From c5ece93665ced51458d00af06a95b533703e7ed3 Mon Sep 17 00:00:00 2001
From: Marigold <mojmir.vinkler@gmail.com>
Date: Thu, 14 Nov 2024 10:07:32 +0100
Subject: [PATCH] wip

---
 lib/repack/owid/repack/__init__.py | 29 ++---------------------------
 lib/repack/tests/test_repack.py    |  8 ++++++--
 2 files changed, 8 insertions(+), 29 deletions(-)

diff --git a/lib/repack/owid/repack/__init__.py b/lib/repack/owid/repack/__init__.py
index 2eb4c0ca7e2..38029175a0f 100644
--- a/lib/repack/owid/repack/__init__.py
+++ b/lib/repack/owid/repack/__init__.py
@@ -76,7 +76,6 @@ def repack_series(s: pd.Series) -> pd.Series:
 
 def to_int(s: pd.Series) -> pd.Series:
     # values could be integers or strings
-    # s = _to_float64(s)
     s = s.astype("Float64")
     v = s.astype("Int64")
 
@@ -93,7 +92,6 @@ def shrink_integer(s: pd.Series) -> pd.Series:
     Take an Int64 series and make it as small as possible.
     """
     assert s.dtype == "Int64"
-    # assert s.dtype.name.replace("[pyarrow]", "") in ("Int64", "int64", "UInt64", "uint64"), s.dtype
 
     if s.isnull().all():
         # shrink all NaNs to Int8
@@ -114,17 +112,6 @@ def shrink_integer(s: pd.Series) -> pd.Series:
     return s
 
 
-# def _to_float64(s: pd.Series) -> pd.Series:
-#     """Convert pandas series to float if possible. It handles object types and string types as well."""
-#     # Directly convert to float64[pyarrow]
-#     if "float" in s.dtype.name.lower():
-#         return s.astype("Float64")
-#     else:
-#         # Convert object types to float first and then to Float64
-#         # TODO: is this necessary?
-#         return s.astype("Float64")
-
-
 def to_float(s: pd.Series) -> pd.Series:
     return shrink_float(s.astype("Float64"))
 
@@ -154,10 +141,7 @@ def to_category(s: pd.Series) -> pd.Series:
     return s.astype("category")
 
 
-# TODO: can it be simplified?
-def series_eq(
-    lhs: pd.Series, rhs: pd.Series, cast: Optional[Any] = None, rtol: float = 1e-5, atol: float = 1e-8
-) -> bool:
+def series_eq(lhs: pd.Series, rhs: pd.Series, rtol: float = 1e-5, atol: float = 1e-8) -> bool:
     """
     Check that series are equal, but unlike normal floating point checks where
     NaN != NaN, we want missing or null values to be reported as equal to each
@@ -168,16 +152,7 @@ def series_eq(
     if len(lhs) != len(rhs):
         return False
 
-    # improve performance by calling native astype method
-    if cast == float:
-        func = lambda s: s.astype(float)  # noqa: E731
-    elif cast is None:
-        func = lambda s: s  # noqa: E731
-    else:
-        # NOTE: this would be extremely slow in practice
-        func = lambda s: s.apply(cast)  # noqa: E731
-
-    return np.allclose(func(lhs), func(rhs), rtol=rtol, atol=atol, equal_nan=True)
+    return np.allclose(lhs, rhs, rtol=rtol, atol=atol, equal_nan=True)
 
 
 def _safe_dtype(dtype: Any) -> str:
diff --git a/lib/repack/tests/test_repack.py b/lib/repack/tests/test_repack.py
index 3dbba13b3a5..b09ad955cda 100644
--- a/lib/repack/tests/test_repack.py
+++ b/lib/repack/tests/test_repack.py
@@ -228,11 +228,15 @@ def test_repack_float64_all_nans():
 def test_series_eq():
     a = pd.Series([1, np.nan], dtype="float64")
     b = pd.Series([2, np.nan], dtype="float64")
-    assert not repack.series_eq(a, b, cast=float)
+    assert not repack.series_eq(a, b)
 
     a = pd.Series([1, np.nan], dtype="float64")
     b = pd.Series([1, np.nan], dtype="float64")
-    assert repack.series_eq(a, b, cast=float)
+    assert repack.series_eq(a, b)
+
+    a = pd.Series([1, np.nan], dtype="float64")
+    b = pd.Series([1, np.nan], dtype="float64").astype("Float64")
+    assert repack.series_eq(a, b)
 
 
 def test_repack_object_np_str():