python 3.8 external pins pandas<1.1.0, numpy<1.19.0 (#962)

dunckerr · web-flow · commit 7c4b378deaae · 2022-09-03T15:44:02.000+01:00
diff --git a/CHANGES.md b/CHANGES.md
@@ -1,8 +1,9 @@
 ## Changelog
 
-### 1.80.5
+### 1.80.5 (2022-09-05)
   * Feature: #950 remove all traces of python 2.7 and six package
   * Feature: #959 add python 3.8 support
+  * Feature: #962 pin pandas<1.1.0, numpy<1.19.0
 
 ### 1.80.4 (2022-01-25)
   * Bugfix:  #940 fix rows per chunk causing divide by zero
diff --git a/README.md b/README.md
@@ -3,7 +3,7 @@
 
 [![Documentation Status](https://readthedocs.org/projects/arctic/badge/?version=latest)](https://arctic.readthedocs.io/en/latest/?badge=latest)
 [![CircleCI](https://circleci.com/gh/man-group/arctic/tree/master.svg?style=shield)](https://app.circleci.com/pipelines/github/man-group/arctic?branch=master)
-[![PyPI](https://img.shields.io/pypi/v/arctic)](https://pypi.org/project/arctic/)
+[![PyPI](https://img.shields.io/pypi/v/arctic)](https://pypi.org/project/arctic)
 [![Python](https://img.shields.io/badge/Python-3.6|3.7|3.8-green.svg)](https://github.com/man-group/arctic)
 [![Join the chat at https://gitter.im/man-group/arctic](https://badges.gitter.im/Join%20Chat.svg)](https://gitter.im/man-group/arctic?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
 
@@ -19,9 +19,6 @@ Arctic has been under active development at [Man Group](https://www.man.com/) si
 
 ---
 
-
-:loudspeaker::loudspeaker::loudspeaker: **BUILDING THE NEXT GENERATION OF ARCTIC** :loudspeaker::loudspeaker::loudspeaker:
-
 ## Man Group is currently building the next generation of Arctic.
 
 This will offer the same intuitive Python-centric API whilst utilizing a custom C++ storage engine and modern S3 compatible object storage to provide a timeseries database that is:
@@ -141,10 +138,10 @@ Arctic storage implementations are **pluggable**.  VersionStore is the default.
 
 Arctic currently works with:
 
- * Python 3.6, 3.7
- * pymongo >= 3.6.0  <= 3.11.0
- * Pandas  >= 0.22.0 <= 1.0.3
- * MongoDB >= 2.4.x  <= 4.2.8
+ * python 3.6, 3.7, 3.8
+ * pymongo >= 3.6.0 <= 3.11.0
+ * pandas >= 0.22.0 <1.1.0
+ * MongoDB >= 2.4.x <= 4.2.8
 
 
 Operating Systems:
diff --git a/arctic/date/_util.py b/arctic/date/_util.py
@@ -2,7 +2,6 @@
 import datetime
 import sys
 from datetime import timedelta
-
 import pandas as pd
 
 from ._daterange import DateRange
@@ -168,11 +167,13 @@ def datetime_to_ms(d):
     try:
         millisecond = d.microsecond // 1000
 
