# We need to break some cycles with optional dependencies for bootstrapping;

# given that a conditional is needed, we take the opportunity to omit as many

# optional dependencies as possible for bootstrapping.

%bcond_without bootstrap

# When not bootstrapping, run tests?

%bcond_without tests

%{?with_bootstrap:%undefine with_tests}

# When running tests, run ones that are marked as slow?

%bcond_without slow_tests

# When running tests, run ones that cannot be run in parallel?

%bcond_without single_tests

Name:           python-pandas

Version:        2.2.1

Release:        7%{?dist}

Summary:        Python library providing high-performance data analysis tools

# Drop support for i686 in preparation for `libarrow`

# https://bugzilla.redhat.com/show_bug.cgi?id=2263999

ExcludeArch:    %{ix86}

# The entire source is BSD-3-Clause and covered by LICENSE, except:

#

# - pandas/util/version/__init__.py is (Apache-2.0 OR BSD-2-Clause): see

#   LICENSES/PACKAGING_LICENSE

# - pandas/_libs/src/headers/portable.h is (BSD-3-Clause AND MIT), because it

#   contains some trivial content under the overall BSD-3-Clause license but

#   also some macros from MUSL libc under the MIT license: see

#   LICENSES/MUSL_LICENSE

# - pandas/_libs/src/parser/tokenizer.c is (BSD-3-Clause AND Python-2.0.1): see

#   LICENSES/PSF_LICENSE

# - pandas/io/sas/sas7bdat.py is (BSD-3-Clause and MIT), because it is mostly

#   under the overall BSD-3-Clause license but is also based on

#   https://bitbucket.org/jaredhobbs/sas7bdat: see LICENSES/SAS7BDAT_LICENSE

# - pandas/core/accessor.py is (BSD-3-Clause AND Apache-2.0), because it is

#   partially under the overall BSD-3-Clause license but is also based on

#   xarray: see LICENSES/XARRAY_LICENSE

# - pandas/_libs/src/klib/khash.h is MIT: see LICENSES/KLIB_LICENSE

# - pandas/_libs/window/aggregations.pyx is (BSD-3-Clause AND BSD-2-Clause):

#   see “Bottleneck license” in LICENSES/OTHER

#

# In the python3-pandas+test subpackage:

#

# - pandas/tests/io/data/spss/*.sav are MIT: see LICENSES/HAVEN_LICENSE and

#   LICENSES/HAVEN_MIT

# - pandas/tests/window/test_rolling.py is (BSD-3-Clause AND BSD-2-Clause)

#   since test_rolling_std_neg_sqrt is from Bottleneck: see “Bottleneck license”

#   in LICENSES/OTHER

#

# Additionally:

#

# - pandas/_libs/tslibs/parsing.pyx is BSD-3-Clause rather than

#   (BSD-3-Clause AND (BSD-3-Clause OR Apache-2.0)), because it appears that at

#   least some trivial content in the code copied from dateutil in the

#   dateutil_parse() function (as of

#   https://github.com/dateutil/dateutil/pull/732) is by dateutil contributors

#   who have not agreed to re-license their previously submitted code: see

#   LICENSES/DATEUTIL_LICENSE.

# - LICENSES/OTHER suggests that some code may be derived from

#   google-api-python-client under Apache-2.0, but a search for attribution

#   comments did not turn up anything specific

# - pandas/_libs/tslibs/src/datetime/np_datetime.{h,c} are still BSD-3-Clause,

#   but see also LICENSES/NUMPY_LICENSE

# - pandas/io/clipboard/ is still BSD-3-Clause, but see also “Pyperclip v1.3

#   license” in LICENSES/OTHER

# - pandas/_testing/__init__.py is still BSD-3-Clause, but see also

#   LICENSES/SCIPY_LICENSE

# - pandas/_libs/src/ujson/lib/ is still BSD-3-Clause, but under

#   LICENSES/ULTRAJSON_LICENSE

#

# Additionally, the following are not packaged and so do not affect the overall

# License field:

#

# - scripts/no_bool_in_generic.py is MIT: see LICENSES/PYUPGRADE_LICENSE

License:        BSD-3-Clause AND (Apache-2.0 OR BSD-2-Clause) AND (BSD-3-Clause AND Apache-2.0) AND (BSD-3-Clause AND MIT) AND (BSD-3-Clause AND Python-2.0.1) AND MIT AND (BSD-3-Clause AND BSD-2-Clause)

