseenv['[']
 R_LENGTH = rpy2.rinterface.baseenv['length']
-R_EQUAL = rpy2.rinterface.baseenv['==']
+R_IDENTICAL = rpy2.rinterface.baseenv['identical']
 
 
 def _cmp_simple(v1, v2):
@@ -97,7 +97,7 @@ def test_rpolar_to_pypolars_dataframe(self):
             ([1, 2], polars.Int64, 'Int64', _cmp_simple),  # Fails.
             ([1.1, 2.1], polars.Float32, 'Float32', _cmp_float),
             ([1.1, 2.1], polars.Float64, 'Float64', _cmp_float),
-            (['wx', 'yz'], polars.Utf8, 'Utf8', _cmp_simple),
+            (['wx', 'yz'], polars.Utf8, 'String', _cmp_simple),
             (['wx', 'yz', 'wx'], polars.Categorical,
              'Categorical', _cmp_simple)
         ])
@@ -107,7 +107,7 @@ def test_converter_py2rpy_dataframe(self, values, dtype, rpotype, cmp):
         with rpy2polars.converter.context():
             globalenv['podataf'] = podataf
         r_podataf = globalenv['podataf']
-        assert tuple(r_podataf.rclass) == ('RPolarsDataFrame',)
+        assert tuple(r_podataf.rclass) == ('polars_data_frame', 'polars_object')
 
         assert tuple(
             R_DOLLAR(r_podataf, 'schema').names
@@ -116,21 +116,18 @@ def test_converter_py2rpy_dataframe(self, values, dtype, rpotype, cmp):
             R_DOLLAR(r_podataf, 'schema'), 1
         )[0]
         type_in_library = R_DOLLAR(
-            R_DOLLAR(
-                getattr(
-                    rpy2polars.rpack_polars, 'pl'
-                ),
-                'dtypes'
+            getattr(
+                rpy2polars.rpack_polars, 'pl'
             ),
             rpotype
         )
-        assert R_EQUAL(
-            field,
+        assert R_IDENTICAL(
+            field.rclass,
             # `r-polars` is a bit inconsistent in the way it declares
             # types. Some are R functions while others are non-callable
             # objects.
             type_in_library() if 'function' in type_in_library.rclass
-            else type_in_library
+            else type_in_library.rclass
         )
 
     @pytest.mark.parametrize(
@@ -173,6 +170,6 @@ def test_rpl_to_pl(self, rstr, cls):
 
     def test_pl_to_rpl(self):
         plobj = polars.DataFrame({'a': [1, 2, 3]})
-        cls = rpy2.robjects.ExternalPointer
+        cls = rpy2.robjects.environments.Environment
         rplobj = rpy2polars.pl_to_rpl(plobj)
         assert isinstance(rplobj, cls)