-        # python3.8 workaround https://github.com/pandas-dev/pandas/issues/32174
+        # https://github.com/pandas-dev/pandas/issues/32526
+        # https://github.com/pandas-dev/pandas/issues/32174
         if sys.version_info < (3, 8, 0):
             return calendar.timegm(_add_tzone(d).utctimetuple()) * 1000 + millisecond
         else:
             tmp = _add_tzone(d)
+            # convert to Datetime seems to be the only reliable option
             if isinstance(tmp, pd.Timestamp):
                 return calendar.timegm(tmp.to_pydatetime().utctimetuple()) * 1000 + millisecond
             else:
diff --git a/arctic/serialization/numpy_records.py b/arctic/serialization/numpy_records.py
@@ -95,7 +95,13 @@ def _index_from_records(self, recarr):
         if len(index) == 1:
             rtn = Index(np.copy(recarr[str(index[0])]), name=index[0])
             if isinstance(rtn, DatetimeIndex) and 'index_tz' in recarr.dtype.metadata:
-                rtn = rtn.tz_localize('UTC').tz_convert(recarr.dtype.metadata['index_tz'])
+                if PD_VER >= '1.0.4':
+                    if isinstance(recarr.dtype.metadata['index_tz'], list):
+                        rtn = rtn.tz_localize('UTC').tz_convert(recarr.dtype.metadata['index_tz'][0])
+                    else:
+                        rtn = rtn.tz_localize('UTC').tz_convert(recarr.dtype.metadata['index_tz'])
+                else:
+                    rtn = rtn.tz_localize('UTC').tz_convert(recarr.dtype.metadata['index_tz'])
         else:
             level_arrays = []
             index_tz = recarr.dtype.metadata.get('index_tz', [])
diff --git a/arctic/store/_pickle_store.py b/arctic/store/_pickle_store.py
@@ -92,7 +92,7 @@ def write(self, arctic_lib, version, symbol, item, _previous_version):
         version['blob'] = _MAGIC_CHUNKEDV2
         # Python 3.8 onwards uses protocol 5 which cannot be unpickled in Python versions below that, so limiting
         # it to use a maximum of protocol 4 in Python which is understood by 3.4 onwards and is still fairly efficient.
-        # The min() used to allow lower versions to be used in py2 (which supported a max of 2)
+        # pickle version 4 is introduced with  python 3.4 and default with 3.8 onward
         pickle_protocol = min(pickle.HIGHEST_PROTOCOL, 4)
         pickled = pickle.dumps(item, protocol=pickle_protocol)
 
diff --git a/setup.cfg b/setup.cfg
@@ -1,2 +1,5 @@
+[metadata]
+name = arctic
+
 [pycodestyle]
 ignore = E122,E126,E501,E731,W503,E722,W504
diff --git a/setup.py b/setup.py
@@ -60,7 +60,7 @@ def run_tests(self):
     name="arctic",
     version="1.80.5",
     author="Man AHL Technology",
-    author_email="ManAHLTech@ahl.com",
+    author_email="arctic@man.com",
     description=("AHL Research Versioned TimeSeries and Tick store"),
     license="GPL",
     keywords=["ahl", "keyvalue", "tickstore", "mongo", "timeseries", ],
@@ -69,15 +69,15 @@ def run_tests(self):
     long_description='\n'.join((long_description, changelog)),
     long_description_content_type="text/markdown",
     cmdclass={'test': PyTest},
-    setup_requires=["numpy<=1.18.4",
+    setup_requires=["numpy<1.19.0",
                     "setuptools-git",
                    ],
     install_requires=["decorator",
                       "enum-compat",
                       "mock",
                       "mockextras",
-                      "pandas<=1.0.3",
-                      "numpy<=1.18.4",
+                      "pandas<1.1.0",
+                      "numpy<1.19.0",
                       "pymongo>=3.6.0, <= 3.11.0",
                       "pytz",
                       "tzlocal",
diff --git a/tests/integration/chunkstore/test_chunkstore.py b/tests/integration/chunkstore/test_chunkstore.py
@@ -15,7 +15,7 @@
 from arctic.date import DateRange
 from arctic.exceptions import NoDataFoundException
 from tests.integration.chunkstore.test_utils import create_test_data
-from tests.util import assert_frame_equal_
+from tests.util import assert_frame_equal_, assert_series_equal_
 
 
 def test_write_dataframe(chunkstore_lib):
@@ -714,7 +714,7 @@ def test_overwrite_series(chunkstore_lib):
 
     chunkstore_lib.write('test', s)
     chunkstore_lib.write('test', s + 1)
-    assert_series_equal(chunkstore_lib.read('test'), s + 1)
+    assert_series_equal_(chunkstore_lib.read('test'), s + 1, check_freq=False)
 