URL:            https://pandas.pydata.org/

# The GitHub archive contains tests; the PyPI sdist does not.

Source0:        https://github.com/pandas-dev/pandas/archive/v%{version}/pandas-%{version}.tar.gz

# https://github.com/pandas-dev/pandas/pull/57389

Patch:          0001-TST-Ensure-Matplotlib-is-always-cleaned-up.patch

# https://github.com/pandas-dev/pandas/pull/57391

Patch:          0002-Fix-evaluations-on-Python-3.12.patch

# Fix big-endian issues:

# https://github.com/pandas-dev/pandas/pull/57393

Patch:          0003-TST-Fix-IntervalIndex-constructor-tests-on-big-endia.patch

# https://github.com/pandas-dev/pandas/issues/57373

# https://github.com/pandas-dev/pandas/pull/57394

Patch:          0004-TST-Fix-test_str_encode-on-big-endian-machines.patch

# https://github.com/pandas-dev/pandas/pull/57397

Patch:          0005-TST-Add-missing-skips-for-unavailable-pyarrow.patch

# https://github.com/pandas-dev/pandas/pull/57548

Patch:          0006-Fix-accidental-loss-of-precision-for-to_datetime-str.patch

# Python 3.13 support https://github.com/pandas-dev/pandas/pull/59065

Patch:          0007-Fix-Python-3.13-test-failures.patch

%global _description %{expand:

pandas is an open source, BSD-licensed library providing

high-performance, easy-to-use data structures and data

analysis tools for the Python programming language.}

%description %_description

%package -n python3-pandas

Summary:        %{summary}

# pandas/_libs/window/aggregations.pyx:

#

#   Moving maximum / minimum code taken from Bottleneck under the terms

#   of its Simplified BSD license

#   https://github.com/pydata/bottleneck

#

# These snippets are extracted from Bottleneck’s internals and cannot be

# replaced by calling the public Bottleneck API, so there is no reasonable path

# to unbundling.

Provides:       bundled(python3dist(bottleneck))

# pandas/_libs/tslibs/parsing.pyx:

#

# Contains a routine, dateutil_parse(), from an unspecified version of dateutil

#

# Cannot be unbundled because the function is forked and compiled as Cython

Provides:       bundled(python3dist(dateutil))

# pandas/_libs/src/klib/khash.h:

#

# From klib (https://github.com/attractivechaos/klib); it is not practical to

# package all of klib separately because it is designed as a copylib, and many

# of its components are not header-only.

Provides:       bundled(klib-khash) = 0.2.6

# pandas/_libs/src/headers/portable.h:

#

# Contains several preprocessor macros from an unspecified version of MUSL libc

#

# Cannot be unbundled because the macros are not directly exposed in the libc

Provides:       bundled(musl-libc)

