fix(deps): drop pytz from dependencies (ibis-project#10976)

## Description of changes

Hi! This PR proposes dropping `pytz` from primary dependencies and
replacing it with stdlib equivalents, as recommended by `pytz`
[docs](https://github.com/stub42/pytz/blob/master/src/README.rst):

> Projects using Python 3.9 or later should be using the support now
included as part of the standard library, and third party packages work
with it such as tzdata. pytz offers no advantages beyond backwards
compatibility with code written for earlier versions of Python.

Given that Ibis only supports Python 3.9 and higher, this should be
pretty safe.

Author: romaingd

conda/environment-arm64-flink.yml

@@ -38,7 +38,6 @@ dependencies:
   - python-dateutil >=2.8.2
   - python-duckdb >=0.8.1
   - python-graphviz >=0.16
-  - pytz >=2022.7
   - regex >=2021.7.6
   - requests >=2
   - rich >=13.8.0

conda/environment-arm64.yml

@@ -39,7 +39,6 @@ dependencies:
   - python-dateutil >=2.8.2
   - python-duckdb >=0.8.1
   - python-graphviz >=0.16
-  - pytz >=2022.7
   - regex >=2021.7.6
   - requests >=2
   - rich >=13.8.0

conda/environment.yml

@@ -40,7 +40,6 @@ dependencies:
   - python-dateutil >=2.8.2
   - python-duckdb >=0.8.1
   - python-graphviz >=0.16
-  - pytz >=2022.7
   - regex >=2021.7.6
   - requests >=2
   - rich >=13.8.0

ibis/backends/bigquery/tests/system/test_client.py

@@ -10,7 +10,6 @@
 import pandas.testing as tm
 import pyarrow as pa
 import pytest
-import pytz
 from google.api_core.exceptions import Forbidden
 
 import ibis
@@ -256,15 +255,15 @@ def test_multiple_project_queries_execute(con):
 
 def test_string_as_timestamp(con):
     timestamp = pd.Timestamp(
-        datetime.datetime(year=2017, month=2, day=6), tz=pytz.timezone("UTC")
+        datetime.datetime(year=2017, month=2, day=6), tz=datetime.timezone.utc
     )
     expr = ibis.literal("2017-02-06").as_timestamp("%F")
     result = con.execute(expr)
     assert result == timestamp
 
     timestamp_tz = pd.Timestamp(
         datetime.datetime(year=2017, month=2, day=6, hour=5),
-        tz=pytz.timezone("UTC"),
+        tz=datetime.timezone.utc,
     )
     expr_tz = ibis.literal("2017-02-06 America/New_York").as_timestamp("%F %Z")
     result_tz = con.execute(expr_tz)

ibis/backends/flink/__init__.py

@@ -2,6 +2,7 @@
 
 import itertools
 import re
+import zoneinfo
 from typing import TYPE_CHECKING, Any
 
 import sqlglot as sg
@@ -1055,7 +1056,6 @@ def _from_pyflink_table_to_pyarrow_batches(
     ):
         import pyarrow as pa
         import pyarrow_hotfix  # noqa: F401
-        import pytz
         from pyflink.java_gateway import get_gateway
         from pyflink.table.serializers import ArrowSerializer
         from pyflink.table.types import create_arrow_schema
@@ -1084,7 +1084,7 @@ def _from_pyflink_table_to_pyarrow_batches(
             pyflink_schema.get_field_names(), get_field_data_types(pyflink_schema)
         )
 
-        timezone = pytz.timezone(
+        timezone = zoneinfo.ZoneInfo(
             table._j_table.getTableEnvironment().getConfig().getLocalTimeZone().getId()
         )
         serializer = ArrowSerializer(

ibis/backends/impala/tests/test_client.py

@@ -5,7 +5,6 @@
 
 import pandas as pd
 import pytest
-import pytz
 
 import ibis
 import ibis.expr.datatypes as dt
@@ -220,7 +219,9 @@ def test_day_of_week(con):
 