 
 def test_overwrite_series_monthly(chunkstore_lib):
diff --git a/tests/integration/store/test_pandas_store.py b/tests/integration/store/test_pandas_store.py
@@ -21,7 +21,7 @@
 # Do not remove PandasStore, used in global scope
 from arctic.store._pandas_ndarray_store import PandasDataFrameStore, PandasSeriesStore, PandasStore
 from arctic.store.version_store import register_versioned_storage
-from tests.util import assert_frame_equal_
+from tests.util import assert_frame_equal_, assert_series_equal_
 
 register_versioned_storage(PandasDataFrameStore)
 
@@ -851,30 +851,30 @@ def test_daterange_large_DataFrame_middle(library):
 
 @pytest.mark.parametrize("df,assert_equal", [
     (DataFrame(index=date_range(dt(2001, 1, 1), freq='D', periods=30000),
-               data=list(range(30000)), columns=['A']), assert_frame_equal),
+               data=list(range(30000)), columns=['A']), assert_frame_equal_),
     (Series(index=date_range(dt(2001, 1, 1), freq='D', periods=30000),
-            data=range(30000)), assert_series_equal),
+            data=range(30000)), assert_series_equal_),
 ])
 def test_daterange(library, df, assert_equal):
     df.index.name = 'idx'
     df.name = 'FOO'
     library.write('MYARR', df)
     # whole array
     saved_arr = library.read('MYARR').data
-    assert_equal(df, saved_arr)
-    assert_equal(df, library.read('MYARR', date_range=DateRange(df.index[0])).data)
-    assert_equal(df, library.read('MYARR', date_range=DateRange(df.index[0], df.index[-1])).data)
-    assert_equal(df, library.read('MYARR', date_range=DateRange()).data)
-    assert_equal(df[df.index[10]:], library.read('MYARR', date_range=DateRange(df.index[10])).data)
-    assert_equal(df[:df.index[10]], library.read('MYARR', date_range=DateRange(end=df.index[10])).data)
-    assert_equal(df[df.index[-1]:], library.read('MYARR', date_range=DateRange(df.index[-1])).data)
-    assert_equal(df[df.index[-1]:], library.read('MYARR', date_range=DateRange(df.index[-1], df.index[-1])).data)
-    assert_equal(df[df.index[0]:df.index[0]], library.read('MYARR', date_range=DateRange(df.index[0], df.index[0])).data)
-    assert_equal(df[:df.index[0]], library.read('MYARR', date_range=DateRange(end=df.index[0])).data)
+    assert_equal(df, saved_arr, check_freq=False)
+    assert_equal(df, library.read('MYARR', date_range=DateRange(df.index[0])).data, check_freq=False)
+    assert_equal(df, library.read('MYARR', date_range=DateRange(df.index[0], df.index[-1])).data, check_freq=False)
+    assert_equal(df, library.read('MYARR', date_range=DateRange()).data, check_freq=False)
+    assert_equal(df[df.index[10]:], library.read('MYARR', date_range=DateRange(df.index[10])).data, check_freq=False)
+    assert_equal(df[:df.index[10]], library.read('MYARR', date_range=DateRange(end=df.index[10])).data, check_freq=False)
+    assert_equal(df[df.index[-1]:], library.read('MYARR', date_range=DateRange(df.index[-1])).data, check_freq=False)
+    assert_equal(df[df.index[-1]:], library.read('MYARR', date_range=DateRange(df.index[-1], df.index[-1])).data, check_freq=False)
+    assert_equal(df[df.index[0]:df.index[0]], library.read('MYARR', date_range=DateRange(df.index[0], df.index[0])).data, check_freq=False)
+    assert_equal(df[:df.index[0]], library.read('MYARR', date_range=DateRange(end=df.index[0])).data, check_freq=False)
     assert_equal(df[df.index[0] - DateOffset(days=1):],
-                 library.read('MYARR', date_range=DateRange(df.index[0] - DateOffset(days=1))).data)
+                 library.read('MYARR', date_range=DateRange(df.index[0] - DateOffset(days=1))).data, check_freq=False)
     assert_equal(df[df.index[-1] + DateOffset(days=1):],
-                 library.read('MYARR', date_range=DateRange(df.index[-1] + DateOffset(days=1))).data)
+                 library.read('MYARR', date_range=DateRange(df.index[-1] + DateOffset(days=1))).data, check_freq=False)
     assert len(library.read('MYARR', date_range=DateRange(dt(1950, 1, 1), dt(1951, 1, 1))).data) == 0
     assert len(library.read('MYARR', date_range=DateRange(dt(2091, 1, 1), dt(2091, 1, 1))).data) == 0
 