# pandas/_libs/tslibs/src/datetime/np_datetime.{h,c}:

#

# Derived from Numpy 1.7

#

# Cannot be unbundled because the routines are forked.

Provides:       bundled(python3dist(numpy)) = 1.7

# pandas/util/version/__init__.py:

#

# Vendored from https://github.com/pypa/packaging/blob/main/packaging/_structures.py

# and https://github.com/pypa/packaging/blob/main/packaging/_structures.py

# changeset ae891fd74d6dd4c6063bb04f2faeadaac6fc6313

# 04/30/2021

#

# Cannot be (reasonably) unbundled because the vendored file is not part of

# packaging’s public API.

Provides:       bundled(python3dist(packaging)) = 20.10.dev0^20210430gitae891fd

# pandas/io/clipboard/:

#

# In https://github.com/pandas-dev/pandas/pull/28471, upstream considered and

# rejected the idea of de-vendoring pyperclip. Furthermore,

# https://github.com/pandas-dev/pandas/commits/main/pandas/io/clipboard and

# https://github.com/pandas-dev/pandas/commits/main/pandas/io/clipboard/__init__.py

# show that the vendored library has accrued Pandas-specific changes.

#

# Version number from:

# https://github.com/pandas-dev/pandas/pull/28471/commits/33cd2d72e0c007c460e59105efda9211441b2ce4

# “Updated internal pyperclip 1.5.27 -> 1.7.0”

Provides:       bundled(python3dist(pyperclip)) = 1.7.0

# pandas/_libs/src/parser/tokenizer.c:

#

# Combines some elements from Python's built-in csv module and Warren

# Weckesser's textreader project on GitHub.

#

# Elements from these are both forked and cannot be unbundled. The textreader

# project is a Python extension but is not on PyPI, and is not the same as

# python3dist(textreader).

Provides:       bundled(python3-libs)

Provides:       bundled(textreader)

# scripts/no_bool_in_generic.py:

#

# The function `visit` is adapted from a function by the same name in pyupgrade:

# https://github.com/asottile/pyupgrade/blob/5495a248f2165941c5d3b82ac3226ba7ad1fa59d/pyupgrade/_data.py#L70-L113

#

# Not packaged (pre-commit hook) therefore not bundled

# Provides:       bundled(python3dist(pyupgrade)) = 2.11.0^20210201git5495a24

# pandas/io/sas/sas7bdat.py

#

# Based on code written by Jared Hobbs:

#   https://bitbucket.org/jaredhobbs/sas7bdat

#

# Cannot be unbundled because the code is modified, not directly copied

Provides:       bundled(python3dist(sas7bdat))

# pandas/_testing/__init__.py: in _create_missing_idx():

#

#   below is cribbed from scipy.sparse

#

# Cannot be unbundled because only a few lines are copied, not a standalone

# function that we can call

Provides:       bundled(python3dist(scipy))

# pandas/_libs/src/ujson/lib/:

#

# This is a stripped-down copy of UltraJSON. It would be an obvious target for

# unbundling, except:

#

# - Pandas uses the C library API, but UltraJSON upstream does not support

#   building and installing it separately from the Python package.

# - In https://github.com/pandas-dev/pandas/issues/24711 it is suggested that

#   Pandas might rely on features of the particular vendored version of

#   UltraJSON. It’s not immediately clear whether this is still true or not.

Provides:       bundled(python3dist(ujson))

# pandas/core/accessor.py

#

#   Ported with modifications from xarray

#   https://github.com/pydata/xarray/blob/master/xarray/core/extensions.py

#   1. We don't need to catch and re-raise AttributeErrors as RuntimeErrors

#   2. We use a UserWarning instead of a custom Warning

#

# Cannot be unbundled because the copied code is forked.

Provides:       bundled(python3dist(xarray))

BuildRequires:  gcc

BuildRequires:  gcc-c++

BuildRequires:  python3-devel

# Runtime dependencies

BuildRequires:  python3dist(numpy) >= 1.26

BuildRequires:  python3dist(python-dateutil) >= 2.8.2

BuildRequires:  python3dist(pytz) >= 2020.1

%if %{with tests}

# From the [test] extra

BuildRequires:  python3dist(hypothesis)

BuildRequires:  python3dist(pytest)

BuildRequires:  python3dist(pytest-xdist)