 
 def test_datetime_to_int_cast(con):
-    timestamp = pytz.utc.localize(datetime.datetime(2021, 9, 12, 14, 45, 33, 0))
+    timestamp = datetime.datetime(
+        2021, 9, 12, 14, 45, 33, 0, tzinfo=datetime.timezone.utc
+    )
     d = ibis.literal(timestamp)
     result = con.execute(d.cast("int64"))
     assert result == pd.Timestamp(timestamp).value // 1000

ibis/backends/tests/test_array.py

@@ -3,11 +3,10 @@
 import math
 import statistics
 from collections import Counter
-from datetime import datetime
+from datetime import datetime, timezone
 from functools import partial
 
 import pytest
-import pytz
 import toolz
 from pytest import param
 
@@ -1474,7 +1473,7 @@ def swap(token):
             ),
         ),
         param(
-            pytz.UTC,
+            timezone.utc,
             id="utc",
             marks=pytest.mark.notyet(
                 ["trino"],
@@ -1569,8 +1568,8 @@ def test_timestamp_range(con, start, stop, step, freq, tzinfo):
     ("start", "stop", "step"),
     [
         param(
-            datetime(2017, 1, 1, tzinfo=pytz.UTC),
-            datetime(2017, 1, 2, tzinfo=pytz.UTC),
+            datetime(2017, 1, 1, tzinfo=timezone.utc),
+            datetime(2017, 1, 2, tzinfo=timezone.utc),
             ibis.interval(hours=0),
             id="pos",
             marks=[
@@ -1587,8 +1586,8 @@ def test_timestamp_range(con, start, stop, step, freq, tzinfo):
             ],
         ),
         param(
-            datetime(2017, 1, 1, tzinfo=pytz.UTC),
-            datetime(2017, 1, 2, tzinfo=pytz.UTC),
+            datetime(2017, 1, 1, tzinfo=timezone.utc),
+            datetime(2017, 1, 2, tzinfo=timezone.utc),
             -ibis.interval(hours=0),
             id="neg",
             marks=[

ibis/common/temporal.py

@@ -8,7 +8,6 @@
 
 import dateutil.parser
 import dateutil.tz
-import pytz
 from public import public
 
 from ibis import util
@@ -199,7 +198,7 @@ def normalize_timezone(tz):
             return dateutil.tz.gettz(tz)
     elif isinstance(tz, (int, float)):
         return datetime.timezone(datetime.timedelta(hours=tz))
-    elif isinstance(tz, (dateutil.tz.tzoffset, pytz._FixedOffset)):
+    elif isinstance(tz, dateutil.tz.tzoffset):
         # this way we have a proper tzname() output, e.g. "UTC+01:00"
         return datetime.timezone(tz.utcoffset(None))
     elif isinstance(tz, datetime.tzinfo):

ibis/common/tests/test_temporal.py

@@ -5,7 +5,6 @@
 
 import dateutil
 import pytest
-import pytz
 
 from ibis.common.patterns import CoercedTo
 from ibis.common.temporal import (
@@ -125,7 +124,7 @@ def test_interval_unit_compatibility():
     ("value", "expected"),
     [
         (None, None),
-        (pytz.UTC, pytz.UTC),
+        (timezone.utc, timezone.utc),
         ("UTC", dateutil.tz.tzutc()),
         ("Europe/Budapest", dateutil.tz.gettz("Europe/Budapest")),
         (+2, timezone(timedelta(seconds=7200))),

ibis/expr/datatypes/tests/test_value.py

@@ -4,10 +4,9 @@
 import enum
 import json
 from collections import OrderedDict
-from datetime import date, datetime, timedelta
+from datetime import date, datetime, timedelta, timezone
 
 import pytest
-import pytz
 
 import ibis.expr.datatypes as dt
 
@@ -94,9 +93,9 @@ def test_infer_mixed_type_fails():
 
 
 def test_infer_timestamp_with_tz():
-    now_utc = datetime.now(pytz.UTC)
-    assert now_utc.tzinfo == pytz.UTC
-    assert dt.infer(now_utc).timezone == str(pytz.UTC)
+    now_utc = datetime.now(timezone.utc)
+    assert now_utc.tzinfo == timezone.utc
+    assert dt.infer(now_utc).timezone == str(timezone.utc)
 