diff --git a/tests/integration/store/test_version_store.py b/tests/integration/store/test_version_store.py
@@ -26,7 +26,7 @@
 from tests.unit.serialization.serialization_test_data import _mixed_test_data
 from ...util import read_str_as_pandas
 from ..test_utils import enable_profiling_for_library
-from tests.util import assert_frame_equal_
+from tests.util import assert_frame_equal_, assert_series_equal_
 
 ts1 = read_str_as_pandas("""         times | near
                    2012-09-08 17:06:11.040 |  1.0
@@ -1743,7 +1743,7 @@ def test_can_write_tz_aware_data_series(library, fw_pointers_cfg):
         # Arctic converts by default the data to UTC, convert back
         read_data = read_data.dt.tz_localize('UTC').dt.tz_convert(read_data.index.tzinfo)
         assert library._versions.find_one({'symbol': 'symTzSer'})['type'] == PandasSeriesStore.TYPE
-        assert_series_equal(myseries, read_data)
+        assert_series_equal_(myseries, read_data, check_freq=False)
 
 
 @pytest.mark.parametrize('write_cfg, read_cfg, append_cfg, reread_cfg', [
diff --git a/tests/integration/tickstore/test_toplevel.py b/tests/integration/tickstore/test_toplevel.py
@@ -4,6 +4,7 @@
 import pandas as pd
 import pytest
 from pandas.util.testing import assert_frame_equal
+from tests.util import assert_frame_equal_
 
 from arctic.date import DateRange, mktz
 from arctic.exceptions import NoDataFoundException, LibraryNotFoundException, OverlappingDataException
@@ -62,7 +63,7 @@ def test_should_return_data_when_date_range_falls_in_a_single_underlying_library
     tickstore_current.write('blah', df)
     res = toplevel_tickstore.read('blah', DateRange(start=dt(2010, 1, 1), end=dt(2010, 1, 6)), list('ABCD'))
 
-    assert_frame_equal(df, res.tz_convert(mktz('Europe/London')))
+    assert_frame_equal_(df, res.tz_convert(mktz('Europe/London')), check_freq=False)
 
 
 def test_should_return_data_when_date_range_spans_libraries(toplevel_tickstore, arctic):
@@ -98,7 +99,7 @@ def test_should_return_data_when_date_range_spans_libraries_even_if_one_returns_
     tickstore_2011.write('blah', df_11)
     res = toplevel_tickstore.read('blah', DateRange(start=dt(2010, 1, 2), end=dt(2011, 1, 4)), list('ABCD'))
     expected_df = df_10[1:]
-    assert_frame_equal(expected_df, res.tz_convert(mktz('Europe/London')))
+    assert_frame_equal_(expected_df, res.tz_convert(mktz('Europe/London')), check_freq=False)
 