%endif

%if %{without bootstrap}

# doc/source/getting_started/install.rst “Recommended dependencies”

# Since these provide large speedups, we make them hard dependencies except

# during bootstrapping.

BuildRequires:  python3dist(numexpr) >= 2.8.4

Requires:       python3dist(numexpr) >= 2.8.4

BuildRequires:  python3dist(bottleneck) >= 1.3.6

Requires:       python3dist(bottleneck) >= 1.3.6

# doc/source/getting_started/install.rst “Optional dependencies”

# We BR all weak dependencies to ensure they are installable.

# Timezones

BuildRequires:  tzdata >= 2022g

Recommends:     tzdata >= 2022g

# Visualization

BuildRequires:  python3dist(matplotlib) >= 3.6.3

Recommends:     python3dist(matplotlib) >= 3.6.3

BuildRequires:  python3dist(jinja2) >= 3.1.2

Recommends:     python3dist(jinja2) >= 3.1.2

BuildRequires:  python3dist(tabulate) >= 0.9

Recommends:     python3dist(tabulate) >= 0.9

# Computation

BuildRequires:  python3dist(scipy) >= 1.10

Recommends:     python3dist(scipy) >= 1.10

# python-numba is not currently packaged:

# BuildRequires:  python3dist(numba) >= 0.56.4

# Recommends:     python3dist(numba) >= 0.56.4

BuildRequires:  python3dist(xarray) >= 2022.12.0

Recommends:     python3dist(xarray) >= 2022.12.0

# Excel files

BuildRequires:  python3dist(xlrd) >= 2.0.1

Recommends:     python3dist(xlrd) >= 2.0.1

BuildRequires:  python3dist(xlsxwriter) >= 3.0.5

Recommends:     python3dist(xlsxwriter) >= 3.0.5

BuildRequires:  python3dist(openpyxl) >= 3.1

Recommends:     python3dist(openpyxl) >= 3.1

# python-calamine is not currently packaged:

# BuildRequires:  python3dist(python-calamine) >= 0.1.7

# Recommends:     python3dist(python-calamine) >= 0.1.7

# python-pyxlsb is not currently packaged:

# BuildRequires:  python3dist(pyxlsb) >= 1.0.10

# Recommends:     python3dist(pyxlsb) >= 1.0.10

# Not in doc/source/getting_started/install.rst, but in environment.yml and in

# some doc-strings:

BuildRequires:  python3dist(odfpy) >= 1.4.1

Recommends:     python3dist(odfpy) >= 1.4.1

# HTML

BuildRequires:  python3dist(beautifulsoup4) >= 4.11.2

Recommends:     python3dist(beautifulsoup4) >= 4.11.2

BuildRequires:  python3dist(html5lib) >= 1.1

Recommends:     python3dist(html5lib) >= 1.1

# lxml handled below:

# XML

BuildRequires:  python3dist(lxml) >= 4.9.2

Recommends:     python3dist(lxml) >= 4.9.2

# SQL databases

BuildRequires:  python3dist(sqlalchemy) >= 2

Recommends:     python3dist(sqlalchemy) >= 2

BuildRequires:  python3dist(psycopg2) >= 2.9.6

Recommends:     python3dist(psycopg2) >= 2.9.6

BuildRequires:  python3dist(pymysql) >= 1.0.2

Recommends:     python3dist(pymysql) >= 1.0.2

# Other data sources

%if 0%{?__isa_bits} != 32

# blosc2 does not support 32-bit architectures:

BuildRequires:  python3dist(tables) >= 3.8

Recommends:     python3dist(tables) >= 3.8

%endif

# Dependencies on blosc and zlib are indirect, via PyTables, so we do not

# encode them here. Note also that the minimum blosc version in the

# documentation seems to be that of the blosc C library, not of the blosc PyPI

# package.

# python-fastparquet is not currently packaged:

# BuildRequires:  python3dist(fastparquet) >= 2022.12.0

# Recommends:     python3dist(fastparquet) >= 2022.12.0

# libarrow does not support 32-bit architectures:

%if 0%{?__isa_bits} != 32

BuildRequires:  python3dist(pyarrow) >= 10.0.1

Recommends:     python3dist(pyarrow) >= 10.0.1

%endif

# python-pyreadstat is not currently packaged:

# BuildRequires:  python3dist(pyreadstat) >= 1.2

# Recommends:     python3dist(pyreadstat) >= 1.2

# Access data in the cloud

BuildRequires:  python3dist(fsspec) >= 2022.11