 
 def test_infer_timedelta():
@@ -127,19 +126,19 @@ def test_infer_timedelta():
         ("2019-01-01 01:02:03.000004", datetime(2019, 1, 1, 1, 2, 3, 4)),
         (
             "2019-01-01 01:02:03.000004+00:00",
-            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=pytz.utc),
+            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=timezone.utc),
         ),
         (
             "2019-01-01 01:02:03.000004+01:00",
-            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=pytz.FixedOffset(60)),
+            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=timezone(timedelta(hours=1))),
         ),
         (
             "2019-01-01 01:02:03.000004-01:00",
-            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=pytz.FixedOffset(-60)),
+            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=timezone(timedelta(hours=-1))),
         ),
         (
             "2019-01-01 01:02:03.000004+01",
-            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=pytz.FixedOffset(60)),
+            datetime(2019, 1, 1, 1, 2, 3, 4, tzinfo=timezone(timedelta(hours=1))),
         ),
         (datetime(2019, 1, 1), datetime(2019, 1, 1)),
         (datetime(2019, 1, 1, 1, 2, 3, 4), datetime(2019, 1, 1, 1, 2, 3, 4)),

ibis/tests/benchmarks/test_benchmarks.py

@@ -13,7 +13,6 @@
 import string
 
 import pytest
-import pytz
 from pytest import param
 
 import ibis
@@ -977,8 +976,8 @@ def test_wide_relocate(benchmark, input, column, relative, cols):
 
 
 def test_duckdb_timestamp_conversion(benchmark, con):
-    start = datetime.datetime(2000, 1, 1, tzinfo=pytz.UTC)
-    stop = datetime.datetime(2000, 2, 1, tzinfo=pytz.UTC)
+    start = datetime.datetime(2000, 1, 1, tzinfo=datetime.timezone.utc)
+    stop = datetime.datetime(2000, 2, 1, tzinfo=datetime.timezone.utc)
     expr = ibis.range(start, stop, ibis.interval(seconds=1)).unnest()
 
     series = benchmark(con.execute, expr)

ibis/tests/expr/test_value_exprs.py

@@ -5,12 +5,11 @@
 import operator
 import uuid
 from collections import OrderedDict
-from datetime import date, datetime, time
+from datetime import date, datetime, time, timezone
 from decimal import Decimal
 from operator import attrgetter, methodcaller
 
 import pytest
-import pytz
 import toolz
 from pytest import param
 
@@ -1116,7 +1115,7 @@ def test_timestamp_with_timezone():
     assert expr.op().value == expected
 
     expr = ibis.timestamp("2017-01-01", timezone="UTC")
-    expected = datetime(2017, 1, 1, tzinfo=pytz.timezone("UTC"))
+    expected = datetime(2017, 1, 1, tzinfo=timezone.utc)
     assert expr.op().value == expected
 
 

ibis/tests/strategies.py

@@ -3,7 +3,6 @@
 import warnings
 
 import hypothesis as h
-import hypothesis.extra.pytz as tzst
 import hypothesis.strategies as st
 import pytest
 
@@ -115,7 +114,7 @@ def time_dtype(nullable=_nullable):
     return st.builds(dt.Time, nullable=nullable)
 
 
-_timezone = st.none() | tzst.timezones().map(str)
+_timezone = st.none() | st.timezones().map(str)
 _interval = st.sampled_from(list(IntervalUnit))
 _timestamp_scale = st.none() | st.integers(min_value=0, max_value=9)
 

pyproject.toml

@@ -51,10 +51,10 @@ dependencies = [
   "atpublic>=2.3",
   "parsy>=2",
   "python-dateutil>=2.8.2",
-  "pytz>=2022.7",
   "sqlglot>=23.4",
   "toolz>=0.11",
   "typing-extensions>=4.3.0",
+  "tzdata>=2022.7",           # fallback time zone data on Windows
 ]
 
 [project.urls]
@@ -138,6 +138,8 @@ exasol = [
   "rich>=12.4.4",
 ]
 flink = [
+  # adding apache-flink would lock pyarrow to an old version (for all backends),
+  # due to the transitional dependency apache-beam imposing a low upper bound.
   "pyarrow>=10.0.1",
   "pyarrow-hotfix>=0.4",
   "numpy>=1.23.2,<3",