 
 def test_should_add_underlying_library_where_none_exists(toplevel_tickstore, arctic):
@@ -194,10 +195,10 @@ def test_should_write_top_level_with_list_of_dicts(arctic):
     expected = pd.DataFrame(np.arange(57, dtype=np.float64), index=dates, columns=list('a'))
     toplevel_tickstore.write('blah', data)
     res = toplevel_tickstore.read('blah', DateRange(start=dt(2010, 12, 1), end=dt(2011, 2, 1)), columns=list('a'))
-    assert_frame_equal(expected, res.tz_convert(mktz('Europe/London')))
+    assert_frame_equal_(expected, res.tz_convert(mktz('Europe/London')), check_freq=False)
     lib2010 = arctic['FEED_2010.LEVEL1']
     res = lib2010.read('blah', DateRange(start=dt(2010, 12, 1), end=dt(2011, 1, 1)))
-    assert_frame_equal(expected[dt(2010, 12, 1): dt(2010, 12, 31)], res.tz_convert(mktz('Europe/London')))
+    assert_frame_equal_(expected[dt(2010, 12, 1): dt(2010, 12, 31)], res.tz_convert(mktz('Europe/London')), check_freq=False)
 
 
 def test_should_write_top_level_with_correct_timezone(arctic):
@@ -212,7 +213,7 @@ def test_should_write_top_level_with_correct_timezone(arctic):
     expected = pd.DataFrame(np.arange(len(dates), dtype=np.float64), index=dates.tz_convert(utc), columns=list('a'))
     toplevel_tickstore.write('blah', data)
     res = toplevel_tickstore.read('blah', DateRange(start=dt(2010, 1, 1), end=dt(2011, 12, 31)), columns=list('a')).tz_convert(utc)
-    assert_frame_equal(expected, res)
+    assert_frame_equal_(expected, res, check_freq=False)
     lib2010 = arctic['FEED_2010.LEVEL1']
     # Check that only one point was written into 2010 being 3am on 31st
     assert len(lib2010.read('blah', DateRange(start=dt(2010, 12, 1), end=dt(2011, 1, 1)))) == 1
diff --git a/tests/unit/store/test_pickle_store.py b/tests/unit/store/test_pickle_store.py
@@ -36,7 +36,7 @@ def test_write_object():
 
     # Python 3.8 onwards uses protocol 5 which cannot be unpickled in Python versions below that, so limiting
     # it to use a maximum of protocol 4 in Python which is understood by 3.4 onwards and is still fairly efficient.
-    # The min() used to allow lower versions to be used in py2 (which supported a max of 2)
+    # pickle version 4 is introduced with  python 3.4 and default with 3.8 onward
     pickle_protocol = min(4, pickle.HIGHEST_PROTOCOL)
     assert coll.update_one.call_args_list == [call({'sha': checksum('sentinel.symbol', {'segment': 0, 'data': Binary(compress(pickle.dumps(sentinel.item, pickle_protocol)))}),
                                                     'symbol': 'sentinel.symbol'},
diff --git a/tests/util.py b/tests/util.py
@@ -11,18 +11,25 @@
 import numpy as np
 import pandas
 from dateutil.rrule import rrule, DAILY
-from pandas.util.testing import assert_frame_equal
+from pandas.util.testing import assert_frame_equal, assert_series_equal
 
 
+# check_freq default True added in pandas 1.1.0
 def assert_frame_equal_(df1, df2, check_freq=True, check_names=True):
-    if pandas.__version__ > '1.0.3':
-        # had to add check_freq because pandas 1.1.5 has different freq metadata behaviour
+    if pandas.__version__ >= '1.1.0':
         assert_frame_equal(df1.sort_index(axis=1), df2.sort_index(axis=1), check_names=check_names, check_freq=check_freq)
     else:
-        # pandas 1.0.3
+        # python 3.x
         assert_frame_equal(df1.sort_index(axis=1), df2.sort_index(axis=1), check_names=check_names)
-    #else: # 0.22.0 python 2.7
-        #assert_frame_equal(df1, df2, check_names=check_names)
+    #else: # python 2.7
+        #assert_frame_equal(df1, df2)
+
+# check_freq default True added in pandas 1.1.0
+def assert_series_equal_(s1, s2, check_freq=True):
+    if pandas.__version__ >= '1.1.0':
+        assert_series_equal(s1, s2, check_freq=check_freq)
+    else:
+        assert_series_equal(s1, s2)
 
 
 def dt_or_str_parser(string):