Recommends:     python3dist(fsspec) >= 2022.11

BuildRequires:  python3dist(gcsfs) >= 2022.11

Recommends:     python3dist(gcsfs) >= 2022.11

# python-pandas-gbq is not currently packaged:

# BuildRequires:  python3dist(pandas-gbq) >= 0.19

# Recommends:     python3dist(pandas-gbq) >= 0.19

# python-s3fs is not currently packaged:

# BuildRequires:  python3dist(s3fs) >= 2022.11

# Recommends:     python3dist(s3fs) >= 2022.11

# Clipboard

BuildRequires:  python3dist(pyqt5)

Recommends:     python3dist(pyqt5)

BuildRequires:  python3dist(qtpy)

Recommends:     python3dist(qtpy)

BuildRequires:  xclip

Recommends:     xclip

BuildRequires:  xsel

Recommends:     xsel

# Compression

BuildRequires:  python3dist(zstandard) >= 0.19

Recommends:     python3dist(zstandard) >= 0.19

# This is just an “ecosystem” package in the upstream documentation, but there

# is an integration test for it. This package historically had a weak

# dependency on it, but this was unnecessary.

BuildRequires:  python3dist(pandas-datareader)

%endif

%description -n python3-pandas %_description

%package -n python3-pandas+test

Summary:        Tests and test extras for Pandas

# See comment above base package License tag for licensing breakdown.

License:        BSD-3-Clause AND MIT

Requires:       python3-pandas%{?_isa} = %{version}-%{release}

%if %{without bootstrap}

# Additional BR’s and weak dependencies below are generally those that don’t

# provide enough added functionality to be weak dependencies of the library

# package, but for which there is some integration support and additional tests

# that can be enabled.

# Additional dependencies from environment.yml: “testing”

# Those not in the “test” extra are treated as weak dependencies for the tests.

BuildRequires:  python3dist(boto3)

Recommends:     python3dist(boto3)

BuildRequires:  python3dist(botocore) >= 1.11

Recommends:     python3dist(botocore) >= 1.11

# Already covered by “test” extra

# BuildRequires:  python3dist(hypothesis) >= 3.82

# Recommends:     python3dist(hypothesis) >= 3.82

# python-moto is not yet packaged

# BuildRequires:  python3dist(moto)

# Recommends:     python3dist(moto)

BuildRequires:  python3dist(flask)

Recommends:     python3dist(flask)

# Already covered by “test” extra

# BuildRequires:  python3dist(pytest) >= 5.0.1

# Requires:       python3dist(pytest) >= 5.0.1

# Already covered by “test” extra

# BuildRequires:  python3dist(pytest-xdist) >= 1.21

# Requires:       python3dist(pytest-xdist) >= 1.21

BuildRequires:  python3dist(pytest-asyncio)

Recommends:     python3dist(pytest-asyncio)

# python-pytest-instafail is not yet packaged

# BuildRequires:  python3dist(pytest-instafail)

# Recommends:     python3dist(pytest-instafail)

# Additional dependencies from environment.yml:

# “Dask and its dependencies (that dont install with dask)”

# Asks for dask-core, but we just have dask

BuildRequires:  python3dist(dask)

Recommends:     python3dist(dask)

BuildRequires:  python3dist(toolz) >= 0.7.3

Recommends:     python3dist(toolz) >= 0.7.3

BuildRequires:  python3dist(partd) >= 0.3.10

Recommends:     python3dist(partd) >= 0.3.10

BuildRequires:  python3dist(cloudpickle) >= 0.2.1

Recommends:     python3dist(cloudpickle) >= 0.2.1

# Additional dependencies from environment.yml: “downstream tests”

BuildRequires:  python3dist(seaborn)

Recommends:     python3dist(seaborn)

BuildRequires:  python3dist(statsmodels)

Recommends:     python3dist(statsmodels)

