head 1.3; access; symbols pkgsrc-2026Q1:1.3.0.22 pkgsrc-2026Q1-base:1.3 pkgsrc-2025Q4:1.3.0.20 pkgsrc-2025Q4-base:1.3 pkgsrc-2025Q3:1.3.0.18 pkgsrc-2025Q3-base:1.3 pkgsrc-2025Q2:1.3.0.16 pkgsrc-2025Q2-base:1.3 pkgsrc-2025Q1:1.3.0.14 pkgsrc-2025Q1-base:1.3 pkgsrc-2024Q4:1.3.0.12 pkgsrc-2024Q4-base:1.3 pkgsrc-2024Q3:1.3.0.10 pkgsrc-2024Q3-base:1.3 pkgsrc-2024Q2:1.3.0.8 pkgsrc-2024Q2-base:1.3 pkgsrc-2024Q1:1.3.0.6 pkgsrc-2024Q1-base:1.3 pkgsrc-2023Q4:1.3.0.4 pkgsrc-2023Q4-base:1.3 pkgsrc-2023Q3:1.3.0.2 pkgsrc-2023Q3-base:1.3 pkgsrc-2023Q2:1.2.0.4 pkgsrc-2023Q2-base:1.2 pkgsrc-2023Q1:1.2.0.2 pkgsrc-2023Q1-base:1.2; locks; strict; comment @# @; 1.3 date 2023.08.28.10.34.02; author adam; state Exp; branches; next 1.2; commitid Qvcvmncue7M9AxCE; 1.2 date 2023.01.28.19.47.55; author he; state Exp; branches; next 1.1; commitid ZiIHzlsUDZWvplbE; 1.1 date 2023.01.28.17.58.05; author he; state Exp; branches; next ; commitid bXO7AFQxl1R3RkbE; desc @@ 1.3 log @py-pandas: updated to 2.0.3 2.0.3 Fixed regressions Bug in Timestamp.weekday`() was returning incorrect results before '0000-02-29' (GH53738) Fixed performance regression in merging on datetime-like columns (GH53231) Fixed regression when DataFrame.to_string() creates extra space for string dtypes (GH52690) Bug fixes Bug in DataFrame.convert_dtype() and Series.convert_dtype() when trying to convert ArrowDtype with dtype_backend="nullable_numpy" (GH53648) Bug in RangeIndex.union() when using sort=True with another RangeIndex (GH53490) Bug in Series.reindex() when expanding a non-nanosecond datetime or timedelta Series would not fill with NaT correctly (GH53497) Bug in read_csv() when defining dtype with bool[pyarrow] for the "c" and "python" engines (GH53390) Bug in Series.str.split() and Series.str.rsplit() with expand=True for ArrowDtype with pyarrow.string (GH53532) Bug in indexing methods (e.g. DataFrame.__getitem__()) where taking the entire DataFrame/Series would raise an OverflowError when Copy on Write was enabled and the length of the array was over the maximum size a 32-bit integer can hold (GH53616) Bug when constructing a DataFrame with columns of an ArrowDtype with a pyarrow.dictionary type that reindexes the data (GH53617) Bug when indexing a DataFrame or Series with an Index with a timestamp ArrowDtype would raise an AttributeError (GH53644) 2.0.2 Fixed regressions Fixed performance regression in GroupBy.apply() (GH53195) Fixed regression in merge() on Windows when dtype is np.intc (GH52451) Fixed regression in read_sql() dropping columns with duplicated column names (GH53117) Fixed regression in DataFrame.loc() losing MultiIndex name when enlarging object (GH53053) Fixed regression in DataFrame.to_string() printing a backslash at the end of the first row of data, instead of headers, when the DataFrame doesn’t fit the line width (GH53054) Fixed regression in MultiIndex.join() returning levels in wrong order (GH53093) Bug fixes Bug in arrays.ArrowExtensionArray incorrectly assigning dict instead of list for .type with pyarrow.map_ and raising a NotImplementedError with pyarrow.struct (GH53328) Bug in api.interchange.from_dataframe() was raising IndexError on empty categorical data (GH53077) Bug in api.interchange.from_dataframe() was returning DataFrame’s of incorrect sizes when called on slices (GH52824) Bug in api.interchange.from_dataframe() was unnecessarily raising on bitmasks (GH49888) Bug in merge() when merging on datetime columns on different resolutions (GH53200) Bug in read_csv() raising OverflowError for engine="pyarrow" and parse_dates set (GH53295) Bug in to_datetime() was inferring format to contain "%H" instead of "%I" if date contained “AM” / “PM” tokens (GH53147) Bug in DataFrame.convert_dtypes() ignores convert_* keywords when set to False dtype_backend="pyarrow" (GH52872) Bug in DataFrame.convert_dtypes() losing timezone for tz-aware dtypes and dtype_backend="pyarrow" (GH53382) Bug in DataFrame.sort_values() raising for PyArrow dictionary dtype (GH53232) Bug in Series.describe() treating pyarrow-backed timestamps and timedeltas as categorical data (GH53001) Bug in Series.rename() not making a lazy copy when Copy-on-Write is enabled when a scalar is passed to it (GH52450) Bug in pd.array() raising for NumPy array and pa.large_string or pa.large_binary (GH52590) Bug in DataFrame.