# environment.yml: Needed for downstream xarray.CFTimeIndex test

BuildRequires:  python3dist(cftime)

Recommends:     python3dist(cftime)

# environment.yml: optional

BuildRequires:  python3dist(ipython) >= 7.11.1

Recommends:     python3dist(ipython) >= 7.11.1

# pandas/tests/io/data/spss/*.sav:

#

# From Haven

Provides:       bundled(R-haven)

# pandas/tests/window/test_rolling.py: test_rolling_std_neg_sqrt()

#

#   unit test from Bottleneck

#

# There is no reasonable path to unbundling a single unit test.

Provides:       bundled(python3dist(bottleneck))

%endif

%description -n python3-pandas+test

These are the tests for python3-pandas. This package:

• Provides the “pandas.tests” package

• Makes sure the “test” extra dependencies are installed

• Carries additonal weak dependencies for running the tests

%prep

%autosetup -n pandas-%{version} -p1

# Let versioneer know what version this is

echo '__version__="%{version}"' > _version_meson.py

# Ensure Cython-generated sources are re-generated

rm -vf $(grep -rl '/\* Generated by Cython')

# We just want to build with the numpy in Fedora:

sed -r -i '/\boldest-supported-numpy\b/d' pyproject.toml

# We don't need the python tzdata package because we have the system tzdata package

sed -i '/tzdata>=2022.7/d' pyproject.toml

# Unpin meson

sed -i 's/meson-python==0.13.1/meson-python>=0.13.1/' pyproject.toml

sed -i 's/meson==1.2.1/meson>=1.2.1/' pyproject.toml

# Unpin Cython

sed -i 's/Cython==3.0.5/Cython>=3.0.5/' pyproject.toml

%generate_buildrequires

# the build is expensive, so we don't use -w

# we list the runtime and test BuildRequires manually

%pyproject_buildrequires -R

%build

%pyproject_wheel

%install

%pyproject_install

%pyproject_save_files pandas

%check

%if %{with tests}

m="${m-}${m+ and }not network"

m="${m-}${m+ and }not db"

%if %{without slow_tests}

m="${m-}${m+ and }not slow"

%endif

# Clipboard tests don’t run without a graphical session, and it’s not worth

# using xvfb-run just for them.

m="${m-}${m+ and }not clipboard"

%if %{without single_tests}

m="${m-}${m+ and }not single"

%endif

# This test allocates a huge amount of memory (~12GB), which causes flaky OOM

# failures on some builders. It’s not worth it.

# https://github.com/pandas-dev/pandas/issues/45223#issuecomment-1250912663

k="${k-}${k+ and }not test_bytes_exceed_2gb"

# This test (only) expects the current working directory to be the

# site-packages directory containing the built pandas. This is not how we run

# the tests, because we don’t want to clutter the buildroot with

# testing-related hidden files and directories. We could run tests from

# %%pyproject_build_lib if this were a problem for a lot of tests, but it’s

# easier just to skip it.

k="${k-}${k+ and }not test_html_template_extends_options"

# Those tests started failing as of 2024-04-12. Not sure why, though.

# Dask wasn't updated at the time.

# > return get(descriptor, obj, type(obj))

# E   TypeError: descriptor '__call__' for 'type' objects doesn't apply to a 'property' object

# and

# [XPASS(strict)] pyarrow doesn't support this

k="${k-}${k+ and }not test_dask"

k="${k-}${k+ and }not test_construct_dask_float_array"

k="${k-}${k+ and }not test_multi_thread_string_io_read_csv[pyarrow]"

# Two tests started failing with matplotlib >= 3.9.0

# E   matplotlib._api.deprecation.MatplotlibDeprecationWarning:

# The plot_date function was deprecated in Matplotlib 3.9

# and will be removed in 3.11. Use plot instead.

#

# E   UserWarning: No artists with labels found to put in legend.

# Note that artists whose label start with an underscore are ignored

# when legend() is called with no argument.

k="${k-}${k+ and }not test_mpl_nopandas"

k="${k-}${k+ and }not test_plot_scatter_shape"

%ifarch %{ix86}

# These failures are i686-specific; most are likely 32-bit issues. It’s not

# really worth trying to fix them.

# E   AssertionError: DataFrame.iloc[:, 2] (column name="C") are different

# E

# E   DataFrame.iloc[:, 2] (column name="C") values are different (11.66363 %)

# E   [index]: [0, 1, …

# Fails for [left], [right], [outer], and [inner]

k="${k-}${k+ and }not (TestMerge and test_int64_overflow_how_merge)"

# E       AssertionError: DataFrame.index are different

# E

# E       Attribute "dtype" are different

# E       [left]:  int32

# E       [right]: int64

k="${k-}${k+ and }not (TestMerge and test_int64_overflow_sort_false_order)"

# E           AssertionError: Attributes of DataFrame.iloc[:, 1] (column name="b") are different

# E

# E           Attribute "dtype" are different

# E           [left]:  int32

# E           [right]: int64

k="${k-}${k+ and }not test_frame_setitem_dask_array_into_new_col"

# E       IndexError: index 0 is out of bounds for axis 0 with size 0

k="${k-}${k+ and }not (TestPivotTable and test_pivot_number_of_levels_larger_than_int32)"

k="${k-}${k+ and }not (TestStackUnstackMultiLevel and test_unstack_number_of_levels_larger_than_int32)"

# [XPASS(strict)] Floating point error

k="${k-}${k+ and }not (TestTimedeltas and test_to_timedelta_float)"

%endif

%ifarch s390x

# Note that pandas does not test big-endian support but will happily accept

# patches to improve it:

# https://github.com/pandas-dev/pandas/issues/4737#issuecomment-1090931741

# TODO: Why does this fail?

#

# >                   os.fsync(self._handle.fileno())

# E                   OverflowError: Python int too large to convert to C int

k="${k-}${k+ and }not test_flush"

# TODO: Why does this fail? The differences are large!

k="${k-}${k+ and }not test_rolling_var_numerical_issues"

# These are a cluster of similar pyarrow/parquet tests with apparent endianness

# issues. It is not immediately obvious where the bug is—in the library or in

# the tests?

k="${k-}${k+ and }not (TestBasic and test_dtype_backend[pyarrow])"

k="${k-}${k+ and }not (TestBasic and test_multiindex_with_columns)"

k="${k-}${k+ and }not (TestBasic and test_write_column_index_nonstring[pyarrow])"

k="${k-}${k+ and }not (TestBasic and test_write_column_index_string)"

k="${k-}${k+ and }not (TestBasic and test_write_column_multiindex[pyarrow])"

k="${k-}${k+ and }not (TestBasic and test_write_column_multiindex_nonstring[pyarrow])"

k="${k-}${k+ and }not (TestBasic and test_write_column_multiindex_string)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_basic)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_to_bytes_without_path_or_buf_provided)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_categorical)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_additional_extension_arrays)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_pyarrow_backed_string_array[python])"

k="${k-}${k+ and }not (TestParquetPyArrow and test_pyarrow_backed_string_array[pyarrow])"

k="${k-}${k+ and }not (TestParquetPyArrow and test_additional_extension_types)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_infer_string_large_string_type)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_read_dtype_backend_pyarrow_config)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_read_dtype_backend_pyarrow_config_index)"

k="${k-}${k+ and }not (TestParquetPyArrow and test_roundtrip_decimal)"

k="${k-}${k+ and }not test_to_read_gcs[parquet]"

# Similarly, there are a cluster of similar stata test failures for which the

# root cause is not immediately obvious.

k="${k-}${k+ and }not (TestStata and test_writer_117)"

k="${k-}${k+ and }not (TestStata and test_convert_strl_name_swap)"

k="${k-}${k+ and }not (TestStata and test_strl_latin1)"

# Fails for [118], [119], and [None]

k="${k-}${k+ and }not (TestStata and test_utf8_writer)"