__getitem__() not preserving dtypes for MultiIndex partial keys (GH51895) 2.0.1 Fixed regressions Fixed regression for subclassed Series when constructing from a dictionary (GH52445) Fixed regression in SeriesGroupBy.agg() failing when grouping with categorical data, multiple groupings, as_index=False, and a list of aggregations (GH52760) Fixed regression in DataFrame.pivot() changing Index name of input object (GH52629) Fixed regression in DataFrame.resample() raising on a DataFrame with no columns (GH52484) Fixed regression in DataFrame.sort_values() not resetting index when DataFrame is already sorted and ignore_index=True (GH52553) Fixed regression in MultiIndex.isin() raising TypeError for Generator (GH52568) Fixed regression in Series.describe() showing RuntimeWarning for extension dtype Series with one element (GH52515) Fixed regression when adding a new column to a DataFrame when the DataFrame.columns was a RangeIndex and the new key was hashable but not a scalar (GH52652) Bug fixes Bug in Series.dt.days that would overflow int32 number of days (GH52391) Bug in arrays.DatetimeArray constructor returning an incorrect unit when passed a non-nanosecond numpy datetime array (GH52555) Bug in ArrowExtensionArray with duration dtype overflowing when constructed from data containing numpy NaT (GH52843) Bug in Series.dt.round() when passing a freq of equal or higher resolution compared to the Series would raise a ZeroDivisionError (GH52761) Bug in Series.median() with ArrowDtype returning an approximate median (GH52679) Bug in api.interchange.from_dataframe() was unnecessarily raising on categorical dtypes (GH49889) Bug in api.interchange.from_dataframe() was unnecessarily raising on large string dtypes (GH52795) Bug in pandas.testing.assert_series_equal() where check_dtype=False would still raise for datetime or timedelta types with different resolutions (GH52449) Bug in read_csv() casting PyArrow datetimes to NumPy when dtype_backend="pyarrow" and parse_dates is set causing a performance bottleneck in the process (GH52546) Bug in to_datetime() and to_timedelta() when trying to convert numeric data with a ArrowDtype (GH52425) Bug in to_numeric() with errors='coerce' and dtype_backend='pyarrow' with ArrowDtype data (GH52588) Bug in ArrowDtype.__from_arrow__() not respecting if dtype is explicitly given (GH52533) Bug in DataFrame.describe() not respecting ArrowDtype in include and exclude (GH52570) Bug in DataFrame.max() and related casting different Timestamp resolutions always to nanoseconds (GH52524) Bug in Series.describe() not returning ArrowDtype with pyarrow.float64 type with numeric data (GH52427) Bug in Series.dt.tz_localize() incorrectly localizing timestamps with ArrowDtype (GH52677) Bug in arithmetic between np.datetime64 and np.timedelta64 NaT scalars with units always returning nanosecond resolution (GH52295) Bug in logical and comparison operations between ArrowDtype and numpy masked types (e.g. "boolean") (GH52625) Fixed bug in merge() when merging with ArrowDtype one one and a NumPy dtype on the other side (GH52406) Fixed segfault in Series.to_numpy() with null[pyarrow] dtype (GH52443) Other DataFrame created from empty dicts had columns of dtype object. It is now a RangeIndex (GH52404) Series created from empty dicts had index of dtype object. It is now a RangeIndex (GH52404) Implemented Series.str.split() and Series.str.rsplit() for ArrowDtype with pyarrow.string (GH52401) Implemented most str accessor methods for ArrowDtype with pyarrow.string (GH52401) Supplying a non-integer hashable key that tests False in api.types.is_scalar() now raises a KeyError for RangeIndex.get_loc(), like it does for Index.get_loc(). Previously it raised an InvalidIndexError (GH52652). @ text @$NetBSD: patch-pandas___libs_window_aggregations.pyx,v 1.2 2023/01/28 19:47:55 he Exp $ On NetBSD, signbit and sqrt cannot be imported from "libc", because even though is included, Python.h later includes which in turn #undef's all the relevant definitions from , forcing the use of the std:: variants of those functions. So ... re-do how we import signbit() and sqrt(). Submitted upstream: https://github.com/pandas-dev/pandas/pull/51049 --- pandas/_libs/window/aggregations.pyx.orig 2023-03-16 14:52:19.000000000 +0000 +++ pandas/_libs/window/aggregations.pyx @@@@ -1,10 +1,6 @@@@ # cython: boundscheck=False, wraparound=False, cdivision=True -from libc.math cimport ( - round, - signbit, - sqrt, -) +from libc.math cimport round from libcpp.deque cimport deque from pandas._libs.algos cimport TiebreakEnumType @@@@ -19,6 +15,10 @@@@ from numpy cimport ( ndarray, ) +cdef extern from "" namespace "std": + int signbit(float64_t) nogil + float64_t sqrt(float64_t x) nogil + cnp.import_array() import cython @ 1.2 log @math/py-pandas: note upstream pull request, and remove .orig file. The .orig file would otherwise be installed, cauisng a PLIST mismatch. @ text @d1 1 a1 1 $NetBSD: patch-pandas___libs_window_aggregations.pyx,v 1.1 2023/01/28 17:58:05 he Exp $ d12 1 a12 1 --- pandas/_libs/window/aggregations.pyx.orig 2023-01-19 03:22:10.000000000 +0000 d14 1 a14 1 @@@@ -1,11 +1,8 @@@@ a16 1 cimport cython a22 1 + d26 1 a26 1 @@@@ -20,6 +17,10 @@@@ from numpy cimport ( d36 1 a36 1 from pandas._libs.algos import is_monotonic @ 1.1 log @math/py-pandas: fix aggregations.pyx so that this builds on NetBSD. On NetBSD, signbit and sqrt cannot be imported from "libc", because even though is included, Python.h later includes which in turn #undef's all the relevant definitions from , forcing the use of the std:: variants of those functions. So ... re-do how we import signbit() and sqrt(). @ text @d1 1 a1 1 $NetBSD$ d9 3 @