# These crash, and are probably a blosc2 or PyTables issue.

k="${k-}${k+ and }not test_complibs[blosc2"

%endif

%ifarch x86_64

# These are brittle and fail with tiny floating-point differences on COPR

# builders but not Koji builders, like:

# >           raise_assert_detail(obj, msg, left, right, index_values=index_values)

# E           AssertionError: numpy array are different

# E

# E           numpy array values are different (16.66667 %)

# E           [left]:  [0.09999999999999999, 1.0, 10.0, 100.0, 1000.0, 10000.0]

# E           [right]: [0.1, 1.0, 10.0, 100.0, 1000.0, 10000.0]

k="${k-}${k+ and }not (TestSeriesPlots and test_bar_log)"

k="${k-}${k+ and }not (TestDataFramePlotsSubplots and test_bar_log_no_subplots)"

k="${k-}${k+ and }not (TestDataFramePlotsSubplots and test_bar_log_subplots)"

%endif

# Ensure pytest doesn’t find the “un-built” library. We can get away with this

# approach because the tests are also in the installed library. We can’t simply

# “cd” to the buildroot’s python3_sitearch because testing leaves files in the

# current working directory.

mkdir -p _empty

cd _empty

# See: test_fast.sh

# Workaround for pytest-xdist flaky collection order

# https://github.com/pytest-dev/pytest/issues/920

# https://github.com/pytest-dev/pytest/issues/1075

export PYTHONHASHSEED="$(

  %{python3} -c 'import random; print(random.randint(1, 4294967295))'

)"

# Previously, we ran tests in parallel. Upstream seems to support this;

# however, in practice, there were still some flaky test failures that seem to

# be fixed by eschewing parallelism (-n 1).

#

# If we start running tests in parallel again in the future, note that on

# 32-bit platforms (%%if 0%%{?__isa_bits} == 32) it may be necessary to limit

# the number of concurrent tests to e.g. 8 in order to prevent memory

# exhaustion.

%pytest -v '%{buildroot}%{python3_sitearch}/pandas' \

    -o cache_dir="$PWD/pytest-cache" \

    --no-strict-data-files \

    -m "${m-}" \

    -k "${k-}" \

    -n 1 \

    -r sxX

%else

# Some imports require optional dependencies, and must be excluded during

# bootstrapping.

%{pyproject_check_import \

  %{?with_bootstrap:-e 'pandas.io.formats.style'} \

  %{?with_bootstrap:-e 'pandas.io.formats.style_render'} \

  %{?with_bootstrap:-e 'pandas.core.arrays.arrow.extension_types'} \

  -e 'pandas.conftest' \

  -e 'pandas.tests.*'}

%endif

%files -n python3-pandas -f %{pyproject_files}

# While pyproject_files automatically handles the LICENSE file in the Python

# package’s dist-info directory, we also want to package the entire LICENSES/

# directory to include third-party license text.  We include a second copy of

# the LICENSE file since it would be surprising to see a license directory for

# the package without the overall license file in it.

%license LICENSE LICENSES/

%doc README.md

%exclude %{python3_sitearch}/pandas/tests

%files -n python3-pandas+test

%{python3_sitearch}/pandas/tests

%ghost %{python3_sitearch}/*.dist-info

%changelog

* Fri Jul 19 2024 Fedora Release Engineering <releng@fedoraproject.org> - 2.2.1-7

- Rebuilt for https://fedoraproject.org/wiki/Fedora_41_Mass_Rebuild

* Wed Jun 26 2024 Orion Poplawski <orion@nwra.com> - 2.2.1-6

- Add upstream patch to fix tests with Python 3.13

* Tue Jun 18 2024 Python Maint <python-maint@redhat.com> - 2.2.1-5

- Rebuilt for Python 3.13

* Sat Jun 08 2024 Python Maint <python-maint@redhat.com> - 2.2.1-4

- Bootstrap for Python 3.13

* Mon May 06 2024 Sandro <devel@penguinpee.nl> - 2.2.1-3

- Stop building for i686